# Comfy: full site text
> Every English page on comfy.org as markdown, in one file, except the supported-models directory, which has its own catalog at https://comfy.org/p/supported-models/llms.txt. The curated index is https://comfy.org/llms.txt; each page below starts with its canonical URL.
# Comfy - Professional Control of Visual AI
> Comfy is the AI creation engine for visual professionals who demand control over every model, every parameter, and every output.
[Get started for free](https://cloud.comfy.org/?utm_source=comfy_org&utm_medium=website&utm_campaign=free_tier&utm_content=hero_get_started_free)
3D ANGLE
Horizontal angle Vertical angle Zoom
COLOR
HUE
SATURATION
 OUTPUT
# Professional Control of Visual AI
3D ANGLE
Horizontal angle Vertical angle Zoom
COLOR
HUE
SATURATION
OUTPUT
[Get started for free](https://cloud.comfy.org/?utm_source=comfy_org&utm_medium=website&utm_campaign=free_tier&utm_content=hero_get_started_free)
## DEVELOPER PLATFORM BETA
Deploy your ComfyUI workflow as a production API. Thousands of models on one platform.
[Get Started](https://comfy.org/platform) [Read the docs](https://docs.comfy.org/development/overview)
## MiniMax Commercial License
Comfy is the only official reseller of MiniMax commercial-use licenses, for running MiniMax models locally on your own hardware. Full commercial rights to your outputs, fine-tuning and LoRA training, and client and downstream work included.
[REQUEST LICENSE](https://comfy.org/contact) [TRY H3 ON COMFY CLOUD](https://cloud.comfy.org/?share=a781503cf508)
Only official reseller Audio and music included Available globally
FEATURED · STAFF PICK
## Product Advertisement Video
 Rob Product Video
FEATURED · STAFF PICK
## MiniMax H3: Image to Video
 ComfyUI Video Image to Video
FEATURED · STAFF PICK
## MiniMax H3: Reference to Video
ComfyUI Video Reference to Video
FEATURED · STAFF PICK
## MiniMax H3: Text to Video
ComfyUI Text to Video Video
FEATURED · STAFF PICK
## Topaz: Image Enhance Bloom 2
ComfyUI API Image Upscale
FEATURED · STAFF PICK
## FLUX 3 Video: Text to Video
ComfyUI Partner Nodes Text to Video
HOW WORKS
Connect models, processing steps, and outputs on a canvas where every decision is visible and every step is inspectable.
Start from a community template or build from scratch.
Industries
Powered by 60,000+ nodes, thousands of workflows, and a community that builds faster than any one company could.
[EXPLORE WORKFLOWS](https://comfy.org/workflows/)
## Get started in minutes
Go as deep as you want to.
1
### Download or Sign up
[Download Desktop](https://comfy.org/download) or [Try Cloud for free](https://cloud.comfy.org/?utm_source=comfy_org&utm_medium=website&utm_campaign=free_tier&utm_content=getstarted_try_cloud) .
2
### Load a workflow
Start from [a community template](https://comfy.org/workflows/) or build your own.
3
### Generate & Iterate
Run, adjust, scale when you're ready.
## The AI creation engine for complete control
Over every model, every node, every step, every output.
[Comfy Desktop
Run ComfyUI on your own hardware.
SEE DESKTOP FEATURES](https://comfy.org/download) [Comfy Cloud
The full power of ComfyUI from anywhere.
SEE CLOUD FEATURES](https://comfy.org/cloud) [Developer Platform
Deploy and scale ComfyUI workflows.
SEE PLATFORM FEATURES](https://comfy.org/platform) [Comfy Enterprise
Enterprise-grade infrastructure for the creative engine inside your organization.
SEE ENTERPRISE FEATURES](https://comfy.org/enterprise)
Customer Stories
## See ComfyUI in the real world
Videos & case studies from teams building with ComfyUI
[WATCH STORY](https://comfy.org/customers/videos/black-math) [SEE ALL CASE STUDIES](https://comfy.org/customers)
New Model Release
## Seedance 2.5
ByteDance's cinematic video model — multi-shot sequences with native audio, text or image in. You direct on the canvas; Seedance renders the cut.
[Explore Seedance 2.5](https://comfy.org/seedance-2.5) [Try Workflow](https://cloud.comfy.org/?template=api_seedance2_5_r2v)
Partner Nodes
---
# About Us - Comfy
> Comfy builds ComfyUI, the open-source engine for visual AI, and the desktop, cloud, API, and enterprise products around it. Meet the team and the mission.
ABOUT
# Build the tools that reward creative skill
The team behind Comfy is small, intense, and building what we intend to be our life’s work.
[SEE OPEN ROLES](https://comfy.org/careers)
00:00
OUR STORY
## An open-source project that became the professional standard for visual AI.
In January, 2023, no tool on the market could chain two AI models together into a repeatable workflow. One developer in Quebec City built his own in two weeks and open-sourced it.
OUR
INVESTORS
If it can be open, it should be open. That’s how we build an ecosystem that moves faster than any company could think for, and it’s how we think about code, content, and skins. We make money by building things worth paying for, not by controlling what we believe should be free.
‘Comfy has innovated a new and powerful ecosystem for creativity without compromising creative control. It has been amazing to watch technical artists and curious creative minds leverage Comfy to explore the full surface area of their ideas.’
Scott Belsky, Founder of Behance
OUR VALUES
## We believe everyone deserves to design exactly what they dream, own their process, and never compromise because of their software.
SHIP IT
Work hard for what makes it into the world. Ideas don’t substitute for output.
SHARE IT
Every decision and its reasoning gets shared with the team.
OPEN-SOURCE IT
If it can be open, it should be open. That’s how we build an ecosystem that moves faster than any company could imagine.
RESPECT THE CRAFT
We build for people who prefer precision over shortcuts. The learning curve is high but the payoff is worth it.
SHIP IT
Work hard for what makes it into the world. Ideas don’t substitute for output.
SHARE IT
Every decision and its reasoning gets shared with the team.
OPEN-SOURCE IT
If it can be open, it should be open. That’s how we build an ecosystem that moves faster than any company could imagine.
RESPECT THE CRAFT
We build for people who prefer precision over shortcuts. The learning curve is high but the payoff is worth it.
Why join
?
IF YOU...
Want to build tools that empower others to create.
Want your output to matter more than your title.
Want high agency in a flat organization.
Would rather ship and learn than plan and wait.
Why join
?
IF YOU...
Want to build tools that empower others to create.
Want your output to matter more than your title.
Want high agency in a flat organization.
Would rather ship and learn than plan and wait.
CAREERS
## Join the team
[SEE OPEN ROLES](https://comfy.org/careers)
Don’t see your dream role here? hiring@comfy.org
---
# Comfy.org Affiliate Program - Become a Partner
> Earn 30% recurring commission for 3 months on every Comfy Cloud subscription you refer. Apply to become a Comfy Partner.
AFFILIATE
# Earn 30% recurring commission for 3 months.
- 30% recurring commission for 3 months
- 60-day cookie window
- $100 minimum payout
- Monthly payouts
[APPLY NOW](https://forms.gle/RS8L2ttcuGap4Q1v6)
## How it works
1. Apply
Submit a quick form. Most applicants approved same day.
2. Share
Get your unique tracking link. Share via content, social, email, however you reach your audience.
3. Earn
30% recurring commission for 3 months on every Comfy Cloud subscriber you refer. Tracked in real-time. Paid monthly.
## Who we're looking for
### If you are...
- A ComfyUI tutorial creator or workflow builder
- An AI tool reviewer on YouTube, TikTok, blogs
- A tech blogger covering AI creative tools
- A newsletter operator in the AI/creative space
- Anyone with an audience interested in AI image, video, or 3D
## Why ComfyUI for affiliate creators
01
ComfyUI is the most powerful open-source AI creative platform
02
Comfy Cloud lets you run ComfyUI in the browser, no GPU needed, all models pre-loaded
03
Node-based workflows give users full creative control unlike prompt-only tools
04
1,000+ community custom node packages
[APPLY NOW](https://forms.gle/RS8L2ttcuGap4Q1v6)
## Brand logos for your content
Banners, screenshots, and talking points are in your affiliate dashboard after approval. Get the [full brand assets here](https://comfy.org/brand).
- ### Core Logo
- ### Logomark
- ### Icon
- 
### Amplified Logomark
[Go to brand portal](https://comfy.org/brand)
## Frequently asked questions
Real-time dashboard via our partner portal.
All Comfy Cloud paid subscription plans (Standard, Creator, Pro, Teams).
Most applications approved within 24 hours.
Monthly, within the first 10 business days. Minimum balance $100. Paid via Stripe Express or PayPal.
If they upgrade, your commission increases. If they downgrade, it adjusts accordingly. Commission is based on actual amounts received by Comfy.org, net of refunds.
Yes. We support both tracking links and unique coupon codes.
We use server-side tracking, so conversions are tracked regardless.
Logos and banners on this page, plus screenshots and talking points in your affiliate dashboard after approval.
## Ready to start earning?
[APPLY NOW](https://forms.gle/RS8L2ttcuGap4Q1v6)
[Read the affiliate program terms](https://comfy.org/affiliates/terms)
---
# Affiliate Terms - Comfy
> Comfy.org Affiliate Program Terms and Conditions.
Effective Date: May 16, 2026
On this page
- 1. Program Overview
- 2. Eligible Products
- 3. Commission Structure
- 4. Attribution Rules
- 5. Prohibited Activities
- 6. Content Guidelines and Intellectual Property Rights
- 7. Termination
- 8. Program Modifications
- 9. Indemnification and Limitation of Liability
- 10. Governing Law
- 11. Miscellaneous
## 1. Program Overview
The [Comfy.org](https://comfy.org/) Affiliate Program (" **Program**") allows approved participants (" **Affiliates**") to earn commissions by referring new paying customers to Comfy Cloud. By participating in this program, you agree to these terms.
## 2. Eligible Products
Commissions are earned on Comfy Cloud paid subscription plans only. The following are excluded from commission eligibility: free tier signups (unless they later convert to paid), one-time credit purchases, enterprise contracts negotiated directly with Comfy sales, and API-only usage billed outside of standard subscription plans.
## 3. Commission Structure
- Commission rate: 30% recurring on the net subscription amount received by Comfy.org
- Commission duration: 3 months from the referred customer’s first paid subscription
- Cookie/attribution window: 60 days from the referral click
- Minimum payout threshold: $100
- Payout schedule: Monthly, within the first 10 business days of each month after the receipt of applicable payment by Comfy from its referred customer
- Payout method: Via the affiliate tracking platform (Stripe Express or PayPal)
- Commission cessation: To the extent a referred customer’s subscription is canceled, in whole or in part, the affiliate shall correspondingly cease to receive commission payments, even within the 3-month commission window. Refunded or charged-back transactions are not eligible for commission, and any commission previously paid for such transactions will be deducted from future payouts (see Section 4).
## 4. Attribution Rules
- Commissions are attributed on a last-click basis within the 60-day cookie window
- If a referred customer cancels and re-subscribes within 60 days, the original affiliate retains attribution
- If a referred customer upgrades their plan, commission is calculated on the upgraded amount
- If a referred customer downgrades their plan, commission adjusts to the new plan amount
- Refunded transactions are not eligible for commission
- Any commission paid on refunded transactions will be deducted from future payouts to you
## 5. Prohibited Activities
Affiliates must NOT:
- **Self-refer**: Use your own affiliate link to purchase or receive discounts on your own account
- **Bid on branded keywords**: Run paid search campaigns (Google Ads, Bing Ads, etc.) targeting "ComfyUI," "Comfy.org," "Comfy Cloud," or any misspellings or variations thereof
- **Misrepresent**: Impersonate Comfy.org, claim to be an employee, or create assets that could be confused with official Comfy.org materials
- **Spam**: Send unsolicited bulk emails, messages, or engage in any form of spam promotion
- **Cookie stuff**: Use hidden iframes, pop-unders, or any technical means to set cookies without genuine user intent
- **Incentivize clicks**: Offer monetary rewards, points, or other incentives solely for clicking your affiliate link (content recommendations are fine)
- **Use misleading claims**: Make false or exaggerated claims about Comfy.org products, pricing, or features
- **Promote on prohibited content**: Place affiliate links on sites containing illegal content, hate speech, or adult content
- **Contrary to laws**: Place affiliate links in any market that is prohibited as a region under the laws of the United States of America.
## 6. Content Guidelines and Intellectual Property Rights
- Affiliates must clearly disclose the affiliate relationship in accordance with FTC guidelines (US) and equivalent regulations in their jurisdiction
- Recommended disclosure: "This page contains affiliate links. I may earn a commission if you sign up through my link."
- Affiliates may use Comfy.org logos and brand assets only as provided in the official affiliate asset kit, and may not modify them
- Comfy.org retains all rights, including in any of its intellectual property apart from the limited use rights granted herein
## 7. Termination
- Either party may terminate the affiliate relationship at any time with 14 days’ prior written notice
- Comfy.org reserves the right to immediately terminate and withhold commissions if an affiliate violates any of the prohibited activities listed above
- Upon termination, any unpaid commissions above the minimum threshold will be paid in the next regular payout cycle
- Commissions on referred customers will cease at the time of termination, even if within the 3-month commission window
## 8. Program Modifications
- Comfy.org reserves the right to modify these terms, commission rates, or program structure with 30 days notice to active affiliates
- Continued participation after notice constitutes acceptance of modified terms
## 9. Indemnification and Limitation of Liability
- You will indemnify Comfy.org from any third party claim arising out of your breach of these terms.
- Comfy.org’s liability to any affiliate shall not (i) exceed the total commissions paid to that affiliate in the preceding 12 months, and (ii) include any indirect, consequential, punitive or any other type of special damages.
- Comfy.org is not responsible for tracking failures caused by user browser settings, ad blockers, or VPNs, though we employ server-side tracking to minimize these issues
## 10. Governing Law
These terms shall be governed by and construed in accordance with the laws of the State of Delaware, without regard to conflict of law principles. All disputes arising under this Agreement shall be resolved exclusively in the state or federal courts in the State of Delaware.
## 11. Miscellaneous
**(a) Entire Agreement.** These terms constitutes the sole and entire agreement (including the attached schedules and exhibits) of the Parties with respect to the subject matter of these terms, and supersedes all prior and contemporaneous understandings, agreements, representations, and warranties, both written and oral, with respect to the subject matter. **(b) Relationship of Parties.** Each party is an independent contractor with regard to these terms. Nothing contained in These terms shall be construed as creating any agency, partnership, joint venture, or other form of joint enterprise, employment, or fiduciary relationship between the Parties. Neither party, by virtue of these terms, will have any right, power, or authority to act or create an obligation, express or implied, on behalf of the other party. **(c) Assignment.** Neither party shall assign any of its rights or delegate any of its obligations hereunder without the prior written consent of the other party, which consent shall not be unreasonably withheld, conditioned or delayed. **(d) Severability.** If any term or provision of these terms is invalid, illegal, or unenforceable in any jurisdiction, such invalidity, illegality, or unenforceability shall not affect any other term or provision of these terms or invalidate or render unenforceable such term or provision in any other jurisdiction. Upon a determination that any term or provision is invalid, illegal or unenforceable, the Parties hereto shall negotiate in good faith to modify these terms to effect the original intent of the Parties as closely as possible in order that the transactions contemplated hereby be consummated as originally contemplated to the greatest extent possible. **(e) Waiver.** No waiver by either party of any of the provisions hereof shall be effective unless explicitly set forth in writing and signed by the party so waiving. **(f) Notice.** Each party shall deliver all notices, requests, consents, claims, demands, waivers, and other communications under these terms in writing to the email utilized for the primary contact for the other party. **(g) Cumulative Remedies.** All rights and remedies provided in these terms are cumulative and not exclusive, and the exercise by a party of any right or remedy does not preclude the exercise of any other rights or remedies that may now or subsequently be available at Law, in equity, by statute, in any other agreement between the Parties or otherwise. **(h) No Third-Party Beneficiaries.** These terms benefits solely the Parties to these terms and their respective permitted successors and assigns and nothing in these terms, express or implied, confers on any other person or entity any legal or equitable right, benefit, or remedy of any nature whatsoever under or by reason of these terms.
---
# Brand — Comfy
> The Comfy brand portal: logos, color, typography, and voice. Everything you need to build something that looks and sounds like Comfy.
Brand Portal
# Create with ComfyUI
Logo, color, type, and voice. Everything you need to build something that looks and sounds like us.
[View brand guidelines](https://drive.google.com/file/d/1EDt03JTfF_nbbY_H2n67aaUj6k11v3bS/view) [Download logos](https://media.comfy.org/website/comfy-org-brand-assets.zip)
## One mark, many dimensions.
Logos come in light and dark options. Use as provided. Do not distort, recolor, or outline. Make sure the logo is legible against its background.
- Core Logo
- Logomark
- Icon
- 
Amplified Logomark
## Every color earns its place.
Our color palette helps build brand recognition. When people think of Comfy, we want them to associate it with the following colors.
- Comfy Yellow
**hex**
**rgb**
**hsl**
**cmyk**
- Comfy Ink
**hex**
**rgb**
**hsl**
**cmyk**
- Comfy Canvas
**hex**
**rgb**
**hsl**
**cmyk**
- Comfy Plum
**hex**
**rgb**
**hsl**
**cmyk**
- Warm White
**hex**
**rgb**
**hsl**
**cmyk**
- Warm Gray
**hex**
**rgb**
**hsl**
**cmyk**
- Cool Gray
**hex**
**rgb**
**hsl**
**cmyk**
- Mauve
**hex**
**rgb**
**hsl**
**cmyk**
## Precise, never cute.
**Direct**
We state things. We don’t hedge, qualify, or suggest. Short sentences. Active voice. One idea at a time.
**Precise**
We use the real names for things. Nodes, samplers, seeds, checkpoints. We don’t talk around the product or reach for metaphor when the technical term is already good.
**Human-first**
The human creates. Comfy makes every step visible. We never write as though the AI is doing the work.
**Anti-hype**
We don’t write “stunning,” “revolutionary,” or “effortless.” We don’t promise magic. Our tagline says exactly what we mean: Method, not magic.
Do
Route your prompt through a ControlNet. Wire the output to the VAE decode.
Comfy runs on your hardware. Nothing leaves your machine.
Don’t
Simply connect your AI blocks and watch the magic happen!
Oops! Something went wrong. Please try again later.
## Trademark guidelines.
Comfy and ComfyUI are trademarks of Comfy Org. You’re welcome to reference them in content that accurately describes your work with our platform. Tutorials, reviews, integrations, and affiliate content all qualify.
A few rules: don’t modify the logo, don’t use the Comfy name in your own product or company name, and don’t present your content in a way that implies official endorsement or partnership beyond what’s been agreed.
For permissions outside these guidelines, [Contact Us](https://support.comfy.org/hc/en-us)
## Questions?
For press, partnerships, or anything outside these guidelines, [Contact Us](https://support.comfy.org/hc/en-us)
---
# Careers - Comfy
> Join the team building the operating system for generative AI. Open roles in engineering, design, marketing, and more.
CAREERS
# Building an operating system for Gen AI
We're building the world's leading visual AI platform — open, modular, and designed for those who want control, quality and scale in their creative process.
From solo creators to enterprise teams, millions of people rely on ComfyUI to push the boundaries of what creative AI can do.
We're a team of builders. We raise the bar for each other, move fast, and care deeply about the open-source community that's grown alongside us.
If you want to shape the future of creativity, come build with us.
## Roles
ENGINEERING [Member of Technical Staff, Frontend
Engineering San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/c3e0584d-5490-491f-aae4-b5922ef63fd2) [Member of Technical Staff, Backend
Engineering San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/732f8b39-076d-4847-afe3-f54d4451607e) [Product Manager, ComfyUI
Engineering San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/9e4b9029-c3e9-436b-82c4-a1a9f1b8c16e) [Developer Relations
Engineering San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/3e0ec22c-758d-4a47-aebe-ac4583ae3d8a) [Member of Technical Staff, Release Engineering
Engineering San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/bbcd32a3-27fb-40f7-94ff-8ec1166b5694)
SALES [Senior Data Scientist, Product & Growth
Sales San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/57ff81f6-7df4-4f9d-be21-ce55f348bdad) [Forward Deployed Creative
Sales San Francisco
San Francisco](https://jobs.ashbyhq.com/comfy-org/b8faf3c0-a21c-4bed-8651-93daa6bfe81c)
Why join
?
IF YOU...
Want to build tools that empower others to create.
Like working on foundational tech that's already powering real-world videos, images, music, and apps.
Care about open, free alternatives to closed AI platforms.
Believe artists, hackers, and solo builders should have real control over their tools.
Want to work in a small, sharp, no-BS team that moves fast and ships often.
Why join
?
IF YOU...
Want to build tools that empower others to create.
Like working on foundational tech that's already powering real-world videos, images, music, and apps.
Care about open, free alternatives to closed AI platforms.
Believe artists, hackers, and solo builders should have real control over their tools.
Want to work in a small, sharp, no-BS team that moves fast and ships often.
## Q&A
We're a small, intense team that values ownership, speed, and craft. Everyone ships real work, helps each other raise the bar, and cares deeply about the open-source community we serve. We move fast, communicate directly, and trust each other to do great work.
We care about what you can do, not where you went to school. Show us your work — side projects, open-source contributions, or anything that demonstrates your skills and taste. We value builders who ship.
Click on any open role above to apply through our job board. Include your resume and any relevant links to your work (GitHub, portfolio, etc.).
Typically: application review → intro call → technical assessment or work sample → team interviews → offer. We try to move quickly and respect your time throughout the process.
Most roles are based in San Francisco, but we have team members in other locations too. Each role listing specifies the location. We do regular team offsites to stay connected.
Use ComfyUI. Build something with it. Contribute to the open-source project. The best candidates are the ones who already understand the product and community because they're part of it.
We can sponsor visas for exceptional candidates. Let us know your situation in your application and we'll work through the details together.
We do our best to provide meaningful feedback to candidates who go through our interview process. Due to volume, we may not be able to provide detailed feedback at the application stage.
---
# ChatGPT Images 2.5 on Comfy: Generate and Edit Images
> Run ChatGPT Images 2.5 in ComfyUI. Generate detailed images from text, edit existing visuals, and direct results with reference images in a repeatable workflow.
Turn a plain language brief into a finished image, or revise an existing visual without rebuilding it from scratch. ChatGPT Images 2.5 brings generation, reference guided creation, and precise editing into ComfyUI workflows through Partner Nodes.
[RUN CHATGPT IMAGES 2.5](https://cloud.comfy.org/?template=api_openai_gpt_image_25_flare_t2i&utm_source=comfy.org&utm_medium=referral&utm_campaign=gpt-image-2-5) [READ THE GUIDE](https://docs.comfy.org/tutorials/partner-nodes/openai/gpt-image-2-5)
Text to Image Image Editing Reference Images
## Made with ChatGPT Images 2.5
Premium Pay-as-you-go
A polished vaporwave world with crisp typography, reflective marble, and neon light.
Premium Pay-as-you-go
A deadpan security-camera frame of four sharply dressed visitors from another world.
Premium Pay-as-you-go
A curious goldfish meets an extreme fisheye lens in a sunlit field.
Premium Pay-as-you-go
A roaring V8 rendered as an expressive watercolor study in heat and motion.
Premium Pay-as-you-go
A cinematic canyon pursuit freezes at the instant a muscle car clears the gap.
Premium Pay-as-you-go
A solitary hero watches a red sun rise over a restrained anime landscape.
## Choose a plan
Access cloud-powered ComfyUI workflows with straightforward, usage-based pricing.
Start Comfy Cloud for free. Upgrade when you're ready.
5 free runs on real GPUs — no credit card required.
[TRY FREE](https://cloud.comfy.org/?template=api_openai_gpt_image_25_flare_t2i&utm_source=comfy.org&utm_medium=referral&utm_campaign=gpt-image-2-5)
STANDARD
$20
/month
Billed monthly
- Included: 30 minute max workflow runtime
- Included: Add more credits anytime
- Not included: Import your own models
- Not included: Longer workflow runtime (up to 1 hr)
4,200 monthly credits
Generates ~380 5s videos*
[SUBSCRIBE TO STANDARD](https://cloud.comfy.org/?pricing=standard&cycle=monthly)
CREATOR MOST POPULAR
$35
/month
Billed monthly
- Included: 30 minute max workflow runtime
- Included: Add more credits anytime
- Included: Import your own models
- Not included: Longer workflow runtime (up to 1 hr)
7,400 monthly credits
Generates ~670 5s videos*
[SUBSCRIBE TO CREATOR](https://cloud.comfy.org/?pricing=creator&cycle=monthly)
PRO
$100
/month
Billed monthly
- Included: 30 minute max workflow runtime
- Included: Add more credits anytime
- Included: Import your own models
- Included: Longer workflow runtime (up to 1 hr)
21,100 monthly credits
Generates ~1,915 5s videos*
[SUBSCRIBE TO PRO](https://cloud.comfy.org/?pricing=pro&cycle=monthly)
TEAM
Built for teams collaborating on workflows together.
$665
$700 /month
Save 5% ($35)
Billed monthly
42.2K
84.4K
147.7K
295.4K
527.5K
147,700 monthly credits
Generates ~13,405 5s videos*
Everything in Pro, plus:
- Included: Invite members
- Included: Members can run workflows concurrently
- Included: Shared credit pool for all members
- Included: Role-based permissions
Coming soon...
- Coming soon: Shared workflows & assets
- Coming soon: Projects
[SUBSCRIBE TO TEAM](https://cloud.comfy.org/?pricing=team&stop=team_700&cycle=monthly)
ENTERPRISE
Need more members? Looking for more flexibility or custom features?
[Contact Us](https://comfy.org/contact)
*Based on 5s videos created with the Wan 2.2 Image-to-Video template using default settings (81 frames, 18fps, 640x640, 4-step sampler)
## Q&A
ChatGPT Images 2.5 is an OpenAI image model for generating new visuals and editing existing images from natural-language instructions. In ComfyUI, it runs through Partner Nodes and can be combined with the rest of your workflow.
Open [Comfy Cloud](https://cloud.comfy.org/?template=api_openai_gpt_image_25_flare_t2i&utm_source=comfy.org&utm_medium=referral&utm_campaign=gpt-image-2-5), add the OpenAI image Partner Node to a workflow, enter a prompt, and connect any reference images you want to use.
Yes. Supply an image and describe the change you want, from replacing objects and restyling a scene to focused masked edits. You can keep the result in the same graph for further processing.
Yes. Reference images can guide the subject, composition, materials, or visual language of a generation. Describe which qualities should carry into the result.
Write a clear creative brief: name the subject, setting, composition, lighting, medium, and any text that must appear. For edits, state what should change and what must stay untouched.
The generated image becomes one step in a larger, repeatable pipeline. Route it into masking, compositing, upscaling, animation, or another model while keeping every step visible and adjustable.
## One engine, every way to run it
Build workflows in the browser today. Batch campaigns with the API, or connect image generation to the rest of your creative pipeline.
[Comfy Desktop
Run ComfyUI on your own hardware.
LEARN MORE](https://comfy.org/download) [Comfy Cloud
The full power of ComfyUI from anywhere.
LEARN MORE](https://comfy.org/cloud) [Developer Platform
Deploy and scale ComfyUI workflows.
LEARN MORE](https://comfy.org/platform) [Comfy Enterprise
Enterprise-grade infrastructure for the creative engine inside your organization.
LEARN MORE](https://comfy.org/enterprise)
### Comfy MCP: now turn your agent into a creative technologist.
Your AI assistant can access the ecosystem, build workflows, and generate images, video, audio, or 3D.
[GET STARTED](https://comfy.org/mcp)
## 4+ million Comfy creators say
"Comfy has innovated a new and powerful ecosystem for creativity without compromising creative control. It has been amazing to watch technical artists and curious creative minds leverage Comfy to explore the full surface area of their ideas."
Scott Belsky,
Founder of Behance
"The best part for me is the node-based workflow: it offers a lot of possibilities and enables many different combinations. I also appreciate the optimization options and the automatic memory handling, even when working with tight models."
Richard N.,
Owner
"I appreciate being able to use models that my computer can no longer keep up with. I prefer the node-based approach because it is adjustable, rather than being behind a closed door."
MaryAnn E.,
Broker at Pop RV's
"The UI and UX of ComfyUI are intuitive and user-friendly, making navigation straightforward and efficient. Its integrations with other tools streamline workflows and enhance overall productivity."
Alan M.,
Co-Founder
"I use ComfyUI to help me create VFX for my animated show. It's easy to use, especially with importing a Comfy pipeline on the cloud, and I really like the clean UI and how straightforward it is."
Kirk H.
"I love that ComfyUI is extremely fast and very reliable. It gives me so much freedom to create the characters I want."
Fred C.
"The one-click workflows are something I do love, as they simplify many processes for me. The setup was very easy, even for someone who isn't familiar with it all."
Matthew P.
"I like how in-depth ComfyUI can be. It makes me feel like the software itself isn't limited, which encourages me to keep experimenting and learning."
Verified User
"A holistic approach and visualisation through an endless whiteboard."
Nikolai K.,
3D Artist
"Having full control of the GenAI process, and being able to use it unlimited times, integrating it into your workflow whenever you need to create."
Leonardo S.,
Social Media Marketing Expert
"Helps me run AI models locally and generate images, video, and audio, all free of cost."
Verified User,
Computer Software
---
# Comfy CLI - Drive ComfyUI from your terminal and any coding agent
> The command line for the full ComfyUI engine. Generate images, video, audio, and 3D from Claude Code, Codex, Cursor, or any shell, on Comfy Cloud or your own GPU.
CLI
# Drive ComfyUI from your terminal. Or your agent's.
One CLI for the whole engine: generate images, video, audio, and 3D, run real workflow graphs, and pipe every output where you need it. Built for coding agents, CI jobs, and anyone with a terminal.
INSTALL CLI [VIEW DOCS](https://docs.comfy.org/agent-tools/cli)
comfy — zsh
$
WORKS WITH
Claude Code
Codex
Cursor
Gemini CLI
OpenClaw
Hermes
Any shell
GET STARTED
## Set up Comfy CLI
One install, two places to run: Comfy Cloud GPUs, or the ComfyUI on your own machine. Paste one prompt into your coding agent, or run the commands yourself.
Cloud runs spend Comfy credits. Add credits, or use an existing [Comfy Cloud subscription](https://comfy.org/pricing). Browsing templates, models, and nodes is free with a Comfy account.
### Install the CLI
Install from PyPI (Homebrew works too), then set up the cloud connection. It opens a browser sign-in.
pip install comfy-cli
Prefer the manual path? After the install above:
comfy setup --where cloud comfy skills install
### Ask Claude Code to set it up
Paste this into Claude Code. It installs the CLI, runs setup, signs you in, and adds the bundled skills.
Set up the Comfy CLI for me: pip install comfy-cli, then comfy setup --where cloud, then comfy skills install. Follow the guide at https://docs.comfy.org/agent-tools/cli.md and tell me when it's ready.
comfy generate is in beta: flag names, model aliases, and output formats may change. The underlying partner endpoints are stable.
Every command and flag lives in the [command reference](https://docs.comfy.org/comfy-cli/reference).
## A real session, start to finish.
Recorded in one take against Comfy Cloud: a generation with a partner model, a workflow run on cloud GPUs, and the bundled skills. Idle seconds trimmed, output untouched.
Video transcript: comfy setup connects the CLI to Comfy Cloud. comfy generate renders a prompt with a partner model and saves cat.png. comfy run submits a workflow to cloud GPUs, comfy jobs wait tracks it, and comfy download saves the result as f5308bc5_000.png. comfy skills list prints the bundled skills. The session ends with two files saved: cat.png and f5308bc5_000.png.
AGENT SKILLS
## Skills your agent already knows.
One command installs five bundled skills that teach Claude Code, Cursor, and any AGENTS.md-aware tool the whole CLI. Below are six production briefs. Every prompt is ready to paste.
comfy skills install
BUNDLED SKILLS
- comfy workflows, models, and generation from the shell
- comfy-fragments typed fragments and YAML recipes for big workflows
- comfy-debug from failed run to fixed, fast
- comfy-relay surfaces every workflow artifact in chat
- comfy-director narrative, multi-shot video direction
### BATCH PACK SHOTS
comfy-fragments
Run our pack-shot workflow for every SKU in ./skus.csv, 4 angles each, route it to cloud, and download everything to ./shots.
Run our pack-shot workflow for every SKU in ./skus.csv, 4 angles each, route it to cloud, and download everything to ./shots.
### KEYFRAME BOARDS
comfy-director
Board a 30-second spot from BRIEF.md: 8 keyframes, consistent hero character, 16:9, contact sheet at the end.
Board a 30-second spot from BRIEF.md: 8 keyframes, consistent hero character, 16:9, contact sheet at the end.
### CHARACTER SHEETS
comfy
Load our saved character workflow and generate a 4-view turnaround for each hero in ./characters.
Load our saved character workflow and generate a 4-view turnaround for each hero in ./characters.
### SET EXTENSION
comfy
Extend this plate to a 21:9 crop, matte-painting style. Keep the horizon line and the light direction.
Extend this plate to a 21:9 crop, matte-painting style. Keep the horizon line and the light direction.
### 3D HERO PROPS
comfy
Build a 3D hero prop from sketch.png, game-ready, and hand me the .glb.
Build a 3D hero prop from sketch.png, game-ready, and hand me the .glb.
### CAMPAIGN KEY ART
comfy
Match frame_04.png's palette and produce campaign key art in 1:1, 4:5, and 16:9.
Match frame_04.png's palette and produce campaign key art in 1:1, 4:5, and 16:9.
More agent skills live in the [Comfy Skills repository](https://github.com/Comfy-Org/comfy-skills).
AI MODELS
## Run the world’s leading AI models
New models are added as they launch.
[Seedance 2.5](https://comfy.org/workflows/) [MiniMax H3](https://comfy.org/workflows/) [Gemini Omni 1.1 Flash](https://comfy.org/workflows/) [Wan 3.0](https://comfy.org/workflows/) [Wan 2.2](https://comfy.org/workflows/) [GPT Image 2](https://comfy.org/workflows/)
[EXPLORE WORKFLOWS](https://comfy.org/workflows/)
THE AGENT CONTRACT
## Built for agents, not adapted for them.
The contract lives in the CLI itself. No wrapper, no guesswork.
--json
### Structured output
Every command accepts --json and returns one envelope shape: ok, data, error, and a hint your agent can act on. Pipe it into anything that reads JSON.
comfy --json discover
### Self-describing
One call dumps the entire command tree, output schemas, and error codes. Your agent reads the contract instead of guessing flags.
comfy run --workflow
### Real workflow graphs
Not one-shot endpoints. A run is a workflow JSON your agent can edit, validate, rerun, and hand to a teammate. It's the same graph that opens on the ComfyUI canvas.
## Why build on Comfy CLI?
The terminal counterpart to Comfy MCP: the same engine and the same account, scripted.
### Open source, end to end.
comfy-cli is open source, and so is the engine it drives. Install from PyPI, read the source when you need to, and never build on a black box.
### The full engine, not a sandbox.
Multi-step workflow graphs, thousands of community nodes, and partner models through the same commands. A creation engine, not a menu of hosted endpoints.
### Reproducible by design.
A run is a workflow file plus a seed. Check it into git, diff it in review, replay it in CI, and get the same output every time.
### Your GPUs, or ours.
Every command auto-detects where to run: signed in, jobs route to Comfy Cloud GPUs; otherwise they hit the ComfyUI on your machine. Override per call with --where.
### Works alongside MCP.
[Set up Comfy MCP](https://comfy.org/mcp)
One account, two surfaces. Chat through Comfy MCP, script through the CLI, and let your agent pick the right one per task. Cloud outputs land in the same asset library either way.
## How it works
01
### INSTALL
pip install comfy-cli, then comfy setup. Sign in with OAuth for cloud GPUs, and comfy skills install teaches your agent the rest.
02
### DISCOVER
comfy --json discover hands your agent the full command tree; templates, models, and nodes are one search away.
03
### CREATE
Queue a run, wait on the job, download the outputs. Or pipe all three into a single line.
INSTALL CLI [VIEW DOCS](https://docs.comfy.org/agent-tools/cli)
Products
## The AI creation engine for complete control
Over every model, every node, every step, every output.
[Comfy Desktop
Run ComfyUI on your own hardware.
SEE DESKTOP FEATURES](https://comfy.org/download) [Comfy Cloud
The full power of ComfyUI from anywhere.
SEE CLOUD FEATURES](https://comfy.org/cloud) [Developer Platform
Deploy and scale ComfyUI workflows.
SEE PLATFORM FEATURES](https://comfy.org/platform) [Comfy Enterprise
Enterprise-grade infrastructure for the creative engine inside your organization.
SEE ENTERPRISE FEATURES](https://comfy.org/enterprise)
## Q&As
Any coding agent that can run shell commands. comfy skills install adds bundled skills for Claude Code, Cursor, and any AGENTS.md-aware tool, and the paste-prompt setup works in Codex, Gemini CLI, OpenClaw, Hermes, and the rest. For humans: any terminal, with tab completion via comfy --install-completion.
Both drive the same engine with the same account. Comfy MCP is for interactive agent work in chat: results come back in the session. The CLI is for automation: batch runs, CI, shell pipelines, and outputs saved to disk with --download or emitted as --json. Many teams run both and let the agent pick per task. Comfy MCP lives at [https://comfy.org/mcp](https://comfy.org/mcp).
The CLI is free and open source. With a Comfy account you can browse templates, models, and nodes at no cost, and runs on your own hardware are free. Partner models are the exception: they run on partner infrastructure and spend Comfy credits. Cloud runs spend credits too, either added directly or through an existing [Comfy Cloud subscription](https://comfy.org/cloud/pricing).
Every command auto-detects where to run: signed in to Comfy Cloud, commands route to cloud; otherwise they run against your local server. Override per call with --where local or --where cloud, set the COMFY_WHERE environment variable, or persist a default with comfy set-default --where cloud.
Browser OAuth is the recommended path: comfy cloud login opens a sign-in, and the CLI refreshes tokens for you. API keys are the fallback for headless and CI use where no browser exists: create one at [https://platform.comfy.org/profile/api-keys](https://platform.comfy.org/profile/api-keys), then set COMFY_API_KEY or pass --api-key per call.
--download saves generate results to the path you name. comfy download pulls a run's outputs into ./outputs/, and it reads prompt IDs straight from piped stdin, so comfy run --workflow flux.json --wait | comfy download is the whole loop. On cloud, outputs also land in your Comfy Cloud asset library.
Yes. comfy-cli lives at [https://github.com/Comfy-Org/comfy-cli](https://github.com/Comfy-Org/comfy-cli) and installs from PyPI with pip install comfy-cli, or from Homebrew. The ComfyUI engine it drives is open source too.
comfy generate is in beta: flag names, model aliases, and output formats may change while we iterate. The underlying partner endpoints are stable. File feedback on the comfy-cli GitHub repo.
Yes, that's what it's for. comfy setup --where cloud --api-key comfyui-... --non-interactive configures a runner without a browser, every command takes --json, and comfy validate --workflow catches unknown nodes and missing models before you spend compute.
• Generate images, video, audio, and 3D through partner models like Flux, Seedance, Kling, and Ideogram • Build, edit, validate, and run workflow graphs; save and rerun them • Queue batches, wait on jobs, and download every output • Search templates, models, and nodes with full schemas • Compose multi-stage workflows from typed fragments Run comfy skills install once, and the bundled skills teach your agent all of it.
---
# Comfy Cloud Nodes: Run Open Models on Our GPUs, From Your ComfyUI
> Drop a Comfy Cloud node into a local graph and run open models on Comfy Cloud GPUs. No subscription, no downloads, billed per GPU-second in credits.
CLOUD NODES
BETA
# Your graph. Our GPUs.
Comfy Cloud nodes run the newest open models on our GPUs from inside your own ComfyUI. Drop one into the graph you already have and wire it like any other node.
- No subscription
- Nothing to download, no VRAM floor
- Billed per GPU-second in credits
[Get Started](https://comfy.org/download) [Read the Docs](https://docs.comfy.org/cloud-nodes/overview)
## Four steps to your first run
01
### Update ComfyUI
Update to v0.34.5 or later. The nodes are part of ComfyUI core, so there is nothing to install separately.
02
### Restart ComfyUI
The nodes only register at startup, so they will not appear until you restart.
03
### Find a node
Search the node library for “Comfy Cloud”.
04
### Wire it in and run
Wire it like any other node. The model runs on our GPUs and the result comes back into the same canvas.
[How to Update ComfyUI](https://docs.comfy.org/installation/update_comfyui)
## Eight nodes, six models
Image, video and audio at launch. Every image and clip below is a real output from the model on the card, generated on Comfy Cloud.
- 
### Flux 2
1 node
- ### MiniMax H3
3 nodes
- 
### Z-Image Turbo
1 node
- 
### Mage Flow
1 node
- 
### Mage Flow Turbo
1 node
- 
### MiniMax Music 3
1 node
The full node list and their inputs live in the docs. [See the node reference](https://docs.comfy.org/cloud-nodes/overview)
## What actually runs on our side
### One curated workflow per node, pinned to a fixed model version.
- Pinned versions, so a saved graph keeps its result.
- Only the node’s inputs leave your machine.
- Results return as a normal IMAGE, VIDEO or AUDIO output.
- Charged by execution time, not per generation.
## Maximum control. Offload only the compute.
Your graph, your nodes, your outputs. A Comfy Cloud node hands off a single model run and hands the result straight back.
### Your workflow stays yours
The graph you built, the nodes around it and the outputs you route onward all stay on your machine, exactly as they are.
### Only the compute moves
A Cloud node sends the inputs for one model and returns the result. Everything else in the workflow keeps running locally.
### Reach for any model
The newest open models run on our GPUs at full size, so what your hardware can do stops deciding what you can make.
### One node, not a migration
You add a node to the graph you already have. Nothing else about your setup changes.
Products
## The AI creation engine for complete control
Over every model, every node, every step, every output.
[Comfy Desktop
Run ComfyUI on your own hardware.
SEE DESKTOP FEATURES](https://comfy.org/download) [Comfy Cloud
The full power of ComfyUI from anywhere.
SEE CLOUD FEATURES](https://comfy.org/cloud) [Developer Platform
Deploy and scale ComfyUI workflows.
SEE PLATFORM FEATURES](https://comfy.org/platform) [Comfy Enterprise
Enterprise-grade infrastructure for the creative engine inside your organization.
SEE ENTERPRISE FEATURES](https://comfy.org/enterprise)
## Frequently asked questions
Because the curated set is still changing: nodes may gain or lose options, and a workflow may be retired. It is not a statement about reliability, runs are real and are billed the same as any other partner node.
No. These behave like every other partner node: an account and credits are enough. There is no plan floor.
Eight nodes at launch. Four text-to-image (Flux 2, Mage Flow, Mage Flow Turbo, Z-Image Turbo), three MiniMax H3 video nodes (text, image, and first-last frame to video), and MiniMax Music 3 for audio.
A curated workflow per node, pinned to an exact model version so a saved graph keeps producing the same thing. Your graph, prompts and inputs stay on your machine apart from what the node sends.
Per GPU-second in credits, shown on the node before you run. A short image generation is a few seconds; video is longer.
Not yet. Each node exposes a fixed set of weights we host. Bringing your own is the obvious next step and is not in this first release.
No, and it should not. Keep running what your machine handles well. These nodes exist for the models it cannot, without asking you to leave your workflow.
## Run the model, not the hardware
Update to v0.34.5 or later, search the node library for Comfy Cloud, and wire one in.
[Get Started](https://comfy.org/download) [Read the Docs](https://docs.comfy.org/cloud-nodes/overview)
---
# Comfy Cloud - AI in the Cloud
> The easiest way to start with ComfyUI. Pre-loaded models. Pre-installed custom nodes. Concurrent jobs. The full power of ComfyUI on Blackwell RTX 6000 Pros. Open a tab and start creating.
CLOUD
# The full power of ComfyUI — from anywhere.
The easiest way to start with ComfyUI. Pre-loaded models. Pre-installed custom nodes. Concurrent jobs. The full power of ComfyUI on Blackwell RTX 6000 Pros. Open a tab and start creating.
[TRY COMFY CLOUD FOR FREE](https://cloud.comfy.org/)
## Why professionals choose Cloud
### Powerful GPUs with end-to-end security built-in
Comfy Cloud works on any device. Pay only for running workflows, not idle time. With Comfy Cloud, you get security and infrastructure built-in with access to the most popular custom nodes.
### All models. Commercial license guaranteed.
ONLY ON CLOUD
Run open-source models like Wan 2.2, Flux, LTX and Qwen alongside partner models like Nano Banana, Seedance, Seedream, Grok, Kling, Hunyuan 3D, GPT Image 2 and more. Every model on Comfy Cloud is cleared for commercial use. No license ambiguity. All through one credit balance.
### More control than any other visual AI tool
Every node exposed. Every setting adjustable. ComfyUI gives you the full inference pipeline. Choose your sampler, your scheduler, your model chain. The cloud simplifies the setup and supercharges the hardware underneath it.
### Community workflows, unlimited customization through pre-installed custom nodes
Browse, run, and remix workflows built by thousands of creators. Start from proven templates instead of blank canvases. Upload custom LoRAs or finetuned foundational models from CivitAI and Hugging Face. The nodes powering ~90% of local ComfyUI workflows are now in the cloud.
AI MODELS
## Run the world’s leading AI models
New models are added as they launch.
[Seedance 2.0](https://comfy.org/workflows/) [
Nano Banana Pro](https://comfy.org/workflows/) [Grok Video](https://comfy.org/workflows/) [
Qwen Image Edit](https://comfy.org/workflows/) [Wan 2.2](https://comfy.org/workflows/) [GPT Image 2](https://comfy.org/workflows/)
[EXPLORE WORKFLOWS](https://comfy.org/workflows/)
## Built for creators who need quality, control, and simplicity.
[
CREATORS
### From idea to output in minutes.
For those who want to generate images, video, and 3D with more control than prompt-based tools offer — without spending a weekend configuring your machine.](https://cloud.comfy.org/) [
TEAMS & STUDIOS
### Onboard your team today.
No IT setup, no GPU procurement, no environment headaches. Everyone works from the same up-to-date platform. Share workflows via App Mode links — your team runs them instantly, no training required.](https://cloud.comfy.org/)
## Simple, credit-based pricing
One balance for Cloud GPU time and Partner Node API models. Build and edit workflows for free — credits are consumed only when the GPU runs.
[SEE PRICING PLANS](https://comfy.org/pricing)
## The AI creation engine for complete control
Over every model, every node, every step, every output.
[Comfy Desktop
Run ComfyUI on your own hardware.
SEE DESKTOP FEATURES](https://comfy.org/download) [Developer Platform
Deploy and scale ComfyUI workflows.
SEE PLATFORM FEATURES](https://comfy.org/platform) [Comfy Enterprise
Enterprise-grade infrastructure for the creative engine inside your organization.
SEE ENTERPRISE FEATURES](https://comfy.org/enterprise)
## FAQs
Comfy Cloud is a version of ComfyUI that we officially host — no setup, no GPU required. Run your workflows instantly on high-performance cloud GPUs.
Cloud runs on powerful remote GPUs and is accessible from any device. Comfy Desktop runs entirely on your computer, giving you full control and offline use.
Comfy Cloud has zero setup, is easy to share with your team, and is faster than most GPUs you can run on a desktop workstation. You can immediately run the best models and workflows from the community on Comfy Cloud. Comfy Desktop is infinitely customizable, works offline, and you don't need to worry about queue times. However, depending on what you want to create, you might need to have a good GPU and some amount of technical knowledge to install community-created custom nodes.
No, you can start creating instantly from your browser, no matter what computer you use.
Comfy Cloud runs on Blackwell RTX 6000 Pros — 96GB VRAM, with a library of 900+ pre-installed models and support for many of the most-used custom nodes from the ComfyUI community. We expand node support regularly based on demand and compatibility.
Yes, your workflows work across Desktop and Cloud. Just note that only the most popular custom nodes are supported for now, but more will be added soon.
You can always check Cloud to see the list of extensions and models that we support, for free. Most popular ones are available, and new ones are added over time.
You can always check Cloud to see the list of extensions and models that we support, for free. Currently, we support a wide variety of preinstalled models. For those on the Creator, Pro, or Team plans, you can bring in your own fine-tuned LoRAs from CivitAI or HuggingFace to perfect your own style. Direct file upload for larger models is on our roadmap.
Each workflow has a max runtime of 30 minutes on Standard and Creator, raised to 1 hour on Pro. Jobs over the limit are cancelled automatically to keep the system fair and stable. You can queue up to 100 workflows at once, and run 1 / 3 / 5 concurrently via API on Standard / Creator / Pro. Need higher API rate limits? Contact enterprise@comfy.org.
By default, all your inputs, outputs, and workflows are private to your account. For enhanced security features or enterprise-level options, please contact our team via support@comfy.org for more details.
Yes, absolutely. ComfyUI will always be free and open source. You can deploy it however you want, such as downloading it from GitHub, using Docker, custom setups, etc. Comfy Cloud is an optional hosted service for those who prefer convenience, accessibility, or don't have powerful GPUs.
Plans start at $20/mo with a credit-based model. For full pricing details — credits, plans, Team plan, billing, and refunds — see the [Pricing FAQs](https://comfy.org/cloud/pricing#faq).
For pricing, plans, credits, and billing details, see the [Pricing FAQs](https://comfy.org/cloud/pricing#faq).
---
# Custom-node packs on Comfy Cloud - supported by default
> Browse hundreds of ComfyUI custom-node packs preinstalled on Comfy Cloud. Detailers, ControlNet preprocessors, animation tools, samplers, and more — search by pack or by node name.
CLOUD NODES
# Run your favorite ComfyUI custom nodes on the cloud
Spin up workflows with hundreds of community-built nodes — detailers, ControlNet preprocessors, animation tools, and quality-of-life utilities — preinstalled on Comfy Cloud and ready to run on managed GPUs.
CLOUD NODES
## Find a custom-node pack
Search custom-node packs Sort packs
### [ComfyUI-VideoHelperSuite](https://comfy.org/cloud/supported-nodes/comfyui-videohelpersuite/)
Nodes related to video workflows
[View repository](https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite) • 40 nodes
Included nodes
#### Video Helper Suite 🎥🅥🅗🅢
- VHS_BatchManager
- VHS_LoadImagePath
- VHS_LoadImages
- VHS_LoadImagesPath
- VHS_LoadVideo
- VHS_LoadVideoFFmpeg
- VHS_LoadVideoFFmpegPath
- VHS_LoadVideoPath
- VHS_PruneOutputs
- VHS_SelectFilename
- VHS_SelectLatest
- VHS_Unbatch
- VHS_VideoCombine
- VHS_VideoInfo
- VHS_VideoInfoLoaded
- VHS_VideoInfoSource
#### Video Helper Suite 🎥🅥🅗🅢/audio
- VHS_AudioToVHSAudio
- VHS_LoadAudio
- VHS_LoadAudioUpload
- VHS_VHSAudioToAudio
#### Video Helper Suite 🎥🅥🅗🅢/batched nodes
- VHS_VAEDecodeBatched
- VHS_VAEEncodeBatched
#### Video Helper Suite 🎥🅥🅗🅢/image
- VHS_DuplicateImages
- VHS_GetImageCount
- VHS_MergeImages
- VHS_SelectEveryNthImage
- VHS_SelectImages
- VHS_SplitImages
#### Video Helper Suite 🎥🅥🅗🅢/latent
- VHS_DuplicateLatents
- VHS_GetLatentCount
- VHS_MergeLatents
- VHS_SelectEveryNthLatent
- VHS_SelectLatents
- VHS_SplitLatents
#### Video Helper Suite 🎥🅥🅗🅢/mask
- VHS_DuplicateMasks
- VHS_GetMaskCount
- VHS_MergeMasks
- VHS_SelectEveryNthMask
- VHS_SelectMasks
- VHS_SplitMasks
### [ComfyUI Impact Pack](https://comfy.org/cloud/supported-nodes/comfyui-impact-pack/)
This node pack offers various detector nodes and detailer nodes that allow you to configure a workflow that automatically enhances facial details. And provide iterative upscaler.
[View repository](https://github.com/ltdrdata/ComfyUI-Impact-Pack) • 197 nodes
Included nodes
#### —
- AddMask
- AnyPipeToBasic
- BasicPipeToDetailerPipe
- BasicPipeToDetailerPipeSDXL
- BboxDetectorCombined_v2
- BboxDetectorSEGS
- BitwiseAndMask
- BitwiseAndMaskForEach
- BlackPatchRetryHookProvider
- CfgScheduleHookProvider
- CLIPSegDetectorProvider
- CombineRegionalPrompts
- CoreMLDetailerHookProvider
- CustomSamplerDetailerHookProvider
- DenoiseScheduleHookProvider
- DenoiseSchedulerDetailerHookProvider
- DetailerForEach
- DetailerForEachAutoRetry
- DetailerForEachDebug
- DetailerForEachDebugPipe
- DetailerForEachPipe
- DetailerForEachPipeForAnimateDiff
- DetailerHookCombine
- DetailerPipeToBasicPipe
- EditBasicPipe
- EditDetailerPipe
- EditDetailerPipeSDXL
- EmptySegs
- FaceDetailer
- FaceDetailerPipe
- FromBasicPipe
- FromBasicPipe_v2
- FromDetailerPipe
- FromDetailerPipe_v2
- FromDetailerPipeSDXL
- GITSSchedulerFuncProvider
- ImageListToImageBatch
- ImageMaskSwitch
- ImageReceiver
- ImageSender
- ImpactAssembleSEGS
- ImpactBoolean
- ImpactCombineConditionings
- ImpactCompare
- ImpactConcatConditionings
- ImpactConditionalBranch
- ImpactConditionalBranchSelMode
- ImpactConditionalStopIteration
- ImpactControlBridge
- ImpactControlNetApplyAdvancedSEGS
- ImpactControlNetApplySEGS
- ImpactControlNetClearSEGS
- ImpactConvertDataType
- ImpactCount_Elts_in_SEGS
- ImpactDecomposeSEGS
- ImpactDilate_Mask_SEG_ELT
- ImpactDilateMask
- ImpactDilateMaskInSEGS
- ImpactDummyInput
- ImpactEdit_SEG_ELT
- ImpactExecutionOrderController
- ImpactFlattenMask
- ImpactFloat
- ImpactFrom_SEG_ELT
- ImpactFrom_SEG_ELT_bbox
- ImpactFrom_SEG_ELT_crop_region
- ImpactGaussianBlurMask
- ImpactGaussianBlurMaskInSEGS
- ImpactHFTransformersClassifierProvider
- ImpactIfNone
- ImpactImageBatchToImageList
- ImpactImageInfo
- ImpactInt
- ImpactInversedSwitch
- ImpactIPAdapterApplySEGS
- ImpactIsNotEmptySEGS
- ImpactKSamplerAdvancedBasicPipe
- ImpactKSamplerBasicPipe
- ImpactLatentInfo
- ImpactListBridge
- ImpactLogger
- ImpactLogicalOperators
- ImpactMakeAnyList
- ImpactMakeImageBatch
- ImpactMakeImageList
- ImpactMakeMaskBatch
- ImpactMakeMaskList
- ImpactMakeTileSEGS
- ImpactMinMax
- ImpactNeg
- ImpactNegativeConditioningPlaceholder
- ImpactNodeSetMuteState
- ImpactQueueTrigger
- ImpactQueueTriggerCountdown
- ImpactRemoteBoolean
- ImpactRemoteInt
- ImpactSAM2VideoDetectorSEGS
- ImpactScaleBy_BBOX_SEG_ELT
- ImpactSchedulerAdapter
- ImpactSegsAndMask
- ImpactSegsAndMaskForEach
- ImpactSEGSClassify
- ImpactSEGSConcat
- ImpactSEGSIntersectionFilter
- ImpactSEGSLabelAssign
- ImpactSEGSLabelFilter
- ImpactSEGSMerge
- ImpactSEGSNMSFilter
- ImpactSEGSOrderedFilter
- ImpactSEGSPicker
- ImpactSEGSRangeFilter
- ImpactSEGSToMaskBatch
- ImpactSEGSToMaskList
- ImpactSelectNthItemOfAnyList
- ImpactSetWidgetValue
- ImpactSimpleDetectorSEGS
- ImpactSimpleDetectorSEGS_for_AD
- ImpactSimpleDetectorSEGSPipe
- ImpactSleep
- ImpactStringSelector
- ImpactSwitch
- ImpactValueReceiver
- ImpactValueSender
- ImpactWildcardEncode
- ImpactWildcardProcessor
- IterativeImageUpscale
- IterativeLatentUpscale
- KSamplerAdvancedProvider
- KSamplerProvider
- LamaRemoverDetailerHookProvider
- LatentPixelScale
- LatentReceiver
- LatentSender
- LatentSwitch
- MaskDetailerPipe
- MaskListToMaskBatch
- MaskRectArea
- MaskRectAreaAdvanced
- MasksToMaskList
- MaskToSEGS
- MaskToSEGS_for_AnimateDiff
- MediaPipeFaceMeshToSEGS
- NoiseInjectionDetailerHookProvider
- NoiseInjectionHookProvider
- ONNXDetectorProvider
- ONNXDetectorSEGS
- PixelKSampleHookCombine
- PixelKSampleUpscalerProvider
- PixelKSampleUpscalerProviderPipe
- PixelTiledKSampleUpscalerProvider
- PixelTiledKSampleUpscalerProviderPipe
- PreviewBridge
- PreviewBridgeLatent
- PreviewDetailerHookProvider
- ReencodeLatent
- ReencodeLatentPipe
- RegionalPrompt
- RegionalSampler
- RegionalSamplerAdvanced
- RemoveImageFromSEGS
- RemoveNoiseMask
- SAMDetectorCombined
- SAMDetectorSegmented
- SAMLoader
- SegmDetectorCombined_v2
- SegmDetectorSEGS
- SEGSDetailer
- SEGSDetailerForAnimateDiff
- SEGSLabelFilterDetailerHookProvider
- SEGSOrderedFilterDetailerHookProvider
- SEGSPaste
- SEGSPreview
- SEGSPreviewCNet
- SEGSRangeFilterDetailerHookProvider
- SEGSSwitch
- SegsToCombinedMask
- SEGSToImageList
- SEGSUpscaler
- SEGSUpscalerPipe
- SetDefaultImageForSEGS
- StepsScheduleHookProvider
- StringListToString
- SubtractMask
- SubtractMaskForEach
- TiledKSamplerProvider
- ToBasicPipe
- ToBinaryMask
- ToDetailerPipe
- ToDetailerPipeSDXL
- TwoAdvancedSamplersForMask
- TwoSamplersForMask
- TwoSamplersForMaskUpscalerProvider
- TwoSamplersForMaskUpscalerProviderPipe
- UnsamplerDetailerHookProvider
- UnsamplerHookProvider
- VariationNoiseDetailerHookProvider
- WildcardPromptFromString
### [ComfyUI_essentials](https://comfy.org/cloud/supported-nodes/comfyui_essentials/)
Essential nodes that are weirdly missing from ComfyUI core. With few exceptions they are new features and not commodities.
[View repository](https://github.com/cubiq/ComfyUI_essentials) • 62 nodes
Included nodes
#### essentials/conditioning
- CLIPTextEncodeSDXL+
- ConditioningCombineMultiple+
- SD3NegativeConditioning+
#### essentials/image analysis
- ImageEnhanceDifference+
#### essentials/image batch
- ImageBatchMultiple+
- ImageExpandBatch+
- ImageFromBatch+
- ImageListToBatch+
#### essentials/image manipulation
- ImageComposite+
- ImageCompositeFromMaskBatch+
- ImageCrop+
- ImageFlip+
- ImageRandomTransform+
- ImageRemoveBackground+
- ImageResize+
- ImageSeamCarving+
- ImageTile+
- ImageUntile+
- RemBGSession+
- TransparentBGSession+
#### essentials/image processing
- ImageApplyLUT+
- ImageCASharpening+
- ImageColorMatch+
- ImageColorMatchAdobe+
- ImageDesaturate+
- ImageHistogramMatch+
- ImagePosterize+
- PixelOEPixelize+
#### essentials/image utils
- GetImageSize+
- ImagePreviewFromLatent+
- ImageRemoveAlpha+
- ImageToDevice+
- NoiseFromImage+
#### essentials/mask
- MaskBlur+
- MaskBoundingBox+
- MaskFix+
- MaskFlip+
- MaskFromColor+
- MaskFromList+
- MaskFromRGBCMYBW+
- MaskFromSegmentation+
- MaskPreview+
- MaskSmooth+
- TransitionMask+
#### essentials/mask batch
- MaskBatch+
- MaskExpandBatch+
- MaskFromBatch+
#### essentials/sampling
- FluxSamplerParams+
- InjectLatentNoise+
- KSamplerVariationsStochastic+
- KSamplerVariationsWithNoise+
- PlotParameters+
#### essentials/segmentation
- ApplyCLIPSeg+
- LoadCLIPSegModels+
#### essentials/text
- DrawText+
#### essentials/utilities
- BatchCount+
- ConsoleDebug+
- DebugTensorShape+
- ModelCompile+
- RemoveLatentMask+
- SDXLEmptyLatentSizePicker+
- SimpleMath+
### [ComfyUI_LayerStyle](https://comfy.org/cloud/supported-nodes/comfyui_layerstyle/)
A set of nodes for ComfyUI it generate image like Adobe Photoshop's Layer Style. the Drop Shadow is first completed node, and follow-up work is in progress.
[View repository](https://github.com/chflame163/ComfyUI_LayerStyle) • 19 nodes
Included nodes
#### 😺dzNodes/LayerFilter
- LayerFilter: HDR Effects
#### 😺dzNodes/LayerMask
- LayerMask: DrawRoundedRectangle
- LayerMask: Segformer Clothes Setting
- LayerMask: Segformer Fashion Setting
#### 😺dzNodes/LayerUtility
- LayerUtility: ColorImage
- LayerUtility: GradientImage
#### 😺dzNodes/LayerUtility/Data
- LayerUtility: Any Rerouter
- LayerUtility: Boolean
- LayerUtility: Float
- LayerUtility: Gray Value
- LayerUtility: HSV Value
- LayerUtility: Integer
- LayerUtility: PrintInfo
- LayerUtility: RGB Value
- LayerUtility: Seed
- LayerUtility: String
- LayerUtility: TextBox
- LayerUtility: TextJoin
- LayerUtility: TextJoinV2
### [ComfyUI-WanVideoWrapper](https://comfy.org/cloud/supported-nodes/ComfyUI-WanVideoWrapper/)
ComfyUI wrapper nodes for WanVideo
[View repository](https://github.com/kijai/ComfyUI-WanVideoWrapper) • 51 nodes
Included nodes
#### WanVideoWrapper
- Create CFG Schedule Float List
- Create Schedule Float List
- Draw NLF Poses
- Dummy Comfy Wan Model Object
- Extract Start Frames For Continuations
- FantasyPortrait Model Loader
- FantasyTalking Model Loader
- Load Lynx Resampler
- Load NLF Model
- Load VQVAE
- Lynx Encode Face IP
- Lynx InsightFace Crop
- MTV Crafter Encode Poses
- Multi/InfiniteTalk Model Loader
- MultiTalk Silent Embeds
- WanVideo Block List
- WanVideo CLIP Text Encoder Loader
- WanVideo Control Embeds
- WanVideo Controlnet Loader
- WanVideo EasyCache
- WanVideo Empty Embeds
- WanVideo Enhance-A-Video
- WanVideo Experimental Args
- WanVideo Extra Model Select
- WanVideo FlashVSR Decoder Loader
- WanVideo Free Init Args
- WanVideo Loop Args
- WanVideo Lora Block Edit
- WanVideo MagCache
- WanVideo MiniMax Remover Embeds
- WanVideo Phantom Embeds
- WanVideo RealisDance Latents
- WanVideo ReCamMaster Default Camera
- WanVideo ReCamMaster Generate Orbit Camera
- WanVideo RoPE Function
- WanVideo Sigma To Step
- WanVideo SLG
- WanVideo SVIPro Embeds
- WanVideo T5 Text Encoder Loader
- WanVideo TeaCache
- WanVideo TextEncode Cached
- WanVideo Tiny VAE Loader
- WanVideo Torch Compile Settings
- WanVideo UniAnimate DWPose Detector
- WanVideo UniAnimate Pose Input
- WanVideo UniLumos Embeds
- WanVideo VACE Module Select
- Wav2vec2 Model Loader
- Whisper Model Loader
#### WanVideoWrapper/Ovi
- Ovi MMAudio VAE Loader
- WanVideo Empty MMAudio Latents
### [ComfyUI-Frame-Interpolation](https://comfy.org/cloud/supported-nodes/comfyui-frame-interpolation/)
A custom node suite for Video Frame Interpolation in ComfyUI
[View repository](https://github.com/Fannovel16/ComfyUI-Frame-Interpolation) • 16 nodes
Included nodes
#### ComfyUI-Frame-Interpolation
- VFI FloatToInt
#### ComfyUI-Frame-Interpolation/others
- KSampler Gradually Adding More Denoise (efficient)
#### ComfyUI-Frame-Interpolation/VFI
- AMT VFI
- ATM VFI
- CAIN VFI
- FILM VFI
- FLAVR VFI
- GMFSS Fortuna VFI
- IFRNet VFI
- IFUnet VFI
- M2M VFI
- Make Interpolation State List
- MOMO VFI
- RIFE VFI
- Sepconv VFI
- STMFNet VFI
### [ComfyUI_UltimateSDUpscale](https://comfy.org/cloud/supported-nodes/comfyui_ultimatesdupscale/)
ComfyUI nodes for the Ultimate Stable Diffusion Upscale script by Coyote-A.
[View repository](https://github.com/ssitu/ComfyUI_UltimateSDUpscale) • 3 nodes
Included nodes
#### image/upscaling
- UltimateSDUpscale
- UltimateSDUpscaleCustomSample
- UltimateSDUpscaleNoUpscale
### [ComfyUI_IPAdapter_plus](https://comfy.org/cloud/supported-nodes/comfyui_ipadapter_plus/)
ComfyUI reference implementation for the IPAdapter models. The IPAdapter are very powerful models for image conditioning. The style and composition of a reference can be easily transferred to the generation. Think of it as a 1-image lora.
[View repository](https://github.com/cubiq/ComfyUI_IPAdapter_plus) • 30 nodes
Included nodes
#### ipadapter
- IPAdapter
- IPAdapterAdvanced
- IPAdapterBatch
- IPAdapterUnifiedLoader
#### ipadapter/dev
- IPAdapterMS
#### ipadapter/embeds
- IPAdapterCombineEmbeds
- IPAdapterEmbeds
- IPAdapterEmbedsBatch
- IPAdapterEncoder
- IPAdapterLoadEmbeds
- IPAdapterSaveEmbeds
#### ipadapter/faceid
- IPAAdapterFaceIDBatch
- IPAdapterFaceID
- IPAdapterUnifiedLoaderFaceID
#### ipadapter/loaders
- IPAdapterInsightFaceLoader
- IPAdapterModelLoader
- IPAdapterUnifiedLoaderCommunity
#### ipadapter/params
- IPAdapterCombineParams
- IPAdapterFromParams
- IPAdapterRegionalConditioning
#### ipadapter/style_composition
- IPAdapterStyleComposition
- IPAdapterStyleCompositionBatch
#### ipadapter/tiled
- IPAdapterTiled
- IPAdapterTiledBatch
#### ipadapter/utils
- IPAdapterCombineWeights
- IPAdapterNoise
- PrepImageForClipVision
#### ipadapter/weights
- IPAdapterPromptScheduleFromWeightsStrategy
- IPAdapterWeights
- IPAdapterWeightsFromStrategy
### [WAS Node Suite v3](https://comfy.org/cloud/supported-nodes/was-node-suite-comfyui/)
WAS-NS reborn; Tools for image processing, filters, masking, text, logic, numbers, latents, files, 3D scenes, and animation.
[View repository](https://github.com/WASasquatch/was-node-suite-comfyui) • 217 nodes
Included nodes
#### WAS Suite/Animation
- Create Morph Image
- Create Morph Image from Path
- Create Video from Path
- Video Dump Frames
#### WAS Suite/Animation/Writer
- Write to GIF
- Write to Video
#### WAS Suite/Conditioning
- CLIPTextEncode (NSP)
#### WAS Suite/Debug
- Debug Number to Console
- Dictionary to Console
- Export API
- Image Bounds to Console
- Samples Passthrough (Stat System)
- Text to Console
#### WAS Suite/History
- Image History Loader
- Text File History Loader
#### WAS Suite/Image
- Image Batch
- Image Blank
- Image Blend
- Image Blend by Mask
- Image Blending Mode
- Images to Linear
- Images to RGB
#### WAS Suite/Image/Adjustment
- Image Levels Adjustment
- Image Rotate Hue
- Image Shadows and Highlights
#### WAS Suite/Image/AI
- Image Rembg (Remove Background)
- MiDaS Depth Approximation
- MiDaS Mask Image
#### WAS Suite/Image/Analyze
- Image Analyze
- Image Color Palette
- Image to Seed
#### WAS Suite/Image/Bound
- Bounded Image Blend
- Bounded Image Blend with Mask
- Bounded Image Crop
- Bounded Image Crop with Mask
- Image Bounds
- Inset Image Bounds
#### WAS Suite/Image/Filter
- Image Bloom Filter
- Image Canny Filter
- Image Chromatic Aberration
- Image Dragan Photography Filter
- Image Edge Detection Filter
- Image fDOF Filter
- Image Film Grain
- Image Filter Adjustments
- Image Gradient Map
- Image High Pass Filter
- Image Lucy Sharpen
- Image Median Filter
- Image Monitor Effects Filter
- Image Nova Filter
- Image SSAO (Ambient Occlusion)
- Image SSDO (Direct Occlusion)
- Image Style Filter
#### WAS Suite/Image/Generate
- Image Generate Gradient
#### WAS Suite/Image/Generate/Noise
- Image Perlin Noise
- Image Perlin Power Fractal
- Image Power Noise
- Image to Noise
- Image Voronoi Noise Filter
#### WAS Suite/Image/Masking
- CLIPSeg Batch Masking
- CLIPSeg Masking
- Convert Masks to Images
- Image to Latent Mask
- Mask Arbitrary Region
- Mask Batch
- Mask Batch to Mask
- Mask Ceiling Region
- Mask Crop Dominant Region
- Mask Crop Minority Region
- Mask Crop Region
- Mask Dilate Region
- Mask Dominant Region
- Mask Erode Region
- Mask Fill Holes
- Mask Floor Region
- Mask Gaussian Region
- Mask Invert
- Mask Minority Region
- Mask Paste Region
- Mask Smooth Region
- Mask Threshold Region
- Masks Add
- Masks Combine Batch
- Masks Combine Regions
- Masks Subtract
- SAM Image Mask
- SAM Model Loader
- SAM Parameters
- SAM Parameters Combine
#### WAS Suite/Image/Process
- Create Grid Image
- Create Grid Image from Batch
- Image Crop Face
- Image Crop Location
- Image Crop Square Location
- Image Mix RGB Channels
- Image Paste Crop
- Image Paste Crop by Location
- Image Paste Face
- Image Pixelate
- Image Remove Background (Alpha)
- Image Remove Color
- Image Seamless Texture
- Image Select Channel
- Image Select Color
- Image Threshold
- Image Tiled
#### WAS Suite/Image/Transform
- Image Displacement Warp
- Image Flip
- Image Padding
- Image Resize
- Image Rotate
- Image Stitch
- Image Transpose
#### WAS Suite/Integer
- Integer place counter
#### WAS Suite/IO
- Cache Node
- Image Load
- Image Save
- Image Send HTTP
- Load Cache
- Load Image Batch
- Load Text File
- Save Text File
#### WAS Suite/Latent
- Blend Latents
- Latent Batch
#### WAS Suite/Latent/Generate
- Latent Noise Injection
#### WAS Suite/Latent/Transform
- Latent Upscale by Factor (WAS)
- Tensor Batch to Image
#### WAS Suite/Loaders
- BLIP Model Loader
- Checkpoint Loader (Simple)
- CLIPSeg Model Loader
- Load Lora
- Lora Loader
- MiDaS Model Loader
- unCLIP Checkpoint Loader
- Upscale Model Loader
#### WAS Suite/Loaders/Advanced
- Checkpoint Loader
- Diffusers Hub Model Down-Loader
- Diffusers Model Loader
#### WAS Suite/Logic
- Boolean To Text
- CLIP Input Switch
- CLIP Vision Input Switch
- Conditioning Input Switch
- Control Net Model Input Switch
- Image Aspect Ratio
- Image Input Switch
- Latent Input Switch
- Logic Boolean
- Logic Boolean Primitive
- Logic Comparison AND
- Logic Comparison OR
- Logic Comparison XOR
- Logic NOT
- Lora Input Switch
- Model Input Switch
- Number Input Condition
- Number Input Switch
- Text Contains
- Text Input Switch
- Upscale Model Switch
- VAE Input Switch
#### WAS Suite/Number
- Constant Number
- Number Counter
- Number PI
- Random Number
- Seed
- True Random.org Number Generator
#### WAS Suite/Number/Functions
- Number Multiple Of
#### WAS Suite/Number/Operations
- Image Size to Number
- Latent Size to Number
- Number Operation
- Number to Float
- Number to Int
- Number to Seed
- Number to String
- Number to Text
#### WAS Suite/Sampling
- KSampler (WAS)
- KSampler Cycle
#### WAS Suite/Text
- Prompt Multiple Styles Selector
- Prompt Styles Selector
- Text Concatenate
- Text Dictionary Convert
- Text Dictionary Get
- Text Dictionary Keys
- Text Dictionary New
- Text Dictionary To Text
- Text Dictionary Update
- Text List
- Text List Concatenate
- Text List to Text
- Text Load Line From File
- Text Multiline
- Text Multiline (Code Compatible)
- Text Random Line
- Text Random Prompt
- Text String
#### WAS Suite/Text/AI
- BLIP Analyze Image
#### WAS Suite/Text/Operations
- String to Text
- Text Shuffle
- Text Sort
- Text String Truncate
- Text to Conditioning
- Text to Number
- Text to String
#### WAS Suite/Text/Parse
- Text Parse A1111 Embeddings
- Text Parse Noodle Soup Prompts
#### WAS Suite/Text/Search
- Text Compare
- Text Find
- Text Find and Replace
- Text Find and Replace by Dictionary
- Text Find and Replace Input
#### WAS Suite/Text/Tokens
- Text Add Token by Input
- Text Add Tokens
- Text Parse Tokens
#### WAS Suite/Utilities
- Bus Node
- Hex to HSL
- HSL to Hex
### [ComfyUI-RMBG](https://comfy.org/cloud/supported-nodes/comfyui-rmbg/)
A sophisticated ComfyUI custom node engineered for advanced image background removal and precise segmentation of objects, faces, clothing, and fashion elements. This tool leverages a diverse array of models, including RMBG-2.0, INSPYRENET, BEN, BEN2, BiRefNet, SDMatte models, SAM, SAM2, SAM3 and GroundingDINO, while also incorporating a new feature for real-time background replacement and enhanced edge detection for improved accuracy.
[View repository](https://github.com/1038lab/ComfyUI-RMBG) • 9 nodes
Included nodes
#### 🧪AILab/🎭Inpaint
- Reference Latent Mask (RMBG) 🖼️🎭
#### 🧪AILab/🖼️IMAGE
- IC LoRA Concat (RMBG) 🖼️🎭
- Image/Mask Preview (RMBG) 🖼️🎭
- Mask Enhancer (RMBG) 🎭
- Mask Overlay (RMBG) 🖼️🎭
#### 🧪AILab/🧽RMBG
- Accessories Segment (RMBG)
- BiRefNet (RMBG)
- Lama Remover (RMBG)
- Remove Background (RMBG)
### [ComfyUI-Advanced-ControlNet](https://comfy.org/cloud/supported-nodes/comfyui-advanced-controlnet/)
Nodes for scheduling ControlNet strength across timesteps and batched latents, as well as applying custom weights and attention masks.
[View repository](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet) • 2 nodes
Included nodes
#### Adv-ControlNet 🛂🅐🅒🅝/Reference
- Reference ControlNet (Finetune) 🛂🅐🅒🅝
#### Adv-ControlNet 🛂🅐🅒🅝/SparseCtrl
- SparseCtrl Index Method 🛂🅐🅒🅝
### [ComfyUI-AnimateDiff-Evolved](https://comfy.org/cloud/supported-nodes/comfyui-animatediff-evolved/)
Improved AnimateDiff integration for ComfyUI.
[View repository](https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved) • 11 nodes
Included nodes
#### Animate Diff 🎭🅐🅓/conditioning
- Timesteps Conditioning 🎭🅐🅓
#### Animate Diff 🎭🅐🅓/context opts/context extras/contextref
- ContextRef Mode◆First 🎭🅐🅓
- ContextRef Mode◆Indexes 🎭🅐🅓
- ContextRef Mode◆Sliding 🎭🅐🅓
- ContextRef Tune◆Attn 🎭🅐🅓
- ContextRef Tune◆Attn+Adain 🎭🅐🅓
#### Animate Diff 🎭🅐🅓/extras
- Empty Latent Image (Big Batch) 🎭🅐🅓
#### Animate Diff 🎭🅐🅓/iteration opts
- Default Iteration Options 🎭🅐🅓
#### Animate Diff 🎭🅐🅓/sample settings/image inject
- Image Injection Options 🎭🅐🅓
#### Animate Diff 🎭🅐🅓/scheduling
- Float to Floats 🎭🅐🅓
- Value Scheduling 🎭🅐🅓
### [ComfyUI-QwenVL](https://comfy.org/cloud/supported-nodes/ComfyUI-QwenVL/)
ComfyUI-QwenVL custom node: Integrates the Qwen-VL series, including Qwen2.5-VL, Qwen3-VL, Qwen3.5-VL, Qwen3.6-VL (MoE), and Qwen3.8-VL, with GGUF support for advanced multimodal AI in text generation, image understanding, and video analysis.
[View repository](https://github.com/1038lab/ComfyUI-QwenVL) • 1 node
Included nodes
#### 🧪AILab/QwenVL
- QwenVL Prompt Enhancer
### [ComfyUI-MelBandRoFormer](https://comfy.org/cloud/supported-nodes/ComfyUI-MelBandRoFormer/)
ComfyUI wrapper nodes for WanVideo
[View repository](https://github.com/kijai/ComfyUI-MelBandRoFormer) • 2 nodes
Included nodes
#### Mel-Band RoFormer
- MelBandRoFormerModelLoader
- MelBandRoFormerSampler
### [ComfyUI-Logic](https://comfy.org/cloud/supported-nodes/comfyui-logic/)
An extension to ComfyUI that introduces logic nodes and conditional rendering capabilities.
[View repository](https://github.com/theUpsider/ComfyUI-Logic) • 7 nodes
Included nodes
#### Logic
- Bool
- Compare
- DebugPrint
- Float
- If ANY execute A else B
- Int
- String
### [ComfyUI-WanAnimatePreprocess](https://comfy.org/cloud/supported-nodes/ComfyUI-WanAnimatePreprocess/)
ComfyUI nodes for WanAnimate input processing
[View repository](https://github.com/kijai/ComfyUI-WanAnimatePreprocess) • 1 node
Included nodes
#### WanAnimatePreprocess
- ONNX Detection Model Loader
### [Basic data handling](https://comfy.org/cloud/supported-nodes/basic_data_handling/)
Basic Python functions for manipulating data that every programmer is used to, lightweight with no additional dependencies. Supported data types: - ComfyUI native: BOOLEAN, FLOAT, INT, STRING, and data lists - Python types as custom data types: DICT, LIST, SET, DATETIME, TIMEDELTA - PyTorch: TENSOR Feature categories: - Boolean logic operations - Type casting/conversion between all supported data types - Comparison operations - Data container manipulation - Flow control (conditionals, branching, execution order) - Mathematical operations - Mathematical formula node in a safe implementation - String manipulation - File system path handling, including STRING, IMAGE and MASK load and save - SET operations - time and date handling - PyTorch Tensor manipulation (arithmetic, slicing, reshaping)
[View repository](https://github.com/StableLlama/ComfyUI-basic_data_handling) • 258 nodes
Included nodes
#### Basic/BOOLEAN
- and
- and (generic)
- nand
- nor
- not
- or
- or (generic)
- xor
#### Basic/cast
- to BOOLEAN
- to DICT
- to FLOAT
- to INT
- to LIST
- to SET
- to STRING
#### Basic/comparison
- !=
- <
- <=
- ==
- >
- >=
- compare length
- is null
- string compare
#### Basic/Data List
- all
- any
- append
- convert to LIST
- convert to SET
- create Data List
- create Data List from BOOLEANs
- create Data List from FLOATs
- create Data List from INTs
- create Data List from STRINGs
- create Data List of Lists
- enumerate
- extend
- filter
- filter select
- first
- last
- max
- min
- pop random
- range
- shuffle
- sum
#### Basic/DICT
- compare
- contains key
- create DICT
- create DICT from BOOLEANs
- create DICT from FLOATs
- create DICT from INTs
- create DICT from STRINGs
- create from items (data list)
- create from items (LIST)
- create from LISTs
- exclude keys
- filter by keys
- from keys
- get
- get keys values
- get multiple
- invert
- items
- keys
- length
- merge
- pop
- pop item
- pop random
- remove
- set
- setdefault
- update
- values
#### Basic/FLOAT
- add
- create FLOAT
- divide
- divide (zero safe)
- from hex
- integer ratio
- is integer
- multiply
- power
- round
- subtract
- to hex
#### Basic/flow control
- continue flow
- flow select
- force calculation
- force execution order
- if/elif/.../else
- if/else
- is connected
- switch/case
#### Basic/INT
- add
- bit count
- bit length
- create INT
- divide
- divide (zero safe)
- from bytes
- modulus
- multiply
- power
- subtract
- to bytes
#### Basic/LIST
- append
- contains
- convert to Data List
- convert to SET
- count
- create LIST
- create LIST from BOOLEANs
- create LIST from FLOATs
- create LIST from INTs
- create LIST from STRINGs
- extend
- first
- get item
- index
- insert
- last
- length
- max
- min
- pop
- pop random
- range
- remove
- reverse
- set item
- shuffle
- slice
- sort
#### Basic/maths
- abs
- acos
- asin
- atan
- atan2
- ceil
- cos
- degrees
- e (2.71828)
- exp
- floor
- formula
- log
- log10
- max
- min
- pi (3.14159)
- radians
- sin
- sqrt
- tan
#### Basic/SET
- add
- all
- any
- contains
- convert to Data List
- convert to LIST
- create SET
- create SET from BOOLEANs
- create SET from FLOATs
- create SET from INTs
- create SET from STRINGs
- difference
- discard
- enumerate
- intersection
- is disjoint
- is subset
- is superset
- length
- pop
- pop random
- remove
- sum
- symmetric difference
- union
#### Basic/STRING
- capitalize
- casefold
- center
- concat
- count
- endswith
- escape
- expandtabs
- find
- format_map
- in (contains)
- join (from data list)
- join (from LIST)
- length
- ljust
- lower
- lstrip
- removeprefix
- removesuffix
- replace
- rfind
- rjust
- rsplit (from data list)
- rsplit (from LIST)
- rstrip
- split (to data list)
- split (to LIST)
- splitlines (from data list)
- splitlines (to LIST)
- startswith
- strip
- swapcase
- title
- unescape
- upper
- zfill
#### Basic/STRING/is
- isalnum
- isalpha
- isascii
- isdecimal
- isdigit
- isidentifier
- islower
- isnumeric
- isprintable
- isspace
- istitle
- isupper
#### Basic/STRING/regex
- find all (data list)
- find all (LIST)
- search groups (data list)
- search groups (LIST)
- search named groups
- split (data list)
- split (LIST)
- substitute
- test
#### Basic/tensor
- Tensor Binary Op
- Tensor Create
- Tensor Info
- Tensor Join
- Tensor Permute
- Tensor Reshape
- Tensor Slice
- Tensor Unary Op
#### Basic/time
- Add Time Delta
- Create Time Delta
- Extract Time Components
- Format Time String
- Parse Time String
- Subtract Time Delta
- Time Difference
- Time Now
- Time Now (UTC)
- Time to Unix Timestamp
- Unix Timestamp to Time
### [ComfyUI_LayerStyle_Advance](https://comfy.org/cloud/supported-nodes/ComfyUI_LayerStyle_Advance/)
The nodes detached from ComfyUI Layer Style are mainly those with complex requirements for dependency packages.
[View repository](https://github.com/chflame163/ComfyUI_LayerStyle_Advance) • 77 nodes
Included nodes
#### 😺dzNodes/LayerMask
- LayerMask: BBoxJoin
- LayerMask: BenUltra
- LayerMask: BiRefNetUltra
- LayerMask: BiRefNetUltraV2
- LayerMask: DrawBBoxMask
- LayerMask: DrawBBoxMaskV2
- LayerMask: EVFSAMUltra
- LayerMask: Florence2Ultra
- LayerMask: HumanPartsUltra
- LayerMask: LoadBenModel
- LayerMask: LoadBiRefNetModel
- LayerMask: LoadBiRefNetModelV2
- LayerMask: LoadFlorence2Model
- LayerMask: LoadSAM2Model
- LayerMask: LoadSegmentAnythingModels
- LayerMask: MaskByDifferent
- LayerMask: MediapipeFacialSegment
- LayerMask: ObjectDetectorFL2
- LayerMask: ObjectDetectorGemini
- LayerMask: ObjectDetectorGeminiV2
- LayerMask: ObjectDetectorMask
- LayerMask: ObjectDetectorYOLO8
- LayerMask: ObjectDetectorYOLOWorld
- LayerMask: PersonMaskUltra
- LayerMask: PersonMaskUltra V2
- LayerMask: SAM2Ultra
- LayerMask: SAM2UltraV2
- LayerMask: SAM2VideoUltra
- LayerMask: SegmentAnythingUltra
- LayerMask: SegmentAnythingUltra V2
- LayerMask: SegmentAnythingUltra V3
- LayerMask: TransparentBackgroundUltra
- LayerMask: YoloV8Detect
#### 😺dzNodes/LayerUtility
- LayerUtility: Collage
- LayerUtility: DeepSeekAPI
- LayerUtility: DeepSeekAPIV2
- LayerUtility: Gemini
- LayerUtility: GeminiImageEdit
- LayerUtility: GeminiV2
- LayerUtility: GetColorTone
- LayerUtility: GetColorToneV2
- LayerUtility: ImageAutoCrop
- LayerUtility: ImageAutoCrop V2
- LayerUtility: ImageAutoCrop V3
- LayerUtility: ImageRewardFilter
- LayerUtility: JimengI2IAPI
- LayerUtility: JoyCaption2
- LayerUtility: JoyCaption2ExtraOptions
- LayerUtility: JoyCaption2Split
- LayerUtility: JoyCaptionBeta1
- LayerUtility: JoyCaptionBeta1ExtraOptions
- LayerUtility: LaMa
- LayerUtility: LlamaVision
- LayerUtility: LoadJoyCaption2Model
- LayerUtility: LoadJoyCaptionBeta1Model
- LayerUtility: LoadSmolLM2Model
- LayerUtility: LoadSmolVLMModel
- LayerUtility: PhiPrompt
- LayerUtility: SmolLM2
- LayerUtility: SmolVLM
- LayerUtility: ZhipuGLM4
- LayerUtility: ZhipuGLM4V
#### 😺dzNodes/LayerUtility/Prompt
- LayerUtility: Florence2Image2Prompt
- LayerUtility: PromptEmbellish
- LayerUtility: PromptTagger
- LayerUtility: QWenImage2Prompt
- LayerUtility: UserPromptGeneratorReplaceWord
- LayerUtility: UserPromptGeneratorTxt2ImgPrompt
- LayerUtility: UserPromptGeneratorTxt2ImgPromptWithReference
#### 😺dzNodes/LayerUtility/SystemIO
- LayerUtility: AddBlindWaterMark
- LayerUtility: CreateQRCode
- LayerUtility: DecodeQRCode
- LayerUtility: LoadPSD
- LayerUtility: SaveImagePlus
- LayerUtility: SaveImagePlusV2
- LayerUtility: SD3NegativeConditioning
- LayerUtility: ShowBlindWaterMark
### [ComfyUI-enricos-nodes](https://comfy.org/cloud/supported-nodes/comfyui-enricos-nodes/)
pass up to 8 images and visually place, rotate and scale them to build the perfect composition. group move and group rescale. remember their position and scaling value across generations to easy swap images. use the buffer zone to to park an asset you don't want to use or easily reach transformations controls
[View repository](https://github.com/erosDiffusion/ComfyUI-enricos-nodes) • 4 nodes
Included nodes
#### image
- 💜 Compositor (V3)
- 💜 Compositor Tools (V3) Experimental
- 💜 Compositor Transforms Output (V3)
#### image/color
- 💜 Compositor Color Picker
### [ComfyUI-CogVideoXWrapper](https://comfy.org/cloud/supported-nodes/comfyui-cogvideoxwrapper/)
Diffusers wrapper for CogVideoX -models: https://github.com/THUDM/CogVideo
[View repository](https://github.com/kijai/ComfyUI-CogVideoXWrapper) • 24 nodes
Included nodes
#### CogVideoWrapper
- CogVideoContextOptions
- CogVideoControlNet
- CogVideoDecode
- CogVideoEnhanceAVideo
- CogVideoImageEncode
- CogVideoImageEncodeFunInP
- CogVideoLoraSelect
- CogVideoLoraSelectComfy
- CogVideoSampler
- CogVideoTextEncode
- CogVideoTextEncodeCombine
- CogVideoTransformerEdit
- CogVideoXFasterCache
- CogVideoXFunResizeToClosestBucket
- CogVideoXModelLoader
- CogVideoXVAELoader
- DownloadAndLoadCogVideoControlNet
- DownloadAndLoadCogVideoGGUFModel
- DownloadAndLoadCogVideoModel
- DownloadAndLoadToraModel
- ToraEncodeOpticalFlow
- ToraEncodeTrajectory
#### MochiWrapper
- CogVideoXTorchCompileSettings
#### PyramidFlowWrapper
- CogVideoLatentPreview
### [ComfyUI-iTools](https://comfy.org/cloud/supported-nodes/comfyui-itools/)
Whether you're a casual user or a power user, iTools provides the quality-of-life improvements you need to make your ComfyUI experience more efficient and enjoyable.
[View repository](https://github.com/MohammadAboulEla/ComfyUI-iTools) • 12 nodes
Included nodes
#### iTools
- iTools Add Text Overlay
- iTools Checkerboard 🏁
- iTools Grid Filler 📲
- iTools Instructor 👨🏻🏫
- iTools KSampler
- iTools Line Loader
- iTools Load Image 🏕️
- iTools Preview Bridge ⛳
- iTools Prompt Record 🪶
- iTools Regex Editor
- iTools Text Preview
- iTools Text Replacer
### [WhatDreamsCost-ComfyUI](https://comfy.org/cloud/supported-nodes/WhatDreamsCost-ComfyUI/)
LTX Director and a variety of other custom ComfyUI nodes and workflows
[View repository](https://github.com/WhatDreamsCost/WhatDreamsCost-ComfyUI) • 1 node
Included nodes
#### WhatDreamsCost
- Speech Length Calculator
### [comfyui_face_parsing](https://comfy.org/cloud/supported-nodes/comfyui_face_parsing/)
This is a set of custom nodes for ComfyUI. The nodes utilize the [a/face parsing model](https://huggingface.co/jonathandinu/face-parsing) to parse face and provide detailed segmantation. To improve face segmantation accuracy, [a/yolov8 face model](https://huggingface.co/Bingsu/adetailer/) is used to first extract face from an image. There are also auxiliary nodes for image and mask processing. A guided filter is also provided for skin smoothing.
[View repository](https://github.com/Ryuukeisyou/comfyui_face_parsing) • 27 nodes
Included nodes
#### face_parsing
- BBoxDecompose(FaceParsing)
- BBoxDetect(FaceParsing)
- BBoxDetectorLoader(FaceParsing)
- BBoxListItemSelect(FaceParsing)
- BBoxResize(FaceParsing)
- ColorAdjust(FaceParsing)
- FaceParse(FaceParsing)
- FaceParsingModelLoader(FaceParsing)
- FaceParsingProcessorLoader(FaceParsing)
- FaceParsingResultsParser(FaceParsing)
- GuidedFilter(FaceParsing)
- ImageCropWithBBox(FaceParsing)
- ImageCropWithBBoxList(FaceParsing)
- ImageInsertWithBBox(FaceParsing)
- ImageListSelect(FaceParsing)
- ImagePadWithBBox(FaceParsing)
- ImageResizeCalculator(FaceParsing)
- ImageResizeWithBBox(FaceParsing)
- ImageSize(FaceParsing)
- MaskBatchComposite(FaceParsing)
- MaskBlackOut(FaceParsing)
- MaskBorderDissolve(FaceParsing)
- MaskComposite(FaceParsing)
- MaskCropWithBBox(FaceParsing)
- MaskInsertWithBBox(FaceParsing)
- MaskListSelect(FaceParsing)
- MaskToBBoxList(FaceParsing)
### [ComfyUI-SCAIL-Pose](https://comfy.org/cloud/supported-nodes/ComfyUI-SCAIL-Pose/)
ComfyUI nodes for SCAIL pose input processing
[View repository](https://github.com/kijai/ComfyUI-SCAIL-Pose) • 4 nodes
Included nodes
#### WanAnimatePreprocess
- ConvertOpenPoseKeypointsToDWPose
- PoseDetectionVitPoseToDWPose
#### WanVideoWrapper
- RenderNLFPoses
- SaveNLFPosesAs3D
### [ComfyUI_Steudio](https://comfy.org/cloud/supported-nodes/comfyui_steudio/)
Divide and Conquer Node Suite: It calculates the optimal upscale resolution and seamlessly divides the image into tiles, ready for individual processing using your preferred workflow. After processing, the tiles are seamlessly merged into a larger image, offering sharper and more detailed visuals.
[View repository](https://github.com/Steudio/ComfyUI_Steudio) • 10 nodes
Included nodes
#### Steudio/Divide and Conquer
- Combine Tiles
- Divide and Conquer Algorithm
- Divide Image and Select Tile
#### Steudio/Utils
- Display UI
- Load Images into List
- Ratio Calculator
- Ratio to Size
- Seed Shifter
- Sequence Generator
- Simple Config
### [SAM3](https://comfy.org/cloud/supported-nodes/comfyui-sam3/)
Warning, uses experimental package `comfy-env` to attempt a one click isolated install. Will download and use pixi package manager. SAM 3 (Segment Anything Model 3) integration for ComfyUI. Image and video segmentation with text prompts, boxes, and points
[View repository](https://github.com/PozzettiAndrea/ComfyUI-SAM3) • 14 nodes
Included nodes
#### SAM3
- (down)Load SAM3 Model
- SAM3 BBox Collector 📦
- SAM3 Multi-Region Collector 🎯
- SAM3 Multiprompt Segmentation
- SAM3 Point Collector
- SAM3 Point Segmentation
#### SAM3/Grounding
- SAM3 Text Segmentation
#### SAM3/prompts
- SAM3 Combine Boxes
- SAM3 Combine Points
- SAM3 Create Box
- SAM3 Create Point
#### SAM3/video
- SAM3 Propagate
- SAM3 Video Output
- SAM3 Video Segmentation
### [ComfyUI-WanVaceAdvanced](https://comfy.org/cloud/supported-nodes/wanvaceadvanced/)
Precise control over Vace frame strengths for reference images, control videos. Also allows Phantom embeddings to be combined with VACE accurately.
[View repository](https://github.com/drozbay/ComfyUI-WanVaceAdvanced) • 2 nodes
Included nodes
#### WanVaceAdvanced
- StringToFloatListRanged
- WVAOptions
### [FL Path Animator](https://comfy.org/cloud/supported-nodes/comfyui-fl-path-animator/)
Interactive path animator with visual editor for creating animated shapes that follow user-drawn paths. WAN ATI compatible for stable video generation.
[View repository](https://github.com/filliptm/ComfyUI_FL-Path-Animator.git) • 1 node
Included nodes
#### 🎨 FL Path Animator
- FL_PathAnimator
### [radiance](https://comfy.org/cloud/supported-nodes/radiance/)
Radiance is a 32-bit color space collection of nodes for ComfyUI.
[View repository](https://github.com/fxtdstudios/radiance) • 74 nodes
Included nodes
#### FXTD Studios/Radiance/Analyze
- ◎ RadianceFalseColor
- ◎ RadianceQC
- ◎ RadianceQCExport
- ◎ RadianceVectorscope
- ◎ RadianceWaveform
- HDRHistogram
#### FXTD Studios/Radiance/Color
- ACES2OutputTransform
- ARRIWideGamut4
- ColorSpaceConvert
- DaVinciWideGamut
- Float32ColorCorrect
- FXTD_RadianceGradeApply
- GPUTensorOps
- HDRExpandDynamicRange
- HDRExposureBlend
- HDRShadowHighlightRecovery
- HDRToneMap
- ImageToFloat32
- RadianceApplyGradeInfo
- RadianceGPUColorMatrix
- RadianceGrade
- RadianceGradeMatch
- RadianceHighlightSynthesis
- RadianceLogCurveDecode
- RadianceLogCurveEncode
- RadianceLUTApply
- RadianceLUTBlend
- RadianceOCIOCDL
- RadianceOCIOColorTransform
- RadianceOCIODisplayView
- RadianceSceneLinearWorkflow
- RadianceWhiteBalance
#### FXTD Studios/Radiance/Data
- ◎ RadianceDepthMapGenerator
- RadianceDNAReader
- RadianceDNAValidator
- RadianceDNAWriter
#### FXTD Studios/Radiance/Draw
- ◎ RadianceBlendComposite
- ◎ RadianceMetadataOverlay
#### FXTD Studios/Radiance/Filter
- ◎ RadianceDenoise
- RadianceCompressionArtifacts
- RadianceDepthOfField
- RadianceFilmGrain
- RadianceMotionBlur
- RadianceRollingShutter
#### FXTD Studios/Radiance/Generate
- ◎ RadianceCinemaStudio
- RadianceCinematicPromptEncoder
- RadianceControlApply
- RadianceLoraStack
- RadianceSamplerPro
- RadianceUnifiedLoader
- RadianceVAEDecode
- RadianceVAEEncode
#### FXTD Studios/Radiance/Image
- HDR360Generate
- RadianceAIUpscale
- RadianceBitDepthConvert
- RadianceDownscale32bit
- RadianceProUpscale
- RadianceResolution
- RadianceUpscaleBySize
#### FXTD Studios/Radiance/IO
- ◎ RadianceDigitalCinemaRead
- ◎ RadianceDigitalCinemaWrite
#### FXTD Studios/Radiance/Layout
- RadianceAdvancedReroute
- RadianceDebugProbe
- RadianceGate
- RadianceMux
- RadianceNote
- RadianceReroute
- RadianceWorkspace
#### FXTD Studios/Radiance/Masking
- RadianceLoadImageMask
#### FXTD Studios/Radiance/Pipeline
- ◎ RadianceNukeBridge
#### FXTD Studios/Radiance/Utilities
- ◎ RadianceShowText
#### FXTD Studios/Radiance/Video
- ◎ RadianceFlickerAnalyze
- ◎ RadianceTemporalSmooth
#### FXTD Studios/Radiance/Views
- FXTD_RadianceViewer
### [ComfyUI-Qwen-TTS](https://comfy.org/cloud/supported-nodes/ComfyUI-Qwen-TTS/)
Qwen-TTS nodes for ComfyUI
[View repository](https://github.com/mailzwj/ComfyUI-Qwen-TTS.git) • 3 nodes
Included nodes
#### Qwen
- AudioInfo
- Qwen3TTS
- QwenTTS
### [comfyui-PlyPreview](https://comfy.org/cloud/supported-nodes/comfyui-PlyPreview/)
Gaussian Splat PLY loader + preview nodes for ComfyUI. Forked from GeometryPack and repackaged to avoid upstream overwrites. Includes front-end viewer (gsplat.js) bundled locally.
[View repository](https://github.com/XuanYu-github/comfyui-PlyPreview) • 4 nodes
Included nodes
#### PlyPreview
- PlyPreviewLoadGaussianPLYEnhance
- PlyPreviewLoadGaussianPLYPathEnhance
- PlyPreviewProcessGaussianPLYEnhance
#### PlyPreview/visualization
- PlyPreviewPreviewGaussianEnhance
### [hellorob](https://comfy.org/cloud/supported-nodes/nodesweet/)
Makes ComfyUI more sweet
[View repository](https://github.com/rslosh/comfyui-nodesweet) • 6 nodes
Included nodes
#### nodesweet-hellorob
- AudioReactiveTransform
- AudioWeightsRemap
- BboxDetectorBatch
- BboxDetectorBatchChunked
- BboxDetectorBatchForEach
- MultilineStringRepeater
### [Systms Face Composite](https://comfy.org/cloud/supported-nodes/comfyui-systms-facecomposite/)
Composites refined face images back onto original video frames using bounding box coordinates with support for replace, alpha blend, and feather blending modes.
[View repository](https://github.com/systms-ai/Nodes-) • 1 node
Included nodes
#### WanAnimatePreprocess
- FaceComposite
### [TimeSlice Nodes](https://comfy.org/cloud/supported-nodes/timeslice-nodes/)
Temporal slit-scan / time-displacement effects for ComfyUI — angled, animated and dual-source slicing on a shared engine.
[View repository](https://github.com/systms-ai/TimeSlice-Nodes) • 3 nodes
Included nodes
#### Systms/effects
- TimeSlice Dual
- TimeSlice Effect
- TimeSlice Wave
### [ComfyQR](https://comfy.org/cloud/supported-nodes/ComfyQR/)
Description unavailable.
View repository • 3 nodes
Included nodes
#### ComfyQR
- QR Code
- QR Code (Conformed to Image Size)
- QR Code (Split)
### [AudioTools](https://comfy.org/cloud/supported-nodes/ComfyUI_AudioTools/)
Description unavailable.
View repository • 29 nodes
Included nodes
#### ⚪Lum3on/AudioTools/AI
- Speech Denoise (AI)
- Speech-to-Text + SRT (Whisper)
- Stem Separator (AI)
#### ⚪Lum3on/AudioTools/Analysis
- Audio-Reactive Envelope
- BPM Detector / Reactive
- Loudness Meter (LUFS)
#### ⚪Lum3on/AudioTools/Effects
- Delay / Echo
- Fade In
- Fade Out
- Pitch Shift / Time Stretch
- Reverb
#### ⚪Lum3on/AudioTools/Processing
- Amplify / Gain
- De-Esser
- De-Hum (50/60Hz)
- De-Plosive (Low Cut)
- Mix Audio Tracks
- Noise Gate
- Normalize Audio
- Parametric EQ for Voice
- Remove Silence
- Standardize Audio (Format/Channels)
- Trim Audio
- Vocal Compressor
#### ⚪Lum3on/AudioTools/Utility
- Concatenate Audio
- Pad With Silence
- Stereo Panner
#### ⚪Lum3on/AudioTools/Visualization
- Compare Waveforms
- Display Waveform
- Show Audio Info
### [Fill-Nodes](https://comfy.org/cloud/supported-nodes/ComfyUI_Fill-Nodes/)
Description unavailable.
View repository • 46 nodes
Included nodes
#### 🏵️Fill Nodes/Audio
- FL Audio Beat Visualizer
- FL Audio BPM Analyzer
- FL Audio Crop
- FL Audio Drum Detector
- FL Audio Envelope Visualizer
- FL Audio Music Video Sequencer
- FL Audio Reactive Brightness
- FL Audio Reactive Edge Glow
- FL Audio Reactive Envelope
- FL Audio Reactive Saturation
- FL Audio Reactive Scale
- FL Audio Reactive Speed
- FL Audio Segment Extractor
- FL Audio Separation
- FL Audio Shot Iterator
#### 🏵️Fill Nodes/experiments
- FL Color Picker
#### 🏵️Fill Nodes/games
- FL BulletHell Game
- FL Tetris Game
#### 🏵️Fill Nodes/Image
- FL Anime Line Extractor
- FL Image Add To Batch
- FL Image Blank
- FL Image Crop
- FL Image Size
#### 🏵️Fill Nodes/Ksamplers
- FL Sampler String XYZ
- FL Scheduler String XYZ
#### 🏵️Fill Nodes/PDF
- FL Text To PDF
#### 🏵️Fill Nodes/Prompting
- FL MadLib Generator
- FL Prompt Basic
- FL Prompt Multi
- FL Prompt Selector
- FL Prompt Selector Basic
#### 🏵️Fill Nodes/Utility
- FL Float
- FL Float to Int
- FL Int to Float
- FL Math
- FL Path Type Checker
- FL Random Number
- FL Switch
- FL Switch Big
#### 🏵️Fill Nodes/VFX
- FL Image Pixelator
#### 🏵️Fill Nodes/Video
- FL Video Cadence Compile
#### 🏵️Fill Nodes/WIP
- FL Animated Shape Patterns
- FL Wan Video Blender
- FL Wan Video Blender 3-Way
- FL Wan Video Continuation Blender
- FL Wan Video Continue
### [SchemaNodes](https://comfy.org/cloud/supported-nodes/ComfyUI_SchemaNodes/)
Description unavailable.
View repository • 8 nodes
Included nodes
#### inout/schema
- Schema Audio Parameter
- Schema Boolean Parameter
- Schema Enum Parameter
- Schema Float Parameter
- Schema Image Parameter
- Schema Integer Parameter
- Schema String Parameter
- Schema Video Parameter
### [DepthAnythingV2](https://comfy.org/cloud/supported-nodes/ComfyUI-DepthAnythingV2/)
Description unavailable.
View repository • 1 node
Included nodes
#### DepthAnythingV2
- DownloadAndLoadDepthAnythingV2Model
### [DynamiCrafterWrapper](https://comfy.org/cloud/supported-nodes/ComfyUI-DynamiCrafterWrapper/)
Description unavailable.
View repository • 2 nodes
Included nodes
#### DynamiCrafterWrapper
- (Down)Load CLIPVisionModel
- (Down)Load DynamiCrafterModel
### [EditUtils](https://comfy.org/cloud/supported-nodes/ComfyUI-EditUtils/)
Description unavailable.
View repository • 4 nodes
Included nodes
#### advanced/conditioning
- EditUtils: Flux2Klein Model Config lrzjason
- EditUtils: Model Config lrzjason
- EditUtils: Qwen Model Config lrzjason
#### image
- EditUtils: Load Image With Filename lrzjason
### [FlashVSR_Ultra_Fast](https://comfy.org/cloud/supported-nodes/ComfyUI-FlashVSR_Ultra_Fast/)
Description unavailable.
View repository • 3 nodes
Included nodes
#### FlashVSR
- FlashVSR Init Pipeline
- FlashVSR Ultra-Fast
- FlashVSR Ultra-Fast (Advanced)
### [Fluxtapoz](https://comfy.org/cloud/supported-nodes/ComfyUI-Fluxtapoz/)
Description unavailable.
View repository • 6 nodes
Included nodes
#### fluxtapoz
- Flow Edit Sampler
- Flux Forward ODE Sampler
- Flux Inverse Sampler
- RF-Edit Double Layers Override
- RF-Edit Single Layers Override
#### fluxtapoz/attn
- Flux Attention Override
### [Image-Filters](https://comfy.org/cloud/supported-nodes/ComfyUI-Image-Filters/)
Description unavailable.
View repository • 3 nodes
Included nodes
#### Image-Filters/image
- Image Constant Color (HSV)
- Image Constant Color (RGB)
#### Image-Filters/latent
- Offset Latent Image
### [KJNodes](https://comfy.org/cloud/supported-nodes/ComfyUI-KJNodes/)
Description unavailable.
View repository • 41 nodes
Included nodes
#### KJNodes/audio
- Sound Reactive
#### KJNodes/constants
- BOOL Constant
- Float Constant
- INT Constant
- String Constant
- String Constant Multiline
#### KJNodes/deprecated
- Create Audio Mask
#### KJNodes/experimental
- Flux Block Lora Select
- Interpolate Coords
- Plot Coordinates
#### KJNodes/hunyuanvideo
- Hunyuan Video Block Lora Select
#### KJNodes/image
- Create Gradient From Coords
- Create Shape Image On Path
- Image Batch Extend With Overlap
- Image Grab PIL
- Load & Resize Image
- Screencap mss
#### KJNodes/InstanceDiffusion
- AppendInstanceDiffusionTracking
- CreateInstanceDiffusionTracking
#### KJNodes/latents
- Empty Latent Image Presets
#### KJNodes/ltxv
- LTX2 Block Lora Select
#### KJNodes/masking/generate
- Create Fluid Mask
- Create Gradient Mask
- Create Magic Mask
- Create Shape Mask
- Create Shape Mask On Path
- Create Voronoi Mask
- Float To Mask
#### KJNodes/misc
- Scale Batch Prompt Schedule
- String to Float List
#### KJNodes/model_loaders
- Diffusion Model Selector
#### KJNodes/noise
- Custom Sigmas
- Float To Sigmas
#### KJNodes/text
- Append Strings To List
- Join String Multi
- Join Strings
- Superprompt
#### KJNodes/vae
- VAE Decode Loop KJ
#### KJNodes/wan
- Wan21 Block Lora Select
#### KJNodes/weights
- Weight Schedule Convert
- Weight Schedule Extend
### [KwaiKolorsWrapper](https://comfy.org/cloud/supported-nodes/ComfyUI-KwaiKolorsWrapper/)
Description unavailable.
View repository • 1 node
Included nodes
#### KwaiKolorsWrapper
- (Down)load ChatGLM3 Model
### [LivePortraitKJ](https://comfy.org/cloud/supported-nodes/ComfyUI-LivePortraitKJ/)
Description unavailable.
View repository • 5 nodes
Included nodes
#### LivePortrait
- LivePortrait KeypointsToImage
- LivePortrait Load FaceAlignmentCropper
- LivePortrait Load InsightFaceCropper
- LivePortrait Load MediaPipeCropper
- LivePortrait Retargeting
### [Lotus](https://comfy.org/cloud/supported-nodes/ComfyUI-Lotus/)
Description unavailable.
View repository • 1 node
Included nodes
#### ComfyUI-Lotus
- Load Lotus Model
### [LTXVideo](https://comfy.org/cloud/supported-nodes/ComfyUI-LTXVideo/)
Description unavailable.
View repository • 19 nodes
Included nodes
#### latent
- 🅛🅣🅧 LTXV Spatio Temporal Tiled VAE Decode
#### Lightricks/audio
- 🅛🅣🅧 LTXV Audio Only Empty Video Latent
- 🅛🅣🅧 LTXV Audio Only Model
#### lightricks/LTXV
- 🅛🅣🅧 Gemma 3 Model Loader
- 🅛🅣🅧 Guider Parameters
- 🅛🅣🅧 LTXV Apply STG
- 🅛🅣🅧 STG Advanced Presets
#### ltxtricks
- LTX Attention Bank
- LTX Attn Block Override
- LTX Flow Edit Sampler
#### ltxtricks/attn
- LTX Attention Override
#### LTXV/loaders
- 🅛🅣🅧 Low VRAM Audio VAE Loader
- 🅛🅣🅧 Low VRAM Checkpoint Loader
#### math/conversion
- 🅛🅣🅧 Float To Int
#### prompt
- 🅛🅣🅧 LTXV Multi Prompt Provider
- 🅛🅣🅧 Multi Prompt Provider
#### sampling
- 🅛🅣🅧 LTXV Base Sampler
- 🅛🅣🅧 LTXV In Context Sampler
- 🅛🅣🅧 LTXV Looping Sampler
### [Prompt-Combinator](https://comfy.org/cloud/supported-nodes/ComfyUI-Prompt-Combinator/)
Description unavailable.
View repository • 1 node
Included nodes
#### prompt_combinator
- 🔢 Prompt Combinator
### [Comfyui-SecNodes](https://comfy.org/cloud/supported-nodes/Comfyui-SecNodes/)
Description unavailable.
View repository • 3 nodes
Included nodes
#### SeC
- Coordinate Plotter
- SeC Model Loader
- SeC Video Segmentation
### [comfyui-sharp](https://comfy.org/cloud/supported-nodes/comfyui-sharp/)
Description unavailable.
View repository • 2 nodes
Included nodes
#### SHARP
- Load Image with EXIF (Focal Length)
- Load SHARP Model
### [comfyui-string-converter](https://comfy.org/cloud/supported-nodes/comfyui-string-converter/)
Description unavailable.
View repository • 2 nodes
Included nodes
#### utils
- Convert String To Float
- Convert String To Int
### [UltraShape1](https://comfy.org/cloud/supported-nodes/ComfyUI-UltraShape1/)
Description unavailable.
View repository • 1 node
Included nodes
#### UltraShape/Loaders
- UltraShape Load Model
### [Upscaler-Tensorrt](https://comfy.org/cloud/supported-nodes/ComfyUI-Upscaler-Tensorrt/)
Description unavailable.
View repository • 2 nodes
Included nodes
#### tensorrt
- Load Upscale Tensorrt Model
- Upscaler Tensorrt ⚡
### [comfyui-various](https://comfy.org/cloud/supported-nodes/comfyui-various/)
Description unavailable.
View repository • 28 nodes
Included nodes
#### jamesWalker55
- Datetime String
- Float
- Float Absolute Value
- Float Add
- Float Divide
- Float Maximum
- Float Minimum
- Float Multiply
- Float Subtract
- Float to String
- Integer
- Integer Absolute Value
- Integer Add
- Integer Divide
- Integer Maximum
- Integer Minimum
- Integer Multiply
- Integer Subtract
- Integer to Float
- Integer to String
- String
- String (Multiline)
- String Concatenate
- String Get Line
- String Replace
- String to Float
- String to Integer
- String Unescape
### [ControlAltAI-Nodes](https://comfy.org/cloud/supported-nodes/ControlAltAI-Nodes/)
Description unavailable.
View repository • 9 nodes
Included nodes
#### ControlAltAI Nodes/Flux
- Flux Resolution Calc
#### ControlAltAI Nodes/Flux Region
- Flux Attention Cleanup
- Region Mask Generator
#### ControlAltAI Nodes/HiDream
- HiDream Resolution
#### ControlAltAI Nodes/Logic
- Boolean Basic
- Boolean Reverse
- Integer Settings
- Integer Settings Advanced
#### ControlAltAI Nodes/Utility
- Text Bridge
### [CustomNodeComfyMath](https://comfy.org/cloud/supported-nodes/CustomNodeComfyMath/)
Description unavailable.
View repository • 14 nodes
Included nodes
#### math/conversion
- BoolToInt
- BreakoutVec2
- BreakoutVec3
- BreakoutVec4
- ComposeVec2
- ComposeVec3
- ComposeVec4
- FloatToInt
- FloatToNumber
- IntToBool
- IntToFloat
- IntToNumber
- NumberToFloat
- NumberToInt
### [masquerade-nodes-comfyui](https://comfy.org/cloud/supported-nodes/masquerade-nodes-comfyui/)
Description unavailable.
View repository • 1 node
Included nodes
#### Masquerade Nodes
- Incrementer
### [OneButtonPrompt](https://comfy.org/cloud/supported-nodes/OneButtonPrompt/)
Description unavailable.
View repository • 5 nodes
Included nodes
#### OneButtonPrompt
- Auto Negative Prompt
- Create Prompt Variant
- One Button Artify
- One Button Preset
- One Button SuperPrompt
### [RES4LYF](https://comfy.org/cloud/supported-nodes/RES4LYF/)
Description unavailable.
View repository • 86 nodes
Included nodes
#### image
- Mask Sketch
#### RES4LYF/images
- SetImageSize
- SetImageSizeWithScale
#### RES4LYF/latents
- EmptyLatentImage64
- EmptyLatentImageCustom
#### RES4LYF/masks
- TemporalCrossAttnMask
- TemporalMaskGenerator
- TemporalSplitAttnMask
- TemporalSplitAttnMask (Midframe)
#### RES4LYF/sampler_options
- ClownSamplerSelector
#### RES4LYF/schedulers
- Constant Scheduler
- Sigmas Iteration Karras
- Sigmas Iteration Polyexp
- Tan Scheduler
- Tan Scheduler 2
- Tan Scheduler 2 Simple
#### RES4LYF/sigmas
- Sigmas Abs
- Sigmas AdaptiveNoiseFloor
- Sigmas AdaptiveStep
- Sigmas Add
- Sigmas Append
- Sigmas ArcCosine
- Sigmas ArcSine
- Sigmas ArcTangent
- Sigmas Attractor
- Sigmas CatmullRom
- Sigmas Chaos
- Sigmas Cleanup
- Sigmas CNFInverse
- Sigmas CollatzIteration
- Sigmas Concat
- Sigmas ConwaySequence
- Sigmas CrossProduct
- Sigmas DeleteBelowFloor
- Sigmas DeleteDuplicates
- Sigmas DotProduct
- Sigmas Easing
- Sigmas Fmod
- Sigmas Frac
- Sigmas From Text
- Sigmas GammaBeta
- Sigmas Gaussian
- Sigmas GaussianCDF
- Sigmas GilbreathSequence
- Sigmas HarmonicDecay
- Sigmas Hyperbolic
- Sigmas If
- Sigmas InvLerp
- Sigmas KernelSmooth
- Sigmas LambertW
- Sigmas LangevinDynamics
- Sigmas Lerp
- Sigmas LinearSine
- Sigmas Logarithm2
- Sigmas Modulus
- Sigmas Mult
- Sigmas Noise Inversion
- Sigmas NormalizingFlows
- Sigmas Pad
- Sigmas Percentile
- Sigmas PersistentHomology
- Sigmas Power
- Sigmas QuantileNorm
- Sigmas Quotient
- Sigmas ReactionDiffusion
- Sigmas RiemannianFlow
- Sigmas SetFloor
- Sigmas Sigmoid
- Sigmas SmoothStep
- Sigmas Split
- Sigmas SquareRoot
- Sigmas Start
- Sigmas StepwiseMultirate
- Sigmas TimeStep
- Sigmas Truncate
- Sigmas Unpad
- Sigmas Variance Floor
- Sigmas ZetaEta
- Sigmas2 Add
- Sigmas2 Mult
#### RES4LYF/text
- TextBox1
- TextBox2
- TextBox3
- TextConcatenate
- TextShuffle
#### RES4LYF/utilities
- SeedGenerator
---
# Basic data handling on Comfy Cloud
> Basic data handling is preinstalled on Comfy Cloud — 258 nodes ready to run on managed GPUs. Basic Python functions for manipulating data that every programmer is used to, lightweight with no additional dependencies. Supported data types: - ComfyUI native: BOOLEAN, FLOAT, INT, STRING, and data lists - Python types as custom data types: DICT, LIST, SET, DATETIME, TIMEDELTA - PyTorch: TENSOR Feature categories: - Boolean logic operations - Type casting/conversion between all supported data types - Comparison operations - Data container manipulation - Flow control (conditionals, branching, execution order) - Mathematical operations - Mathematical formula node in a safe implementation - String manipulation - File system path handling, including STRING, IMAGE and MASK load and save - SET operations - time and date handling - PyTorch Tensor manipulation (arithmetic, slicing, reshaping)
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/StableLlama/ComfyUI-basic_data_handling](https://github.com/StableLlama/ComfyUI-basic_data_handling)
**Publisher**
StableLlama
**Downloads**
234,611
**GitHub stars**
49
**Latest version**
1.8.1
**License**
{"file": "LICENSE"}
**Last updated**
Sep 6, 2026
## Nodes in this pack
### Basic/BOOLEAN
- and
Returns the logical AND result of two boolean values. This node takes two boolean inputs and returns their logical AND result.
- and (generic)
Returns the logical N/AND result of one or more values. This node takes a dynamic number of inputs and returns their logical N/AND result. Note that values are evaluated according Python's rules. I.e. an empty string is `false`, an integer 0 is also `false`, etc.
- nand
Returns the logical NAND result of two boolean values. This node takes two boolean inputs and returns their logical NAND result.
- nor
Returns the logical NOR result of two boolean values. This node takes two boolean inputs and returns their logical NOR result.
- not
Returns the logical NOT result of a boolean value. This node takes one boolean input and returns its logical NOT result.
- or
Returns the logical OR result of two boolean values. This node takes two boolean inputs and returns their logical OR result.
- or (generic)
Returns the logical N/OR result of one or more values. This node takes a dynamic number of inputs and returns their logical N/OR result. Note that values are evaluated according Python's rules. I.e. an empty string is `false`, an integer 0 is also `false`, etc.
- xor
Returns the logical XOR result of two boolean values. This node takes two boolean inputs and returns their logical XOR result.
### Basic/cast
- to BOOLEAN
Converts any input to a BOOLEAN. Follows standard Python truthy/falsy rules.
- to DICT
Converts compatible inputs to a DICT. Input must be a mapping or a list of key-value pairs.
- to FLOAT
Converts any numeric input to a FLOAT. Non-numeric or invalid inputs raise a ValueError.
- to INT
Converts any numeric input to an INT. Non-numeric or invalid inputs raise a ValueError.
- to LIST
Converts any input to a LIST. Non-list inputs are wrapped in a list. If input is a ComfyUI data list, it converts the individual items into a Python LIST.
- to SET
Converts any input to a SET. Non-set inputs are converted into a set. If input is a ComfyUI data list, it casts the individual items into a SET.
- to STRING
Converts any input to a STRING. Non-string values are converted using str().
### Basic/comparison
- !=
Checks if two values are not equal. This node takes two inputs of any type and returns True if they are not equal, and False otherwise. For complex objects, structural inequality is tested.
- <
Checks if the first value is less than the second. This node takes two numerical inputs and returns True if the first value is less than the second value, and False otherwise.
- <=
Checks if the first value is less than or equal to the second. This node takes two numerical inputs and returns True if the first value is less than or equal to the second value, and False otherwise.
- ==
Checks if two values are equal. This node takes two inputs of any type and returns True if they are equal, and False otherwise. For complex objects, structural equality is tested.
- >
Checks if the first value is greater than the second. This node takes two numerical inputs and returns True if the first value is greater than the second value, and False otherwise.
- >=
Checks if the first value is greater than or equal to the second. This node takes two numerical inputs and returns True if the first value is greater than or equal to the second value, and False otherwise.
- compare length
Compares the length of a container (string, list, etc) with a value. This node takes a container and a comparison value, and returns a boolean result based on the comparison of the container's length with the value.
- is null
Checks if a value is None/null. This node takes any input value and returns True if the value is None, and False otherwise.
- string compare
Compares two strings using a selected comparison operator. This node takes two string inputs and a comparison operator, and returns a boolean result based on the selected comparison.
### Basic/Data List
- all
Check if all elements in the data list are true. Returns true if all elements are true (or if the list is empty).
- any
Check if any element in the data list is true. Returns true if at least one element is true. Returns false if the list is empty.
- append
Adds an item to the end of a list. This node takes a list and any item as inputs, then returns the modified list with the new item appended.
- convert to LIST
Converts a ComfyUI Data List into a LIST object. This node takes a Data List input (which is typically a list of items with the same type) and converts it to a LIST object (a Python list as a single variable).
- convert to SET
Converts a ComfyUI Data List into a LIST object. This node takes a Data List input (which is typically a list of items with the same type) and converts it to a LIST object (a Python list as a single variable).
- create Data List
Creates a new Data List from items. This node creates and returns a Data List. The list of items is dynamically extended based on the number of inputs provided.
- create Data List from BOOLEANs
Creates a new Data List from BOOLEAN items. This node creates and returns a Data List. The list of items is dynamically extended based on the number of inputs provided.
- create Data List from FLOATs
Creates a new Data List from FLOAT items. This node creates and returns a Data List. The list of items is dynamically extended based on the number of inputs provided.
- create Data List from INTs
Creates a new Data List from INT items. This node creates and returns a Data List. The list of items is dynamically extended based on the number of inputs provided.
- create Data List from STRINGs
Creates a new Data List from STRING items. This node creates and returns a Data List. The list of items is dynamically extended based on the number of inputs provided.
- create Data List of Lists
Creates a new Data List from items. This node creates and returns a Data List. The list of items is dynamically extended based on the number of inputs provided. Each input can be a list, so you'll get a list of lists.
- enumerate
Enumerate a data list, returning a list of [index, value] pairs. Optionally, specify a starting value for the index.
- extend
Extends a list by appending elements from another list. This node takes two lists as input and returns a new list that contains all elements from both lists.
- filter
Filters a Data List using boolean values. This node takes a value Data List and a filter Data List (containing only boolean values). It returns a new Data List containing only the elements from the value list where the corresponding element in the filter list is False. If the lists have different lengths, the last element of the shorter list is repeated till the lengths are matching.
- filter select
Filters a Data List using boolean values. This node takes a value Data List and a filter Data List (containing only boolean values). It returns two new Data Lists containing only the elements from the value list where the corresponding element in the filter list is true or false. If the lists have different lengths, the last element of the shorter list is repeated till the lengths are matching.
- first
Returns the first element in a list. This node takes a list as input and returns the first element of the list. If the list is empty, it returns None.
- last
Returns the last element in a list. This node takes a list as input and returns the last element of the list. If the list is empty, it returns None.
- max
Finds the maximum value in a list of numbers. This node takes a list of numbers (either FLOAT or INT) and returns the maximum value. Returns None if the list is empty or if it contains non-numeric values.
- min
Finds the minimum value in a list of numbers. This node takes a list of numbers (either FLOAT or INT) and returns the minimum value. Returns None if the list is empty or if it contains non-numeric values.
- pop random
Removes and returns a random element from a list. This node takes a list as input and returns the list with the random element removed and the removed element itself. If the list is empty, it returns None for the element. An optional seed can be provided for reproducible selection.
- range
Creates a data list containing a sequence of numbers. This node generates a sequence of numbers similar to Python's range() function. It takes start, stop, and step parameters to define the sequence.
- shuffle
Shuffles the items in a list using a seed for reproducibility. This node takes a list and a seed as input and returns a new shuffled list.
- sum
Sum all elements of the data list. Returns 0 for an empty list.
### Basic/DICT
- compare
Compares two dictionaries and reports differences. This node takes two dictionaries and compares them, returning information about their equality, any keys that exist in only one dictionary, and any keys with different values.
- contains key
Checks if a key exists in a dictionary. This node takes a dictionary and a key as inputs, then returns True if the key exists in the dictionary, and False otherwise.
- create DICT
Creates a new empty dictionary. This node creates and returns a new empty dictionary object.
- create DICT from BOOLEANs
Creates a new empty dictionary. This node creates and returns a new empty dictionary object.
- create DICT from FLOATs
Creates a new empty dictionary. This node creates and returns a new empty dictionary object.
- create DICT from INTs
Creates a new empty dictionary. This node creates and returns a new empty dictionary object.
- create DICT from STRINGs
Creates a new empty dictionary. This node creates and returns a new empty dictionary object.
- create from items (data list)
Creates a dictionary from a list of key-value pairs. This node takes a list of key-value pairs (tuples) and builds a dictionary from them.
- create from items (LIST)
Creates a dictionary from a list of key-value pairs. This node takes a list of key-value pairs (tuples) and builds a dictionary from them.
- create from LISTs
Creates a dictionary from separate lists of keys and values. This node takes a list of keys and a list of values, then creates a dictionary by pairing corresponding elements from each list. If the lists are of different lengths, only pairs up to the length of the shorter list are used.
- exclude keys
Creates a new dictionary excluding the specified keys. This node takes a dictionary and a list of keys, then returns a new dictionary containing all key-value pairs except those with keys in the provided list.
- filter by keys
Creates a new dictionary with only the specified keys. This node takes a dictionary and a list of keys, then returns a new dictionary containing only the key-value pairs for the keys that exist in the original dictionary.
- from keys
Creates a dictionary from a list of keys and a default value. This node takes a list of keys and an optional value, then creates a new dictionary where each key is associated with the value. If no value is provided, None is used.
- get
Retrieves a value from a dictionary using the specified key. This node gets the value for the specified key from a dictionary. If the key is not found and a default value is provided, that default is returned. Otherwise, it returns None.
- get keys values
Returns keys and values as separate lists. This node takes a dictionary and returns two lists: one containing all keys and another containing all corresponding values.
- get multiple
Retrieves multiple values from a dictionary using a list of keys. This node takes a dictionary and a list of keys, then returns a list containing the corresponding values. If a key is not found, the default value is used for that position.
- invert
Creates a new dictionary with keys and values swapped. This node takes a dictionary as input and returns a new dictionary where the keys become values and values become keys. Note that values must be hashable to be used as keys in the new dictionary.
- items
Returns all key-value pairs in a dictionary. This node takes a dictionary and returns a list of tuples, where each tuple contains a key-value pair from the dictionary.
- keys
Returns all keys in a dictionary. This node takes a dictionary and returns a list containing all of its keys.
- length
Returns the number of key-value pairs in a dictionary. This node takes a dictionary as input and returns its length (number of items).
- merge
Merges multiple dictionaries into a single dictionary. This node takes multiple dictionaries as input and combines them into a single dictionary. If there are duplicate keys, values from later dictionaries take precedence.
- pop
Removes and returns a key-value pair from a dictionary. This node takes a dictionary and a key as inputs, removes the specified key from the dictionary, and returns both the modified dictionary and the value associated with the key. If the key is not found and a default value is provided, that default is returned. Otherwise, an error is raised.
- pop item
Removes and returns an arbitrary key-value pair from a dictionary. This node takes a dictionary as input, removes an arbitrary key-value pair, and returns the modified dictionary along with the removed key and value. If operation fails, no error is thrown, but the last argument is `False`, and the dictionary is returned intact.
- pop random
Removes and returns a random key-value pair from a dictionary. This node takes a dictionary as input, removes a random key-value pair, and returns the modified dictionary along with the removed key and value. If the dictionary is empty, it returns empty values. An optional seed can be provided for reproducible selection.
- remove
Removes a key-value pair from a dictionary. This node takes a dictionary and a key as inputs, then returns a new dictionary with the specified key removed. If the key doesn't exist, the dictionary remains unchanged.
- set
Adds or updates a key-value pair in a dictionary. This node takes a dictionary, key, and value as inputs, then returns a modified dictionary with the new key-value pair.
- setdefault
Returns the value for a key, setting a default if the key doesn't exist. This node takes a dictionary, a key, and a default value. If the key exists in the dictionary, the corresponding value is returned. If the key doesn't exist, the default value is inserted for the key and returned.
- update
Updates a dictionary with key-value pairs from another dictionary. This node takes two dictionaries as inputs and returns a new dictionary that contains all key-value pairs from both dictionaries. If there are duplicate keys, the values from the second dictionary take precedence.
- values
Returns all values in a dictionary. This node takes a dictionary and returns a list containing all of its values.
### Basic/FLOAT
- add
Adds two floating-point numbers. This node takes two floats as input and returns their sum.
- create FLOAT
Create a FLOAT from a STRING widget. The input string must be a valid floating-point number and will be directly converted to a FLOAT without any further processing.
- divide
Divides one floating-point number by another. This node takes two floats as input and returns the result of division. It raises a ValueError if the divisor is 0.
- divide (zero safe)
Divides one floating-point number by another. This node takes two floats as input and returns the result of the division. It returns positive or negative infinity if the divisor is 0 (assumed to be +0.0).
- from hex
Converts a hexadecimal string to its corresponding floating-point number. This node takes a hexadecimal float string as input and returns the float.
- integer ratio
Returns the integer ratio of a floating-point number. This node takes a floating-point number and returns two integers, which represent the ratio as numerator and denominator.
- is integer
Checks if a floating-point number is an integer. This node takes a floating-point number as input and returns True if the number is an integer, otherwise False.
- multiply
Multiplies two floating-point numbers. This node takes two floats as input and returns their product.
- power
Raises one floating-point number to the power of another. This node takes two floats as input and returns the result of raising the first float to the power of the second.
- round
Rounds a floating-point number to the specified number of decimal places. This node takes a float and an integer for decimal places as inputs, and returns the rounded result.
- subtract
Subtracts one floating-point number from another. This node takes two floats as input and returns the result of subtracting the second float from the first.
- to hex
Converts a floating-point number to its hexadecimal representation. This node takes a float as input and returns its hexadecimal string representation.
### Basic/flow control
- continue flow
Conditionally enable or disable a flow. This node takes a value and either passes it through or blocks execution based on the 'select' parameter. When 'select' is True, the value passes through; when False, execution is blocked. When a `message` is provided ComfyUI will display it in a dialog. Leave it empty for silent operation.
- flow select
Select the direction of the flow. This node takes a value and directs it to either the "true" or "false" output. Note: for dynamic switching in a Data Flow you might want to use "filter select" instead.
- force calculation
Forces recalculation of the connected nodes. This node passes the input directly to the output but prevents caching by marking itself as an output node and also indicates the out has changed. Use this when you need to ensure nodes are always recalculated.
- force execution order
Force execution order in the workflow. This node is lightweight and does not affect the workflow. It is used to force the execution order of nodes in the workflow. You only need to chain this node with the other execution order nodes in the desired order and add any output of the nodes you want to force execution order on. This node also passes through any input connected to "any node output" as its second output.
- if/elif/.../else
Implements a conditional branch (if/elif/else) in the workflow. This node takes a condition input and two value inputs. If the condition evaluates to True, the first value is returned; otherwise, the elif (else if) is evaluated and returned when true. This continues for all additional elifs. When none is true, the value of the else is returned. This allows conditional data flow in ComfyUI workflows.
- if/else
Implements a conditional branch (if/else) in the workflow. This node takes a condition input and two value inputs. If the condition evaluates to True, the first value is returned; otherwise, the second value is returned. This allows conditional data flow in ComfyUI workflows.
- is connected
Checks if the input is connected. This node returns True if the 'input' is connected to another node's output, and False otherwise.
- switch/case
Implements a switch/case selection in the workflow. This node takes a selector input (an integer) and multiple case value inputs. It returns the value corresponding to the provided index. If the index is out of range, the default value is returned. This allows for selection from multiple options based on a computed index. NOTE: This version of the node will most likely be deprecated in the future.
### Basic/INT
- add
Adds two integers. This node takes two integers as input and returns their sum.
- bit count
Returns the number of 1 bits in the binary representation of an integer. This node takes an integer as input and returns the count of set bits.
- bit length
Returns the number of bits required to represent an integer in binary. This node takes an integer as input and returns the number of bits needed to represent it, excluding the sign and leading zeros.
- create INT
Create an INT from a STRING widget. The input string must be a valid integer number and will be directly converted to an INT without any further processing. Strings starting with "0b" are interpreted as binary numbers, "0o" as octal numbers, and "0x" as hexadecimal numbers. Note: This doesn't handle ones' complement as the data size is unknown.
- divide
Divides one integer by another. This node takes two integers as input and returns the result of integer division. It raises a ValueError if the divisor is 0.
- divide (zero safe)
Divides one integer by another. This node takes two integers as input and returns the result of the integer division. It returns the positive or negative infinity value if the divisor is 0.
- from bytes
Converts a bytes object to an integer. This class method takes bytes, byte order, and signed flag as inputs and returns an integer.
- modulus
Returns the modulus of two integers. This node takes two integers as input and returns the remainder when the first integer is divided by the second.
- multiply
Multiplies two integers. This node takes two integers as input and returns their product.
- power
Raises one integer to the power of another. This node takes two integers as input and returns the result of raising the first integer to the power of the second.
- subtract
Subtracts one integer from another. This node takes two integers as input and returns the result of subtracting the second integer from the first.
- to bytes
Converts an integer to its byte representation. This node takes an integer, byte length, and byte order as inputs and returns the bytes object representation of the integer.
### Basic/LIST
- append
Adds an item to the end of a LIST. This node takes a LIST and any item as inputs, then returns a new LIST with the item appended to the end.
- contains
Checks if a LIST contains a specified value. This node takes a LIST and a value as inputs, then returns True if the value is present in the LIST, and False otherwise.
- convert to Data List
Converts a LIST object into a ComfyUI data list. This node takes a LIST object (Python list as a single variable) and converts it to a ComfyUI data list, allowing its items to be processed individually by nodes that accept data lists.
- convert to SET
Converts a LIST into a SET. This node takes a LIST input and creates a new SET containing all unique elements from the LIST, removing any duplicates.
- count
Counts the number of occurrences of a value in a LIST. This node takes a LIST and a value as inputs, then returns the number of times the value appears in the LIST.
- create LIST
Creates a new LIST from items. This node creates and returns a LIST. The list of items is dynamically extended based on the number of inputs provided.
- create LIST from BOOLEANs
Creates a new LIST from BOOLEAN items. This node creates and returns a LIST. The list of items is dynamically extended based on the number of inputs provided.
- create LIST from FLOATs
Creates a new LIST from FLOAT items. This node creates and returns a LIST. The list of items is dynamically extended based on the number of inputs provided.
- create LIST from INTs
Creates a new LIST from INT items. This node creates and returns a LIST. The list of items is dynamically extended based on the number of inputs provided.
- create LIST from STRINGs
Creates a new LIST from STRING items. This node creates and returns a LIST. The list of items is dynamically extended based on the number of inputs provided.
- extend
Extends a LIST by appending elements from another LIST. This node takes two LIST objects as input and returns a new LIST that contains all elements from both lists.
- first
Returns the first element in a LIST. This node takes a LIST as input and returns the first element of the list. If the LIST is empty, it returns None.
- get item
Retrieves an item at a specified position in a LIST. This node takes a LIST and an index as inputs, then returns the item at the specified index. Negative indices count from the end of the LIST. Out of range indices return None.
- index
Returns the index of the first occurrence of a value in a LIST. This node takes a LIST and a value as inputs, then returns the index of the first occurrence of the value. Optional start and end parameters limit the search to a slice of the LIST. Returns -1 if the value is not present.
- insert
Inserts an item at a specified position in a LIST. This node takes a LIST, an index, and any item as inputs, then returns a new LIST with the item inserted at the specified index.
- last
Returns the last element in a LIST. This node takes a LIST as input and returns the last element of the list. If the LIST is empty, it returns None.
- length
Returns the number of items in a LIST. This node takes a LIST as input and returns its length as an integer.
- max
Returns the maximum value in a LIST. This node takes a LIST of comparable items and returns the maximum value. Returns None if the LIST is empty or if items are not comparable.
- min
Returns the minimum value in a LIST. This node takes a LIST of comparable items and returns the minimum value. Returns None if the LIST is empty or if items are not comparable.
- pop
Removes and returns an item at a specified position in a LIST. This node takes a LIST and an index as inputs, then returns both the new LIST with the item removed and the removed item. If no index is specified, removes and returns the last item. When the LIST is empty, the item is None.
- pop random
Removes and returns a random element from a LIST. This node takes a LIST as input and returns the LIST with the random element removed and the removed element itself. If the LIST is empty, it returns None for the element. An optional seed can be provided for reproducible selection.
- range
Creates a LIST containing a sequence of numbers. This node generates a LIST of numbers similar to Python's range() function. It takes start, stop, and step parameters to define the sequence.
- remove
Removes the first occurrence of a specified value from a LIST. This node takes a LIST and a value as inputs, then returns a new LIST with the first occurrence of the value removed and a success indicator. If the value is not present, the original LIST is returned with success set to False.
- reverse
Reverses the order of items in a LIST. This node takes a LIST as input and returns a new LIST with the items in reversed order.
- set item
Sets an item at a specified position in a LIST. This node takes a LIST, an index, and a value, then returns a new LIST with the item at the specified index replaced by the value.
- shuffle
Shuffles the items in a list using a seed for reproducibility. This node takes a LIST and a seed as input and returns a new shuffled LIST.
- slice
Creates a slice of a LIST. This node takes a LIST and start/stop/step parameters, and returns a new LIST containing the specified slice of the original LIST.
- sort
Sorts the items in a LIST. This node takes a LIST as input and returns a new sorted LIST. Option includes sorting in reverse order.
### Basic/maths
- abs
Returns the absolute value of a number. This node takes a number and returns its absolute value (magnitude without sign).
- acos
Calculates the arc cosine (inverse cosine) of a value. This node takes a value between -1 and 1, and returns the arc cosine in radians or degrees. The arc cosine is the inverse operation of cosine, returning the angle whose cosine is the input value.
- asin
Calculates the arc sine (inverse sine) of a value. This node takes a value between -1 and 1, and returns the arc sine in radians or degrees. The arc sine is the inverse operation of sine, returning the angle whose sine is the input value.
- atan
Calculates the arc tangent (inverse tangent) of a value. This node takes a value and returns the arc tangent in radians or degrees. The arc tangent is the inverse operation of tangent, returning the angle whose tangent is the input value.
- atan2
Calculates the arc tangent of y/x, considering the quadrant. This node takes y and x coordinates and returns the arc tangent in radians or degrees. Unlike atan(y/x), this function handles the case where x is zero and correctly determines the quadrant of the resulting angle.
- ceil
Returns the ceiling of a number. This node takes a number and returns the smallest integer greater than or equal to the input value.
- cos
Calculates the cosine of an angle in radians or degrees. This node takes an angle value and unit type, and returns the cosine of that angle. The cosine of an angle is the ratio of the length of the adjacent side to the length of the hypotenuse in a right-angled triangle.
- degrees
Converts an angle from radians to degrees. This node takes an angle in radians and returns the equivalent angle in degrees.
- e (2.71828)
Returns the mathematical constant e. This node returns the value of e (Euler's number), which is approximately 2.71828.
- exp
Calculates the exponential of a number (e^x). This node takes a number and returns e (Euler's number) raised to the power of that number.
- floor
Returns the floor of a number. This node takes a number and returns the largest integer less than or equal to the input value.
- formula
A node that evaluates a mathematical formula provided as a string without using eval. This node takes an arbitrary number of numerical inputs (single letters like `a`, `b`, `c`, etc.) and safely evaluates the formula with supported operations: +, -, *, /, //, %, **, parentheses, and mathematical functions. NOTE on Operator Precedence: This parser uses standard precedence rules. Unary minus binds tightly, so expressions like `-a ** 2` are interpreted as `(-a) ** 2`. To calculate `-(a ** 2)`, use parentheses: `-(a ** 2)`. The identifiers `e` and `pi` are special. When used as a function call (`e()`, `pi()`), they return their respective mathematical constants. When used as a variable (`e`), they expect a corresponding input. Supported functions: - Basic: abs, floor, ceil, round, min(a,b), max(a,b) - Trigonometric: sin, cos, tan, asin, acos, atan, atan2(y,x), degrees, radians - Hyperbolic: sinh, cosh, tanh, asinh, acosh, atanh - Exponential & Logarithmic: exp, log, log10, log2, sqrt, pow(base,exp) - Constants (must be called with empty parentheses): pi(), e()
- log
Calculates the natural logarithm (base e) of a number. This node takes a positive number and returns its natural logarithm. If the specified base is not 'e', calculates the logarithm with the specified base.
- log10
Calculates the base-10 logarithm of a number. This node takes a positive number and returns its base-10 logarithm.
- max
Returns the maximum value among the inputs. This node takes two or more values and returns the maximum value among them. The function works with both FLOAT and INT types.
- min
Returns the minimum value among the inputs. This node takes two or more values and returns the minimum value among them. The function works with both FLOAT and INT types.
- pi (3.14159)
Returns the mathematical constant π (pi). This node returns the value of π, which is approximately 3.14159.
- radians
Converts an angle from degrees to radians. This node takes an angle in degrees and returns the equivalent angle in radians.
- sin
Calculates the sine of an angle in radians or degrees. This node takes an angle value and unit type, and returns the sine of that angle. The sine of an angle is the ratio of the length of the opposite side to the length of the hypotenuse in a right-angled triangle.
- sqrt
Calculates the square root of a non-negative number. This node takes a non-negative number and returns its square root.
- tan
Calculates the tangent of an angle in radians or degrees. This node takes an angle value and unit type, and returns the tangent of that angle. The tangent of an angle is the ratio of the length of the opposite side to the length of the adjacent side in a right-angled triangle.
### Basic/SET
- add
Adds an item to a SET. This node takes a SET and any item as inputs, then returns a new SET with the item added. If the item is already present, the SET remains unchanged.
- all
Checks if all elements in the SET are true. This node takes a SET as input and returns True if all elements in the SET evaluate to True (or if the SET is empty), and False otherwise.
- any
Checks if any element in the SET is true. This node takes a SET as input and returns True if at least one element in the SET evaluates to True, and False otherwise (including if the SET is empty).
- contains
Checks if a SET contains a specified value. This node takes a SET and a value as inputs, then returns True if the value is present in the SET, and False otherwise.
- convert to Data List
Converts a SET object into a ComfyUI data list. This node takes a SET object (Python set as a single variable) and converts it to a ComfyUI data list, allowing its items to be processed individually by nodes that accept data lists.
- convert to LIST
Converts a SET into a LIST. This node takes a SET input and creates a new LIST containing all elements from the SET. Note that the order of elements in the resulting LIST is arbitrary since SETs are unordered collections.
- create SET
Creates a new SET from items. This node creates and returns a SET. The list of items is dynamically extended based on the number of inputs provided.
- create SET from BOOLEANs
Creates a new SET from items. This node creates and returns a SET. The list of items is dynamically extended based on the number of inputs provided.
- create SET from FLOATs
Creates a new SET from items. This node creates and returns a SET. The list of items is dynamically extended based on the number of inputs provided.
- create SET from INTs
Creates a new SET from items. This node creates and returns a SET. The list of items is dynamically extended based on the number of inputs provided.
- create SET from STRINGs
Creates a new SET from items. This node creates and returns a SET. The list of items is dynamically extended based on the number of inputs provided.
- difference
Returns the difference between two SETs. This node takes two SETs as input and returns a new SET containing elements in the first SET but not in the second SET.
- discard
Removes an item from a SET if it is present. This node takes a SET and any item as inputs, then returns a new SET with the item removed. Unlike remove, no error is raised if the item is not present.
- enumerate
Enumerates elements in a SET. This node takes a SET as input and returns a LIST of tuples where each tuple contains an index and a value from the SET. The start parameter specifies the initial index value (default is 0). Note: Since SETs are unordered, the enumeration order is arbitrary but consistent within a single operation.
- intersection
Returns the intersection of two or more SETs. This node takes multiple SETs as input and returns a new SET containing only elements present in all input SETs.
- is disjoint
Checks if two SETs have no elements in common. This node takes two SETs as input and returns True if they have no elements in common.
- is subset
Checks if set1 is a subset of set2. This node takes two SETs as input and returns True if set1 is a subset of set2 (all elements in set1 are also in set2).
- is superset
Checks if set1 is a superset of set2. This node takes two SETs as input and returns True if set1 is a superset of set2 (set1 contains all elements in set2).
- length
Returns the number of items in a SET. This node takes a SET as input and returns its length (number of elements) as an integer.
- pop
Removes and returns an arbitrary item from a SET. This node takes a SET as input and returns both the new SET with an arbitrary item removed and the removed item. When the SET is empty, the item is None.
- pop random
Removes and returns a random element from a SET. This node takes a SET as input and returns the SET with a random element removed and the removed element itself. If the SET is empty, it returns None for the element. An optional seed can be provided for reproducible selection.
- remove
Removes an item from a SET. This node takes a SET and any item as inputs, then returns a new SET with the item removed and a success indicator. If the item is not present, the original SET is returned with success set to False.
- sum
Calculates the sum of all elements in a SET. This node takes a SET as input and returns the sum of all its elements. The optional start parameter specifies the initial value (default is 0). Note: This operation requires all elements to be numeric or otherwise compatible with addition. If the SET contains mixed or incompatible types, it may raise an error.
- symmetric difference
Returns the symmetric difference between two SETs. This node takes two SETs as input and returns a new SET containing elements in either SET but not in both.
- union
Returns the union of two or more SETs. This node takes multiple SETs as input and returns a new SET containing all elements from all the input SETs.
### Basic/STRING
- capitalize
Converts the first character of the input string to uppercase and all other characters to lowercase.
- casefold
Converts text to lowercase in a way optimized for case-insensitive comparisons. This node is especially useful when comparing text that might contain special characters. Unlike standard lowercase, this handles special cases like converting the German 'ß' to "ss". Use this node when you need the most accurate case-insensitive text matching.
- center
Centers text within a specified width. This node takes a string and centers it within a field of the specified width. By default, it uses spaces as padding characters, but you can specify any single character to use as padding.
- concat
Combines two text strings together, joining them end-to-end.
- count
Counts occurrences of a substring within a string. This node returns the number of times the specified substring appears in the input string. You can optionally specify start and end positions to limit the search range.
- endswith
Checks if a string ends with the specified suffix. This node tests whether the input string ends with a particular substring. Returns True if the string ends with the specified suffix, otherwise False. Optional start and end parameters allow you to check only a specific portion of the string.
- escape
Escapes a string by converting special characters to escape sequences. This node converts characters like newlines, tabs, quotes, and backslashes to their escaped representation (like ' ', ' ', '"', '\'). Useful when you need to prepare strings for formats that require escaped sequences instead of literal special characters.
- expandtabs
Replaces tab characters with spaces. This node replaces all tab characters (' ') in a string with spaces. The tab size (number of spaces per tab) can be adjusted, with a default value of 8 characters.
- find
Finds the first occurrence of a substring in a string. This node searches for the first occurrence of the specified substring within the input string. Returns the lowest index where the substring is found. If the substring is not found, it returns -1. Optional start and end parameters allow you to limit the search to a specific portion of the string.
- format_map
Formats a string using values from a dictionary. This node replaces placeholders in the input string with values from a dictionary. It uses the same placeholder syntax as the format() method, but takes the values from a dictionary mapping keys to values. For example, the string "Hello, {name}" with a dictionary {"name": "World"} would produce "Hello, World".
- in (contains)
Checks if a string contains a specified substring. This node implements the 'in' operator for strings. It checks whether the given substring exists within the input string. Returns True if the substring is found, otherwise False.
- join (from data list)
Joins strings from a data list with a specified separator. This node takes a data list of strings and concatenates them, with the specified separator string between each element. The separator is inserted between the strings, not at the beginning or end of the result.
- join (from LIST)
Joins strings from a LIST with a specified separator. This node takes a data list of strings and concatenates them, with the specified separator string between each element. The separator is inserted between the strings, not at the beginning or end of the result.
- length
Returns the length of a string. This node calculates and returns the number of characters in the input string. It works the same way as Python's built-in len() function for strings.
- ljust
Left-aligns the string within a field of a given width. This node returns a string left-aligned in a field of the specified width. If fillchar is provided, it is used as the padding character instead of a space.
- lower
Converts the string to lowercase. This node returns a copy of the string with all uppercase characters converted to lowercase.
- lstrip
Removes leading characters from the string. This node returns a copy of the string with leading characters removed. If chars is provided, it specifies the set of characters to be removed. If chars is not provided, whitespace characters are removed.
- removeprefix
Removes prefix from the string if present. This node returns a copy of the string with the specified prefix removed if the string starts with that prefix, otherwise returns the original string.
- removesuffix
Removes suffix from the string if present. This node returns a copy of the string with the specified suffix removed if the string ends with that suffix, otherwise returns the original string.
- replace
Replaces occurrences of a substring with another substring. This node returns a copy of the string with all occurrences of substring old replaced by new. If the optional argument count is given, only the first count occurrences are replaced.
- rfind
Finds the highest index of the substring in the string. This node searches for the last occurrence of the specified substring within the input string. Returns the highest index where the substring is found. If the substring is not found, it returns -1. Optional start and end parameters allow you to limit the search to a specific portion of the string.
- rjust
Right-aligns the string within a field of a given width. This node returns a string right-aligned in a field of the specified width. If fillchar is provided, it is used as the padding character instead of a space.
- rsplit (from data list)
Splits the string at the specified separator from the right. This node returns a list of strings by splitting the input string at the specified separator, starting from the right. If maxsplit is provided, at most maxsplit splits are done. If the separator is not specified or is None, any whitespace string is a separator. Creates a data list of strings.
- rsplit (from LIST)
Splits the string at the specified separator from the right. This node returns a list of strings by splitting the input string at the specified separator, starting from the right. If maxsplit is provided, at most maxsplit splits are done. If the separator is not specified or is None, any whitespace string is a separator. Creates a LIST of strings.
- rstrip
Removes trailing characters from the string. This node returns a copy of the string with trailing characters removed. If chars is provided, it specifies the set of characters to be removed. If chars is not provided, whitespace characters are removed.
- split (to data list)
Splits the string at the specified separator. This node splits the input string at the specified separator and returns a data list containing all parts. If maxsplit is provided, at most maxsplit splits are done. If the separator is not specified or is None, any whitespace string is a separator. Creates a data list of strings.
- split (to LIST)
Splits the string at the specified separator. This node splits the input string at the specified separator and returns a data list containing all parts. If maxsplit is provided, at most maxsplit splits are done. If the separator is not specified or is None, any whitespace string is a separator. Creates a LIST of strings.
- splitlines (from data list)
Splits the string at line boundaries. This node returns a list of the lines in the string, breaking at line boundaries. If keepends is True, line breaks are included in the resulting list. Creates a data list of strings.
- splitlines (to LIST)
Splits the string at line boundaries. This node returns a list of the lines in the string, breaking at line boundaries. If keepends is True, line breaks are included in the resulting list. Creates a LIST of strings.
- startswith
Checks if a string starts with the specified prefix. This node tests whether the input string starts with a particular substring. Returns True if the string starts with the specified prefix, otherwise False. Optional start and end parameters allow you to check only a specific portion of the string.
- strip
Removes leading and trailing characters from the string. This node returns a copy of the string with both leading and trailing characters removed. If chars is provided, it specifies the set of characters to be removed. If chars is not provided, whitespace characters are removed.
- swapcase
Swaps case of all characters in the string. This node returns a copy of the string with uppercase characters converted to lowercase and lowercase characters converted to uppercase.
- title
Converts the string to titlecase. This node returns a titlecased version of the string where words start with an uppercase character and the remaining characters are lowercase.
- unescape
Unescapes a string by converting escape sequences to their actual characters. This node converts escape sequences like ' ' (two characters) to actual newlines (one character), ' ' to tabs, '\' to backslashes, etc. Useful for processing strings where escape sequences are represented literally rather than interpreted.
- upper
Converts the string to uppercase. This node returns a copy of the string with all lowercase characters converted to uppercase.
- zfill
Pads the string with zeros on the left. This node returns a copy of the string left filled with ASCII '0' digits to make a string of length width. A leading sign prefix ('+'/'-') is handled by inserting the padding after the sign character rather than before.
### Basic/STRING/is
- isalnum
Checks if all characters in the string are alphanumeric. This node returns True if all characters in the string are alphanumeric (letters or numbers) and there is at least one character, otherwise False.
- isalpha
Checks if all characters in the string are alphabetic. This node returns True if all characters in the string are alphabetic (letters) and there is at least one character, otherwise False.
- isascii
Checks if all characters in the string are ASCII characters. This node returns True if all characters in the string are in the ASCII character set and there is at least one character, otherwise False.
- isdecimal
Checks if all characters in the string are decimal characters. This node returns True if all characters in the string are decimal characters and there is at least one character, otherwise False. Decimal characters include digit characters and characters that can be used to form decimal numbers in various locales.
- isdigit
Checks if all characters in the string are digits. This node returns True if all characters in the string are digits and there is at least one character, otherwise False.
- isidentifier
Checks if the string is a valid identifier in Python. This node returns True if the string is a valid Python identifier, otherwise False. A valid identifier must start with a letter or underscore and can only contain letters, digits, or underscores.
- islower
Checks if all characters in the string are lowercase. This node returns True if all cased characters in the string are lowercase and there is at least one cased character, otherwise False.
- isnumeric
Checks if all characters in the string are numeric. This node returns True if all characters in the string are numeric and there is at least one character, otherwise False. Numeric characters include digit characters and characters that have the Unicode numeric value property.
- isprintable
Checks if all characters in the string are printable. This node returns True if all characters in the string are printable or the string is empty, otherwise False. Printable characters are those which are not control characters.
- isspace
Checks if all characters in the string are whitespace. This node returns True if all characters in the string are whitespace and there is at least one character, otherwise False.
- istitle
Checks if the string is titlecased. This node returns True if the string is titlecased and there is at least one character, otherwise False. A titlecased string has all words start with an uppercase character and continue with lowercase.
- isupper
Checks if all characters in the string are uppercase. This node returns True if all cased characters in the string are uppercase and there is at least one cased character, otherwise False.
### Basic/STRING/regex
- find all (data list)
Returns all non-overlapping matches of a pattern in the string as a list of strings.
- find all (LIST)
Returns all non-overlapping matches of a pattern in the string as a list of strings.
- search groups (data list)
Searches the string for a match to the pattern and returns a LIST of match groups. If no match is found, it returns an empty LIST.
- search groups (LIST)
Searches the string for a match to the pattern and returns a data list of match groups. If no match is found, it returns an empty data list.
- search named groups
Searches the string with the given pattern and returns a DICT of named groups. If no match is found, it returns an empty DICT.
- split (data list)
Splits the string at each match of the pattern and returns a list of substrings.
- split (LIST)
Splits the string at each match of the pattern and returns a list of substrings.
- substitute
Substitutes matches of the pattern in the string with a replacement string.
- test
Tests whether a given regex pattern matches any part of the input string. Returns True if a match is found, otherwise False.
### Basic/tensor
- Tensor Binary Op
Performs binary operations between two tensors or a tensor and a scalar.
- Tensor Create
Creates a PyTorch tensor from various input types. Can accept numbers, lists, or existing tensors.
- Tensor Info
Returns information about a tensor.
- Tensor Join
Joins multiple tensors (concatenate or stack).
- Tensor Permute
Permutes tensor dimensions.
- Tensor Reshape
Reshapes a tensor.
- Tensor Slice
Slices a tensor using a slice string (e.g., ':, 0:10, 5').
- Tensor Unary Op
Performs unary operations on a tensor.
### Basic/time
- Add Time Delta
Adds a TIMEDELTA (duration) to a DATETIME object, resulting in a new DATETIME.
- Create Time Delta
Creates a TIMEDELTA object, which represents a duration and can be used for date calculations.
- Extract Time Components
Extracts individual components (year, month, day, hour, etc.) from a DATETIME object. Weekday is returned as an integer, where Monday is 0 and Sunday is 6.
- Format Time String
Formats a DATETIME object into a string using a specified format code. Common format codes: %Y (year), %m (month), %d (day), %H (hour), %M (minute), %S (second).
- Parse Time String
Parses a string containing a date and time into a DATETIME object, using a specified format code.
- Subtract Time Delta
Subtracts a TIMEDELTA (duration) from a DATETIME object, resulting in a new DATETIME.
- Time Difference
Calculates the difference between two DATETIME objects, returning a TIMEDELTA object.
- Time Now
Returns the current time and date as a DATETIME object. Note: Output changes for every run, providing a fresh timestamp each time.
- Time Now (UTC)
Returns the current time and date as a DATETIME object, in the UTC timezone. Note: Output changes for every run, providing a fresh timestamp each time.
- Time to Unix Timestamp
Converts a DATETIME object to a Unix timestamp (a float representing seconds since the epoch).
- Unix Timestamp to Time
Converts a Unix timestamp (float or int) to a DATETIME object.
---
# ComfyQR on Comfy Cloud
> ComfyQR is preinstalled on Comfy Cloud — 3 nodes ready to run on managed GPUs. Description unavailable.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
—
**Publisher**
—
**Downloads**
—
**GitHub stars**
—
**Latest version**
—
**License**
—
**Last updated**
—
## Nodes in this pack
### ComfyQR
- QR Code
- QR Code (Conformed to Image Size)
- QR Code (Split)
---
# AudioTools on Comfy Cloud
> AudioTools is preinstalled on Comfy Cloud — 29 nodes ready to run on managed GPUs. Description unavailable.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
—
**Publisher**
—
**Downloads**
—
**GitHub stars**
—
**Latest version**
—
**License**
—
**Last updated**
—
## Nodes in this pack
### ⚪Lum3on/AudioTools/AI
- Speech Denoise (AI)
- Speech-to-Text + SRT (Whisper)
- Stem Separator (AI)
### ⚪Lum3on/AudioTools/Analysis
- Audio-Reactive Envelope
- BPM Detector / Reactive
- Loudness Meter (LUFS)
### ⚪Lum3on/AudioTools/Effects
- Delay / Echo
- Fade In
- Fade Out
- Pitch Shift / Time Stretch
- Reverb
### ⚪Lum3on/AudioTools/Processing
- Amplify / Gain
- De-Esser
- De-Hum (50/60Hz)
- De-Plosive (Low Cut)
- Mix Audio Tracks
- Noise Gate
- Normalize Audio
- Parametric EQ for Voice
- Remove Silence
- Standardize Audio (Format/Channels)
- Trim Audio
- Vocal Compressor
### ⚪Lum3on/AudioTools/Utility
- Concatenate Audio
- Pad With Silence
- Stereo Panner
### ⚪Lum3on/AudioTools/Visualization
- Compare Waveforms
- Display Waveform
- Show Audio Info
---
# ComfyUI_essentials on Comfy Cloud
> ComfyUI_essentials is preinstalled on Comfy Cloud — 62 nodes ready to run on managed GPUs. Essential nodes that are weirdly missing from ComfyUI core. With few exceptions they are new features and not commodities.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/cubiq/ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials)
**Publisher**
Matteo
**Downloads**
2,947,151
**GitHub stars**
1,178
**Latest version**
1.1.0
**License**
{"file": "LICENSE"}
**Last updated**
Aug 7, 2024
## Nodes in this pack
### essentials/conditioning
- CLIPTextEncodeSDXL+
- ConditioningCombineMultiple+
- SD3NegativeConditioning+
### essentials/image analysis
- ImageEnhanceDifference+
### essentials/image batch
- ImageBatchMultiple+
- ImageExpandBatch+
- ImageFromBatch+
- ImageListToBatch+
### essentials/image manipulation
- ImageComposite+
- ImageCompositeFromMaskBatch+
- ImageCrop+
- ImageFlip+
- ImageRandomTransform+
- ImageRemoveBackground+
- ImageResize+
- ImageSeamCarving+
- ImageTile+
- ImageUntile+
- RemBGSession+
- TransparentBGSession+
### essentials/image processing
- ImageApplyLUT+
- ImageCASharpening+
- ImageColorMatch+
- ImageColorMatchAdobe+
- ImageDesaturate+
- ImageHistogramMatch+
- ImagePosterize+
- PixelOEPixelize+
### essentials/image utils
- GetImageSize+
- ImagePreviewFromLatent+
Saves the input images to your ComfyUI output directory.
- ImageRemoveAlpha+
- ImageToDevice+
- NoiseFromImage+
### essentials/mask
- MaskBlur+
- MaskBoundingBox+
- MaskFix+
- MaskFlip+
- MaskFromColor+
- MaskFromList+
- MaskFromRGBCMYBW+
- MaskFromSegmentation+
- MaskPreview+
Saves the input images to your ComfyUI output directory.
- MaskSmooth+
- TransitionMask+
### essentials/mask batch
- MaskBatch+
- MaskExpandBatch+
- MaskFromBatch+
### essentials/sampling
- FluxSamplerParams+
- InjectLatentNoise+
- KSamplerVariationsStochastic+
- KSamplerVariationsWithNoise+
- PlotParameters+
### essentials/segmentation
- ApplyCLIPSeg+
- LoadCLIPSegModels+
### essentials/text
- DrawText+
### essentials/utilities
- BatchCount+
- ConsoleDebug+
- DebugTensorShape+
- ModelCompile+
- RemoveLatentMask+
- SDXLEmptyLatentSizePicker+
- SimpleMath+
---
# comfyui_face_parsing on Comfy Cloud
> comfyui_face_parsing is preinstalled on Comfy Cloud — 27 nodes ready to run on managed GPUs. This is a set of custom nodes for ComfyUI. The nodes utilize the [a/face parsing model](https://huggingface.co/jonathandinu/face-parsing) to parse face and provide detailed segmantation. To improve face segmantation accuracy, [a/yolov8 face model](https://huggingface.co/Bingsu/adetailer/) is used to first extract face from an image. There are also auxiliary nodes for image and mask processing. A guided filter is also provided for skin smoothing.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/Ryuukeisyou/comfyui_face_parsing](https://github.com/Ryuukeisyou/comfyui_face_parsing)
**Publisher**
Ryuukeisyou
**Downloads**
85,133
**GitHub stars**
194
**Latest version**
1.0.5
**License**
{"file": "LICENSE"}
**Last updated**
Feb 4, 2025
## Nodes in this pack
### face_parsing
- BBoxDecompose(FaceParsing)
- BBoxDetect(FaceParsing)
- BBoxDetectorLoader(FaceParsing)
- BBoxListItemSelect(FaceParsing)
- BBoxResize(FaceParsing)
- ColorAdjust(FaceParsing)
- FaceParse(FaceParsing)
- FaceParsingModelLoader(FaceParsing)
- FaceParsingProcessorLoader(FaceParsing)
- FaceParsingResultsParser(FaceParsing)
- GuidedFilter(FaceParsing)
- ImageCropWithBBox(FaceParsing)
- ImageCropWithBBoxList(FaceParsing)
- ImageInsertWithBBox(FaceParsing)
- ImageListSelect(FaceParsing)
- ImagePadWithBBox(FaceParsing)
- ImageResizeCalculator(FaceParsing)
- ImageResizeWithBBox(FaceParsing)
- ImageSize(FaceParsing)
- MaskBatchComposite(FaceParsing)
- MaskBlackOut(FaceParsing)
- MaskBorderDissolve(FaceParsing)
- MaskComposite(FaceParsing)
- MaskCropWithBBox(FaceParsing)
- MaskInsertWithBBox(FaceParsing)
- MaskListSelect(FaceParsing)
- MaskToBBoxList(FaceParsing)
---
# Fill-Nodes on Comfy Cloud
> Fill-Nodes is preinstalled on Comfy Cloud — 46 nodes ready to run on managed GPUs. Description unavailable.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
—
**Publisher**
—
**Downloads**
—
**GitHub stars**
—
**Latest version**
—
**License**
—
**Last updated**
—
## Nodes in this pack
### 🏵️Fill Nodes/Audio
- FL Audio Beat Visualizer
- FL Audio BPM Analyzer
- FL Audio Crop
- FL Audio Drum Detector
- FL Audio Envelope Visualizer
- FL Audio Music Video Sequencer
- FL Audio Reactive Brightness
- FL Audio Reactive Edge Glow
- FL Audio Reactive Envelope
- FL Audio Reactive Saturation
- FL Audio Reactive Scale
- FL Audio Reactive Speed
- FL Audio Segment Extractor
- FL Audio Separation
- FL Audio Shot Iterator
### 🏵️Fill Nodes/experiments
- FL Color Picker
### 🏵️Fill Nodes/games
- FL BulletHell Game
- FL Tetris Game
### 🏵️Fill Nodes/Image
- FL Anime Line Extractor
- FL Image Add To Batch
- FL Image Blank
- FL Image Crop
- FL Image Size
### 🏵️Fill Nodes/Ksamplers
- FL Sampler String XYZ
- FL Scheduler String XYZ
### 🏵️Fill Nodes/PDF
- FL Text To PDF
### 🏵️Fill Nodes/Prompting
- FL MadLib Generator
- FL Prompt Basic
- FL Prompt Multi
- FL Prompt Selector
- FL Prompt Selector Basic
### 🏵️Fill Nodes/Utility
- FL Float
- FL Float to Int
- FL Int to Float
- FL Math
- FL Path Type Checker
- FL Random Number
- FL Switch
FL_Switch allows you to choose between two processing paths based on a switch value. The node only evaluates the path that is selected by the switch value. - If switch = False: on_false is evaluated and returned as output, on_true is not evaluated - If switch = True: on_true is evaluated and returned as output, on_false is not evaluated This node is useful for creating conditional workflows where you want to process data differently based on certain conditions, and you only want the selected path to execute.
- FL Switch Big
FL_Switch_Big allows you to choose between multiple processing paths based on a switch condition. The node only evaluates the path that is selected by the switch condition. - The switch_condition is compared against each case value - If a match is found, the corresponding input is evaluated and returned - If no match is found, input_default is evaluated and returned - Inputs that don't match the condition are not evaluated This node is useful for creating conditional workflows with multiple branches.
### 🏵️Fill Nodes/VFX
- FL Image Pixelator
### 🏵️Fill Nodes/Video
- FL Video Cadence Compile
Takes a cadence list (frame counts) and a directory of video files. Crops each video from the middle to match its cadence frame count. Concatenates all processed frames into a single image batch. Handles errors like count mismatches or dimension inconsistencies.
### 🏵️Fill Nodes/WIP
- FL Animated Shape Patterns
Creates animated geometric shape patterns with various algorithms. Generates a batch of images with shapes arranged in interesting patterns.
- FL Wan Video Blender
Blends two input videos with a cross-fade. The resolution of the second clip is resized to match the first.
- FL Wan Video Blender 3-Way
Blends three videos with cross-fade transitions. Perfect for input → continuation → end workflows. All videos resized to match first video.
- FL Wan Video Continuation Blender
Blends videos with a continuation video that has embedded overlaps. Automatically strips overlap frames to prevent duplication. Use this after FL_WanVideoContinue.
- FL Wan Video Continue
Creates a continuation video with symmetrical overlap - uses last N frames from input video at start and first N frames from end video at end.
---
# ComfyUI_IPAdapter_plus on Comfy Cloud
> ComfyUI_IPAdapter_plus is preinstalled on Comfy Cloud — 30 nodes ready to run on managed GPUs. ComfyUI reference implementation for the IPAdapter models. The IPAdapter are very powerful models for image conditioning. The style and composition of a reference can be easily transferred to the generation. Think of it as a 1-image lora.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/cubiq/ComfyUI_IPAdapter_plus](https://github.com/cubiq/ComfyUI_IPAdapter_plus)
**Publisher**
Matteo
**Downloads**
1,511,857
**GitHub stars**
6,125
**Latest version**
2.0.0
**License**
GPL-3.0 license
**Last updated**
Jun 5, 2024
## Nodes in this pack
### ipadapter
- IPAdapter
- IPAdapterAdvanced
- IPAdapterBatch
- IPAdapterUnifiedLoader
### ipadapter/dev
- IPAdapterMS
### ipadapter/embeds
- IPAdapterCombineEmbeds
- IPAdapterEmbeds
- IPAdapterEmbedsBatch
- IPAdapterEncoder
- IPAdapterLoadEmbeds
- IPAdapterSaveEmbeds
### ipadapter/faceid
- IPAAdapterFaceIDBatch
- IPAdapterFaceID
- IPAdapterUnifiedLoaderFaceID
### ipadapter/loaders
- IPAdapterInsightFaceLoader
- IPAdapterModelLoader
- IPAdapterUnifiedLoaderCommunity
### ipadapter/params
- IPAdapterCombineParams
- IPAdapterFromParams
- IPAdapterRegionalConditioning
### ipadapter/style_composition
- IPAdapterStyleComposition
- IPAdapterStyleCompositionBatch
### ipadapter/tiled
- IPAdapterTiled
- IPAdapterTiledBatch
### ipadapter/utils
- IPAdapterCombineWeights
- IPAdapterNoise
- PrepImageForClipVision
### ipadapter/weights
- IPAdapterPromptScheduleFromWeightsStrategy
- IPAdapterWeights
- IPAdapterWeightsFromStrategy
---
# ComfyUI_LayerStyle_Advance on Comfy Cloud
> ComfyUI_LayerStyle_Advance is preinstalled on Comfy Cloud — 77 nodes ready to run on managed GPUs. The nodes detached from ComfyUI Layer Style are mainly those with complex requirements for dependency packages.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/chflame163/ComfyUI_LayerStyle_Advance](https://github.com/chflame163/ComfyUI_LayerStyle_Advance)
**Publisher**
chflame163
**Downloads**
187,743
**GitHub stars**
710
**Latest version**
2.0.38
**License**
{"text": "MIT License"}
**Last updated**
Apr 8, 2026
## Nodes in this pack
### 😺dzNodes/LayerMask
- LayerMask: BBoxJoin
- LayerMask: BenUltra
- LayerMask: BiRefNetUltra
- LayerMask: BiRefNetUltraV2
- LayerMask: DrawBBoxMask
- LayerMask: DrawBBoxMaskV2
- LayerMask: EVFSAMUltra
- LayerMask: Florence2Ultra
- LayerMask: HumanPartsUltra
- LayerMask: LoadBenModel
- LayerMask: LoadBiRefNetModel
- LayerMask: LoadBiRefNetModelV2
- LayerMask: LoadFlorence2Model
- LayerMask: LoadSAM2Model
- LayerMask: LoadSegmentAnythingModels
- LayerMask: MaskByDifferent
- LayerMask: MediapipeFacialSegment
- LayerMask: ObjectDetectorFL2
- LayerMask: ObjectDetectorGemini
- LayerMask: ObjectDetectorGeminiV2
- LayerMask: ObjectDetectorMask
- LayerMask: ObjectDetectorYOLO8
- LayerMask: ObjectDetectorYOLOWorld
- LayerMask: PersonMaskUltra
- LayerMask: PersonMaskUltra V2
- LayerMask: SAM2Ultra
- LayerMask: SAM2UltraV2
- LayerMask: SAM2VideoUltra
- LayerMask: SegmentAnythingUltra
- LayerMask: SegmentAnythingUltra V2
- LayerMask: SegmentAnythingUltra V3
- LayerMask: TransparentBackgroundUltra
- LayerMask: YoloV8Detect
### 😺dzNodes/LayerUtility
- LayerUtility: Collage
- LayerUtility: DeepSeekAPI
- LayerUtility: DeepSeekAPIV2
- LayerUtility: Gemini
- LayerUtility: GeminiImageEdit
- LayerUtility: GeminiV2
- LayerUtility: GetColorTone
- LayerUtility: GetColorToneV2
- LayerUtility: ImageAutoCrop
- LayerUtility: ImageAutoCrop V2
- LayerUtility: ImageAutoCrop V3
- LayerUtility: ImageRewardFilter
- LayerUtility: JimengI2IAPI
- LayerUtility: JoyCaption2
- LayerUtility: JoyCaption2ExtraOptions
- LayerUtility: JoyCaption2Split
- LayerUtility: JoyCaptionBeta1
- LayerUtility: JoyCaptionBeta1ExtraOptions
- LayerUtility: LaMa
- LayerUtility: LlamaVision
- LayerUtility: LoadJoyCaption2Model
- LayerUtility: LoadJoyCaptionBeta1Model
- LayerUtility: LoadSmolLM2Model
- LayerUtility: LoadSmolVLMModel
- LayerUtility: PhiPrompt
- LayerUtility: SmolLM2
- LayerUtility: SmolVLM
- LayerUtility: ZhipuGLM4
- LayerUtility: ZhipuGLM4V
### 😺dzNodes/LayerUtility/Prompt
- LayerUtility: Florence2Image2Prompt
- LayerUtility: PromptEmbellish
- LayerUtility: PromptTagger
- LayerUtility: QWenImage2Prompt
- LayerUtility: UserPromptGeneratorReplaceWord
- LayerUtility: UserPromptGeneratorTxt2ImgPrompt
- LayerUtility: UserPromptGeneratorTxt2ImgPromptWithReference
### 😺dzNodes/LayerUtility/SystemIO
- LayerUtility: AddBlindWaterMark
- LayerUtility: CreateQRCode
- LayerUtility: DecodeQRCode
- LayerUtility: LoadPSD
- LayerUtility: SaveImagePlus
- LayerUtility: SaveImagePlusV2
- LayerUtility: SD3NegativeConditioning
- LayerUtility: ShowBlindWaterMark
---
# ComfyUI_LayerStyle on Comfy Cloud
> ComfyUI_LayerStyle is preinstalled on Comfy Cloud — 19 nodes ready to run on managed GPUs. A set of nodes for ComfyUI it generate image like Adobe Photoshop's Layer Style. the Drop Shadow is first completed node, and follow-up work is in progress.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/chflame163/ComfyUI_LayerStyle](https://github.com/chflame163/ComfyUI_LayerStyle)
**Publisher**
chflame163
**Downloads**
2,108,889
**GitHub stars**
3,159
**Latest version**
2.0.40
**License**
{"text": "MIT License"}
**Last updated**
Aug 1, 2026
## Nodes in this pack
### 😺dzNodes/LayerFilter
- LayerFilter: HDR Effects
### 😺dzNodes/LayerMask
- LayerMask: DrawRoundedRectangle
- LayerMask: Segformer Clothes Setting
- LayerMask: Segformer Fashion Setting
### 😺dzNodes/LayerUtility
- LayerUtility: ColorImage
- LayerUtility: GradientImage
### 😺dzNodes/LayerUtility/Data
- LayerUtility: Any Rerouter
- LayerUtility: Boolean
- LayerUtility: Float
- LayerUtility: Gray Value
- LayerUtility: HSV Value
- LayerUtility: Integer
- LayerUtility: PrintInfo
- LayerUtility: RGB Value
- LayerUtility: Seed
- LayerUtility: String
- LayerUtility: TextBox
- LayerUtility: TextJoin
- LayerUtility: TextJoinV2
---
# SchemaNodes on Comfy Cloud
> SchemaNodes is preinstalled on Comfy Cloud — 8 nodes ready to run on managed GPUs. Description unavailable.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
—
**Publisher**
—
**Downloads**
—
**GitHub stars**
—
**Latest version**
—
**License**
—
**Last updated**
—
## Nodes in this pack
### inout/schema
- Schema Audio Parameter
- Schema Boolean Parameter
- Schema Enum Parameter
- Schema Float Parameter
- Schema Image Parameter
- Schema Integer Parameter
- Schema String Parameter
- Schema Video Parameter
---
# ComfyUI_Steudio on Comfy Cloud
> ComfyUI_Steudio is preinstalled on Comfy Cloud — 10 nodes ready to run on managed GPUs. Divide and Conquer Node Suite: It calculates the optimal upscale resolution and seamlessly divides the image into tiles, ready for individual processing using your preferred workflow. After processing, the tiles are seamlessly merged into a larger image, offering sharper and more detailed visuals.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/Steudio/ComfyUI_Steudio](https://github.com/Steudio/ComfyUI_Steudio)
**Publisher**
steudio
**Downloads**
76,766
**GitHub stars**
147
**Latest version**
2.0.5
**License**
{"file": "LICENSE.txt"}
**Last updated**
May 22, 2025
## Nodes in this pack
### Steudio/Divide and Conquer
- Combine Tiles
- Divide and Conquer Algorithm
Calculate the best dimensions and optionally upscale an image while maintaining minimum tile overlap and scale factor constraints. Steudio
- Divide Image and Select Tile
tile 0 = All tiles tile # = Tile #
### Steudio/Utils
- Display UI
- Load Images into List
- Ratio Calculator
- Ratio to Size
- Seed Shifter
A simple and effective way to generate a “batch” of images with reproducible seed. Steudio
- Sequence Generator
x...y+z | Generates a sequence of numbers from x to y with a step of z. x...y#z | Generates z evenly spaced numbers between x and y. x,y,z | Generates a list of x, y, z.
- Simple Config
---
# ComfyUI_UltimateSDUpscale on Comfy Cloud
> ComfyUI_UltimateSDUpscale is preinstalled on Comfy Cloud — 3 nodes ready to run on managed GPUs. ComfyUI nodes for the Ultimate Stable Diffusion Upscale script by Coyote-A.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/ssitu/ComfyUI_UltimateSDUpscale](https://github.com/ssitu/ComfyUI_UltimateSDUpscale)
**Publisher**
ssit
**Downloads**
1,794,278
**GitHub stars**
1,554
**Latest version**
1.7.2
**License**
{"file": "LICENSE"}
**Last updated**
Mar 28, 2026
## Nodes in this pack
### image/upscaling
- UltimateSDUpscale
Upscales an image and runs image-to-image on tiles from the input image.
- UltimateSDUpscaleCustomSample
Runs image-to-image on tiles from the input image.
- UltimateSDUpscaleNoUpscale
Runs image-to-image on tiles from the input image.
---
# ComfyUI-Advanced-ControlNet on Comfy Cloud
> ComfyUI-Advanced-ControlNet is preinstalled on Comfy Cloud — 2 nodes ready to run on managed GPUs. Nodes for scheduling ControlNet strength across timesteps and batched latents, as well as applying custom weights and attention masks.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet)
**Publisher**
Kosinkadink
**Downloads**
704,660
**GitHub stars**
1,008
**Latest version**
1.6.0
**License**
{"file": "LICENSE"}
**Last updated**
Jul 28, 2026
## Nodes in this pack
### Adv-ControlNet 🛂🅐🅒🅝/Reference
- Reference ControlNet (Finetune) 🛂🅐🅒🅝
### Adv-ControlNet 🛂🅐🅒🅝/SparseCtrl
- SparseCtrl Index Method 🛂🅐🅒🅝
---
# ComfyUI-AnimateDiff-Evolved on Comfy Cloud
> ComfyUI-AnimateDiff-Evolved is preinstalled on Comfy Cloud — 11 nodes ready to run on managed GPUs. Improved AnimateDiff integration for ComfyUI.
[Back to all packs](https://comfy.org/cloud/supported-nodes)
**View repository**
[https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved](https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved)
**Publisher**
Kosinkadink
**Downloads**
612,237
**GitHub stars**
3,536
**Latest version**
1.5.7
**License**
{"file": "LICENSE"}
**Last updated**
Mar 30, 2026
## Nodes in this pack
### Animate Diff 🎭🅐🅓/conditioning
- Timesteps Conditioning 🎭🅐🅓 Deprecated
### Animate Diff 🎭🅐🅓/context opts/context extras/contextref
- ContextRef Mode◆First 🎭🅐🅓
- ContextRef Mode◆Indexes 🎭🅐🅓
- ContextRef Mode◆Sliding 🎭🅐🅓
- ContextRef Tune◆Attn 🎭🅐🅓
- ContextRef Tune◆Attn+Adain 🎭🅐🅓
### Animate Diff 🎭🅐🅓/extras
- Empty Latent Image (Big Batch) 🎭🅐🅓
### Animate Diff 🎭🅐🅓/iteration opts
- Default Iteration Options 🎭🅐🅓
### Animate Diff 🎭🅐🅓/sample settings/image inject
- Image Injection Options 🎭🅐🅓
### Animate Diff 🎭🅐🅓/scheduling
- Float to Floats 🎭🅐🅓
- Value Scheduling 🎭🅐🅓
Create a list of values with automatic interpolation.
[-]
Format:
Scheduling supports two formats: JSON and pythonic.
[-]
JSON:
"idx": float/int_value, ...
[-]
pythonic:
idx = float/int_value, ...
The idx is the index of the frame - first frame is 0, last frame is max_frames-1. An idx may be the following:
[-]
allowed idxs:
[-]
single: A positive integer (e.g. 0, 2) schedules value for frame. A negative integer (e.g. -1, -5) schedules value for frame from the end (-1 would be the last frame). A decimal (e.g. 0.5, 1.0) selects frame based relative location in whole schedule (0.5 would be halfway, 1.0 would be last frame).
[-]
range: Using rules above, single:single chooses uninterpolated prompts from start idx (included) to end idx (excluded). Examples -> 0:12, 0:-5, 2:0.5
[-]
hold: Putting a colon after a single idx stops interpolation until the next provided index. Examples -> 0:, 0.5:, 16:
The values can be written without any special formatting.
[-]
pythonic:
0 = 1.0,
16 = 1.3
[-]
Inputs:
[-]
values: Write your values here.
[-]
max_length: Used to select the intended length of schedule. If set to 0, will use the largest index in the schedule as max_length, but will disable relative indexes (negative and decimal).
[-]
print_schedule: When True, prints output values for each frame.
Split the processing of a very long video into sets of smaller Meta Batches
The Meta Batch Manager allows for extremely long input videos to be processed when all other methods for fitting the content in RAM fail. It does not effect VRAM usage.
It must be connected to at least one Input (a Load Video or Load Images) AND at least one Video Combine
It functions by holding both the inputs and ouputs open between executions, and automatically requeue's the workflow until one of the inputs is unable to provide additional images.
Because each sub execution only contains a subset of the total frames, each sub execution creates a hard window which temporal smoothing can not be applied across. This results in jumps in the output.
[-]
Outputs:
[-]
meta_batch: Add all connected nodes to this Meta Batch
[-]
Widgets:
[-]
frames_per_batch: How many frames to process for each sub execution. If loading as image, each frame will use about 50MB of RAM (not VRAM), and this can safely be set in the 100-1000 range, depending on available memory. When loading and combining from latent space (no blue image noodles exist), this value can be much higher, around the 2,000 to 20,000 range
Load a single image from a given path
[-]
Inputs:
[-]
vae: (optional) If provided the node will output latents instead of images.
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
MASK: The alpha channel of the loaded images.
[-]
Widgets:
[-]
image: The image file to be loaded.
This is a VHS_PATH input. When edited, it provides a list of possible valid files or directories
The current top-most completion may be selected with Tab
You can navigate up a directory by pressing Ctrl+B (or Ctrl+W if supported by browser)
The filter on suggested file types can be disabled by pressing Ctrl+G.
If converted to an input, this functions as a string
[-]
force_size: Allows for conveniently scaling the input without requiring an additional node. Provides options to maintain aspect ratio or conveniently target common training formats for Animate Diff
[-]
custom_width: Allows for an arbitrary width to be entered, cropping to maintain aspect ratio if both are set
[-]
custom_height: Allows for an arbitrary height to be entered, cropping to maintain aspect ratio if both are set
[-]
videopreview: Displays a preview for the selected video input. Will only be shown if Advanced Previews is enabled. This preview will reflect the image_load_cap, skip_first_images, and select_every_nth values chosen. Additional preview options can be accessed with right click.
Loads a sequence of images from a subdirectory of the input folder
[-]
Inputs:
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long sequences into sub batches. See the documentation for Meta Batch Manager
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
MASK: The alpha channel of the loaded images.
[-]
frame_count: The length of images just returned
[-]
Widgets:
[-]
directory: The directory images will be loaded from. Filtered to process jpg, png, ppm, bmp, tif, and webp files
[-]
image_load_cap: The maximum number of images to load. If 0, all images are loaded.
[-]
start_time: A timestamp, in seconds from the start of the video, to start loading frames from.
[-]
choose folder to upload: An upload button is provided to upload a local folder containing images to the input folder
[-]
videopreview: Displays a preview for the selected video input. Will only be shown if Advanced Previews is enabled. This preview will reflect the image_load_cap, skip_first_images, and select_every_nth values chosen. Additional preview options can be accessed with right click.
Loads a sequence of images from an arbitrary path
[-]
Inputs:
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long sequences into sub batches. See the documentation for Meta Batch Manager
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
MASK: The alpha channel of the loaded images.
[-]
frame_count: The length of images just returned
[-]
Widgets:
[-]
directory: The directory images will be loaded from. Filtered to process jpg, png, ppm, bmp, tif, and webp files
This is a VHS_PATH input. When edited, it provides a list of possible valid files or directories
The current top-most completion may be selected with Tab
You can navigate up a directory by pressing Ctrl+B (or Ctrl+W if supported by browser)
The filter on suggested file types can be disabled by pressing Ctrl+G.
If converted to an input, this functions as a string
[-]
image_load_cap: The maximum number of images to load. If 0, all images are loaded.
[-]
skip_first_images: A number of images which are discarded before producing output.
[-]
select_every_nth: Keeps only the first of every n frames and discard the rest.
[-]
videopreview: Displays a preview for the selected video input. Will only be shown if Advanced Previews is enabled. This preview will reflect the image_load_cap, skip_first_images, and select_every_nth values chosen. Additional preview options can be accessed with right click.
Loads a video from the input folder
[-]
Inputs:
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long sequences into sub batches. See the documentation for Meta Batch Manager
[-]
vae: (optional) If provided the node will output latents instead of images. This drastically reduces the required RAM (not VRAM) when working with long (100+ frames) sequences
Using this is strongly encouraged unless connecting to a node that requires a blue image connection such as Apply Controllnet
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
frame_count: The length of images just returned
[-]
audio: The audio from the loaded video
[-]
video_info: Exposes additional info about the video such as the source frame rate, or the total length
[-]
LATENT: The loaded images pre-converted to latents. Only available when a vae is connected
[-]
Widgets:
[-]
video: The video file to be loaded. Lists all files with a video extension in the ComfyUI/Input folder
[-]
force_rate: Drops or duplicates frames so that the produced output has the target frame rate. Many motion models are trained on videos of a specific frame rate and will give better results if input matches that frame rate. If set to 0, all frames are returned. May give unusual results with inputs that have a variable frame rate like animated gifs. Reducing this value can also greatly reduce the execution time and memory requirements.
[-]
force_size: Previously was used to provide suggested resolutions. Instead, custom_width and custom_height can be disabled by setting to 0.
[-]
custom_width: Allows for an arbitrary width to be entered, cropping to maintain aspect ratio if both are set
[-]
custom_height: Allows for an arbitrary height to be entered, cropping to maintain aspect ratio if both are set
[-]
frame_load_cap: The maximum number of frames to load. If 0, all frames are loaded.
[-]
skip_first_frames: A number of frames which are discarded before producing output.
[-]
select_every_nth: Similar to frame rate. Keeps only the first of every n frames and discard the rest. Has better compatibility with variable frame rate inputs such as gifs. When combined with force_rate, select_every_nth_applies after force_rate so the resulting output has a frame rate equivalent to force_rate/select_every_nth. select_every_nth does not apply to skip_first_frames
[-]
format: Updates other widgets so that only values supported by the given format can be entered and provides recommended defaults.
[-]
choose video to upload: An upload button is provided to upload local files to the input folder
[-]
videopreview: Displays a preview for the selected video input. If advanced previews is enabled, this preview will reflect the frame_load_cap, force_rate, skip_first_frames, and select_every_nth values chosen. If the video has audio, it will also be previewed when moused over. Additional preview options can be accessed with right click.
Loads a video from the input folder using ffmpeg instead of opencv
Provides faster execution speed, transparency support, and allows specifying start time in seconds
[-]
Inputs:
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long sequences into sub batches. See the documentation for Meta Batch Manager
[-]
vae: (optional) If provided the node will output latents instead of images. This drastically reduces the required RAM (not VRAM) when working with long (100+ frames) sequences
Using this is strongly encouraged unless connecting to a node that requires a blue image connection such as Apply Controllnet
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
mask: Transparency data from the loaded video
[-]
audio: The audio from the loaded video
[-]
video_info: Exposes additional info about the video such as the source frame rate, or the total length
[-]
LATENT: The loaded images pre-converted to latents. Only available when a vae is connected
[-]
Widgets:
[-]
video: The video file to be loaded. Lists all files with a video extension in the ComfyUI/Input folder
[-]
force_rate: Drops or duplicates frames so that the produced output has the target frame rate. Many motion models are trained on videos of a specific frame rate and will give better results if input matches that frame rate. If set to 0, all frames are returned. May give unusual results with inputs that have a variable frame rate like animated gifs. Reducing this value can also greatly reduce the execution time and memory requirements.
[-]
force_size: Previously was used to provide suggested resolutions. Instead, custom_width and custom_height can be disabled by setting to 0.
[-]
custom_width: Allows for an arbitrary width to be entered, cropping to maintain aspect ratio if both are set
[-]
custom_height: Allows for an arbitrary height to be entered, cropping to maintain aspect ratio if both are set
[-]
frame_load_cap: The maximum number of frames to load. If 0, all frames are loaded.
[-]
start_time: A timestamp, in seconds from the start of the video, to start loading frames from.
[-]
format: Updates other widgets so that only values supported by the given format can be entered and provides recommended defaults.
[-]
choose video to upload: An upload button is provided to upload local files to the input folder
[-]
videopreview: Displays a preview for the selected video input. If advanced previews is enabled, this preview will reflect the frame_load_cap, force_rate, skip_first_frames, and select_every_nth values chosen. If the video has audio, it will also be previewed when moused over. Additional preview options can be accessed with right click.
Loads a video from an arbitrary path using ffmpeg instead of opencv
Provides faster execution speed, transparency support, and allows specifying start time in seconds
[-]
Inputs:
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long sequences into sub batches. See the documentation for Meta Batch Manager
[-]
vae: (optional) If provided the node will output latents instead of images. This drastically reduces the required RAM (not VRAM) when working with long (100+ frames) sequences
Using this is strongly encouraged unless connecting to a node that requires a blue image connection such as Apply Controllnet
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
mask: Transparency data from the loaded video
[-]
audio: The audio from the loaded video
[-]
video_info: Exposes additional info about the video such as the source frame rate, or the total length
[-]
LATENT: The loaded images pre-converted to latents. Only available when a vae is connected
[-]
Widgets:
[-]
video: The video file to be loaded.
You can also select an image to load it as a single frame
This is a VHS_PATH input. When edited, it provides a list of possible valid files or directories
The current top-most completion may be selected with Tab
You can navigate up a directory by pressing Ctrl+B (or Ctrl+W if supported by browser)
The filter on suggested file types can be disabled by pressing Ctrl+G.
If converted to an input, this functions as a string
[-]
force_rate: Drops or duplicates frames so that the produced output has the target frame rate. Many motion models are trained on videos of a specific frame rate and will give better results if input matches that frame rate. If set to 0, all frames are returned. May give unusual results with inputs that have a variable frame rate like animated gifs. Reducing this value can also greatly reduce the execution time and memory requirements.
[-]
force_size: Previously was used to provide suggested resolutions. Instead, custom_width and custom_height can be disabled by setting to 0.
[-]
custom_width: Allows for an arbitrary width to be entered, cropping to maintain aspect ratio if both are set
[-]
custom_height: Allows for an arbitrary height to be entered, cropping to maintain aspect ratio if both are set
[-]
frame_load_cap: The maximum number of frames to load. If 0, all frames are loaded.
[-]
skip_first_frames: A number of frames which are discarded before producing output.
[-]
select_every_nth: Similar to frame rate. Keeps only the first of every n frames and discard the rest. Has better compatibility with variable frame rate inputs such as gifs. When combined with force_rate, select_every_nth_applies after force_rate so the resulting output has a frame rate equivalent to force_rate/select_every_nth. select_every_nth does not apply to skip_first_frames
[-]
format: Updates other widgets so that only values supported by the given format can be entered and provides recommended defaults.
[-]
videopreview: Displays a preview for the selected video input. Will only be shown if Advanced Previews is enabled. This preview will reflect the frame_load_cap, force_rate, skip_first_frames, and select_every_nth values chosen. If the video has audio, it will also be previewed when moused over. Additional preview options can be accessed with right click.
Loads a video from an arbitrary path
[-]
Inputs:
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long sequences into sub batches. See the documentation for Meta Batch Manager
[-]
vae: (optional) If provided the node will output latents instead of images. This drastically reduces the required RAM (not VRAM) when working with long (100+ frames) sequences
Using this is strongly encouraged unless connecting to a node that requires a blue image connection such as Apply Controllnet
[-]
Outputs:
[-]
IMAGE: The loaded images
[-]
frame_count: The length of images just returned
[-]
audio: The audio from the loaded video
[-]
video_info: Exposes additional info about the video such as the source frame rate, or the total length
[-]
LATENT: The loaded images pre-converted to latents. Only available when a vae is connected
[-]
Widgets:
[-]
video: The video file to be loaded.
You can also select an image to load it as a single frame
This is a VHS_PATH input. When edited, it provides a list of possible valid files or directories
The current top-most completion may be selected with Tab
You can navigate up a directory by pressing Ctrl+B (or Ctrl+W if supported by browser)
The filter on suggested file types can be disabled by pressing Ctrl+G.
If converted to an input, this functions as a string
[-]
force_rate: Drops or duplicates frames so that the produced output has the target frame rate. Many motion models are trained on videos of a specific frame rate and will give better results if input matches that frame rate. If set to 0, all frames are returned. May give unusual results with inputs that have a variable frame rate like animated gifs. Reducing this value can also greatly reduce the execution time and memory requirements.
[-]
force_size: Previously was used to provide suggested resolutions. Instead, custom_width and custom_height can be disabled by setting to 0.
[-]
custom_width: Allows for an arbitrary width to be entered, cropping to maintain aspect ratio if both are set
[-]
custom_height: Allows for an arbitrary height to be entered, cropping to maintain aspect ratio if both are set
[-]
frame_load_cap: The maximum number of frames to load. If 0, all frames are loaded.
[-]
skip_first_frames: A number of frames which are discarded before producing output.
[-]
select_every_nth: Similar to frame rate. Keeps only the first of every n frames and discard the rest. Has better compatibility with variable frame rate inputs such as gifs. When combined with force_rate, select_every_nth_applies after force_rate so the resulting output has a frame rate equivalent to force_rate/select_every_nth. select_every_nth does not apply to skip_first_frames
[-]
format: Updates other widgets so that only values supported by the given format can be entered and provides recommended defaults.
[-]
videopreview: Displays a preview for the selected video input. Will only be shown if Advanced Previews is enabled. This preview will reflect the frame_load_cap, force_rate, skip_first_frames, and select_every_nth values chosen. If the video has audio, it will also be previewed when moused over. Additional preview options can be accessed with right click.
Automates deletion of undesired outputs from a Video Combine node.
Video Combine produces a number of file outputs in addition to the final output. Some of these, such as a video file without audio included, are implementation limitations and are not feasible to solve. As an alternative, the Prune Outputs node is added to automate the deletion of these file outputs if they are not desired
[-]
Inputs:
[-]
filenames: A connection from a Video Combine node to indicate which outputs should be pruned
[-]
Widgets:
[-]
options: Which files should be deleted
[-]
Intermediate: Delete any files that were required for intermediate processing but are not the final output, like the no-audio output file when audio is included
[-]
Intermediate and Utility: Delete all produced files that aren't the final output, including the first frame png
Select a single filename from the VHS_FILENAMES output by a Video Combine and return it as a string
Take care when combining this node with Prune Outputs. The VHS_FILENAMES object is immutable and will always contain the full list of output files, but execution order is undefined behavior (currently, Prune Outputs will generally execute first) and SelectFilename may return a path to a file that no longer exists.
[-]
Inputs:
[-]
filenames: A VHS_FILENAMES from a Video Combine node
[-]
Outputs:
[-]
filename: A string representation of the full output path for the chosen file
[-]
Widgets:
[-]
index: The index of which file should be selected. The default, -1, chooses the most complete output
Experimental virtual node to select the most recently modified file from a given folder
Assists in the creation of workflows where outputs from one execution are used elsewhere in subsequent executions.
[-]
Inputs:
[-]
filename_prefix: A path which can consist of a combination of folders and a prefix which candidate files must match
[-]
filename_postfix: A string which chich the selected file must end with. Useful for limiting to a target extension.
[-]
Outputs:
[-]
Filename: A string representing a file path to the most recently modified file.
Unbatch a list of items into a single concatenated item
Useful for when you want a single video output from a complex workflow
Has no relation to the Meta Batch system of VHS
[-]
Inputs:
[-]
batched: Any input which may or may not be batched
[-]
Outputs:
[-]
unbatched: A single output element. Torch tensors are concatenated across dim 0, all other types are added which functions as concatenation for strings and arrays, but may give undesired results for other types
Combine an image sequence into a video
[-]
Inputs:
[-]
images: The images to be turned into a video
[-]
audio: (optional) audio to add to the video
[-]
meta_batch: (optional) Connect to a Meta Batch manager to divide extremely long image sequences into sub batches. See the documentation for Meta Batch Manager
[-]
vae: (optional) If provided, the node will take latents as input instead of images. This drastically reduces the required RAM (not VRAM) when working with long (100+ frames) sequences
Unlike on Load Video, this isn't always a strict upgrade over using a standalone VAE Decode.
If you have multiple Video Combine outputs, then the VAE decode will be performed for each output node increasing execution time
If you make any change to output settings on the Video Combine (such as changing the output format), the VAE decode will be performed again as the decoded result is (by design) not cached
[-]
Widgets:
[-]
frame_rate: The frame rate which will be used for the output video. Consider converting this to an input and connecting this to a Load Video with Video Info(Loaded)->fps. When including audio, failure to properly set this will result in audio desync
[-]
loop_count: The number of additional times the video should repeat. Can cause performance issues when used with long (100+ frames) sequences
[-]
filename_prefix: A prefix to add to the name of the output filename. This can include subfolders or format strings.
[-]
format: The output format to use. Formats starting with, 'image' are saved with PIL, but formats starting with 'video' utilize the video_formats system. 'video' options require ffmpeg and selecting one frequently adds additional options to the node.
[-]
pingpong: Play the video normally, then repeat the video in reverse so that it 'pingpongs' back and forth. This is frequently used to minimize the appearance of skips on very short animations.
[-]
save_output: Specifies if output files should be saved to the output folder, or the temporary output folder
[-]
videopreview: Displays a preview for the processed result. If advanced previews is enabled, the output is always converted to a format viewable from the browser. If the video has audio, it will also be previewed when moused over. Additional preview options can be accessed with right click.
[-]
Common Format Widgets:
[-]
crf: Determines how much to prioritize quality over filesize. Numbers vary between formats, but on each format that includes it, the default value provides visually loss less output
[-]
pix_fmt: The pixel format to use for output. Alternative options will often have higher quality at the cost of increased file size and reduced compatibility with external software.
[-]
yuv420p: The most common and default format
[-]
yuv420p10le: Use 10 bit color depth. This can improve color quality when combined with 16bit input color depth
[-]
yuva420p: Include transparency in the output video
[-]
input_color_depth: VHS supports outputting 16bit images. While this produces higher quality output, the difference usually isn't visible without postprocessing and it significantly increases file size and processing time.
[-]
save_metadata: Determines if metadata for the workflow should be included in the output video file
Splits information on a video into a numerous outputs
[-]
Inputs:
[-]
video_info: A connection to a Load Video node
[-]
Outputs:
[-]
source_fps🟨: The frame rate of the video
[-]
source_frame_count🟨: How many total frames the video contains before accounting for frame rate or select_every_nth
[-]
source_duration🟨: The length of images just returned in seconds
[-]
source_width🟨: The width
[-]
source_height🟨: The height
[-]
loaded_fps🟦: The frame rate after accounting for force_rate and select_every_nth. This output is of particular use as it can be connected to the converted frame_rate input of a Video Combine node to ensure audio remains synchronized.
[-]
loaded_frame_count🟦: The number of frames returned by the current execution. Identical to the frame_count returned by the node itself
[-]
loaded_duration🟦: The duration in seconds of returned images after accounting for frame_load_cap
[-]
loaded_width🟦: The width of the video after scaling. These coordinates are in image space even if loading to latent space
[-]
loaded_height🟦: The height of the video after scaling. These coordinates are in image space even if loading to latent space
Splits information on a video into a numerous outputs describing the file itself after accounting for load options
[-]
Inputs:
[-]
video_info: A connection to a Load Video node
[-]
Outputs:
[-]
loaded_fps🟦: The frame rate after accounting for force_rate and select_every_nth. This output is of particular use as it can be connected to the converted frame_rate input of a Video Combine node to ensure audio remains synchronized.
[-]
loaded_frame_count🟦: The number of frames returned by the current execution. Identical to the frame_count returned by the node itself
[-]
loaded_duration🟦: The duration in seconds of returned images after accounting for frame_load_cap
[-]
loaded_width🟦: The width of the video after scaling. This is the dimension of the corresponding image even if loading as a latent directly
[-]
loaded_height🟦: The height of the video after scaling. This is the dimension of the corresponding image even if loading as a latent directly
Splits information on a video into a numerous outputs describing the file itself without accounting for load options
[-]
Inputs:
[-]
video_info: A connection to a Load Video node
[-]
Outputs:
[-]
source_fps🟨: The frame rate of the video
[-]
source_frame_count🟨: How many total frames the video contains before accounting for frame rate or select_every_nth
[-]
source_duration🟨: The length of images just returned in seconds
[-]
source_width🟨: The original width
[-]
source_height🟨: The original height
utility function for compatibility with external nodes
VHS used to use an internal VHS_AUDIO format for routing audio between inputs and outputs. This format was intended to only be used internally and was designed with a focus on performance over ease of use. Since ComfyUI now has an internal AUDIO format, VHS now uses this format. However, some custom node packs were made that are external to both ComfyUI and VHS that use VHS_AUDIO. This node was added so that those external nodes can still function
[-]
Inputs:
[-]
audio: An input in the standardized AUDIO format
[-]
Outputs:
[-]
vhs_audio: An output in the legacy VHS_AUDIO format for use with external nodes
Loads an audio file from an arbitrary path
[-]
Outputs:
[-]
audio: The loaded audio
[-]
Widgets:
[-]
audio_file: The audio file to be loaded.
This is a VHS_PATH input. When edited, it provides a list of possible valid files or directories
The current top-most completion may be selected with Tab
You can navigate up a directory by pressing Ctrl+B (or Ctrl+W if supported by browser)
The filter on suggested file types can be disabled by pressing Ctrl+G.
If converted to an input, this functions as a string
[-]
seek_seconds: An offset from the start of the sound file that the audio should start from
Loads an audio file from the input directory
Very similar in functionality to the built-in LoadAudio. It was originally added before VHS swapped to use Comfy's internal AUDIO format, but provides the additional options for start time and duration
[-]
Outputs:
[-]
audio: The loaded audio
[-]
Widgets:
[-]
audio: The audio file to be loaded.
[-]
start_time: An offset from the start of the sound file that the audio should start from
[-]
duration: A maximum limit for the audio. Disabled if 0
[-]
choose audio to upload: An upload button is provided to upload an audio file to the input folder
utility function for compatibility with external nodes
VHS used to use an internal VHS_AUDIO format for routing audio between inputs and outputs. This format was intended to only be used internally and was designed with a focus on performance over ease of use. Since ComfyUI now has an internal AUDIO format, VHS now uses this format. However, some custom node packs were made that are external to both ComfyUI and VHS that use VHS_AUDIO. This node was added so that those external nodes can still function
[-]
Inputs:
[-]
vhs_audio: An input in the legacy VHS_AUDIO format produced by an external node
[-]
Outputs:
[-]
vhs_audio: An output in the standardized AUDIO format
Decode latents to images with a manually specified batch size
Some people have ran into VRAM issues when encoding or decoding large batches of images. As a workaround, this node lets you manually set a batch size when decoding latents.
Unless these issues have been encountered, it is simpler to use the native VAE Decode or to decode from a Video Combine directly
[-]
Inputs:
[-]
samples: The latents to be decoded.
[-]
vae: The VAE to use when decoding.
[-]
Outputs:
[-]
IMAGE: The decoded images.
[-]
Widgets:
[-]
per_batch: The maximum number of images to decode in each batch.
Encode images as latents with a manually specified batch size.
Some people have ran into VRAM issues when encoding or decoding large batches of images. As a workaround, this node lets you manually set a batch size when encoding images.
Unless these issues have been encountered, it is simpler to use the native VAE Encode or to encode directly from a Load Video
[-]
Inputs:
[-]
pixels: The images to be encoded.
[-]
vae: The VAE to use when encoding.
[-]
Outputs:
[-]
LATENT: The encoded latents.
[-]
Widgets:
[-]
per_batch: The maximum number of images to encode in each batch.
Append copies of a image to itself so it repeats
[-]
Inputs:
[-]
IMAGES: The image to be repeated
[-]
Outputs:
[-]
IMAGE: The image with repeats
[-]
count: The number of image in the output. Equal to the length of the input image * multiply_by
[-]
Widgets:
[-]
multiply_by: Controls the number of times the mask should repeat. 1, the default, means no change.
Return the number of images in an input as an INT
[-]
Inputs:
[-]
images: The input image
[-]
Outputs:
[-]
count: The number of images in the input
Combine two groups of images into a single group of images
[-]
Inputs:
[-]
images_A: The first group of images
[-]
images_B: The first group of images
[-]
Outputs:
[-]
IMAGE: The combined group of images
[-]
count: The length of the combined group
[-]
Widgets:
[-]
merge_strategy: Determines what the output resolution will be if input resolutions don't match
[-]
match A: Always use the resolution for A
[-]
match B: Always use the resolution for B
[-]
match smaller: Pick the smaller resolution by area
[-]
match larger: Pick the larger resolution by area
[-]
scale_method: Determines what method to use if scaling is required
[-]
crop: When sizes don't match, should the resized image have it's aspect ratio changed, or be cropped to maintain aspect ratio
Keep only 1 image for every interval
[-]
Inputs:
[-]
images: The input image
[-]
Outputs:
[-]
IMAGE: The output images
[-]
count: The number of images in the input
[-]
Widgets:
[-]
select_every_nth: The interval from which one frame is kept. 1 means no frames are skipped.
[-]
skip_first_images: A number of frames which that is skipped from the start. This applies before select_every_nth. As a result, multiple copies of the node can each have a different skip_first_frames to divide the image into groups
Split a set of images into two groups
[-]
Inputs:
[-]
images: The images to be split.
[-]
Outputs:
[-]
IMAGE_A: The first group of images
[-]
A_count: The number of images in group A. This will be equal to split_index unless the images input has length less than split_index
[-]
IMAGE_B: The second group of images
[-]
B_count: The number of images in group B
[-]
Widgets:
[-]
split_index: The index of the first latent that will be in the second output groups.
Append copies of a latent to itself so it repeats
[-]
Inputs:
[-]
latents: The latents to be repeated
[-]
Outputs:
[-]
LATENT: The latent with repeats
[-]
count: The number of latents in the output. Equal to the length of the input latent * multiply_by
[-]
Widgets:
[-]
multiply_by: Controls the number of times the latent should repeat. 1, the default, means no change.
Return the number of latents in an input as an INT
[-]
Inputs:
[-]
latents: The input latent
[-]
Outputs:
[-]
count: The number of latents in the input
Combine two groups of latents into a single group of latents
[-]
Inputs:
[-]
latents_A: The first group of latents
[-]
latents_B: The first group of latents
[-]
Outputs:
[-]
LATENT: The combined group of latents
[-]
count: The length of the combined group
[-]
Widgets:
[-]
merge_strategy: Determines what the output resolution will be if input resolutions don't match
[-]
match A: Always use the resolution for A
[-]
match B: Always use the resolution for B
[-]
match smaller: Pick the smaller resolution by area
[-]
match larger: Pick the larger resolution by area
[-]
scale_method: Determines what method to use if scaling is required
[-]
crop: When sizes don't match, should the resized image have it's aspect ratio changed, or be cropped to maintain aspect ratio
Keep only 1 latent for every interval
[-]
Inputs:
[-]
latents: The input latent
[-]
Outputs:
[-]
LATENT: The output latents
[-]
count: The number of latents in the input
[-]
Widgets:
[-]
select_every_nth: The interval from which one frame is kept. 1 means no frames are skipped.
[-]
skip_first_latents: A number of frames which that is skipped from the start. This applies before select_every_nth. As a result, multiple copies of the node can each have a different skip_first_frames to divide the latent into groups
Split a set of latents into two groups
[-]
Inputs:
[-]
latents: The latents to be split.
[-]
Outputs:
[-]
LATENT_A: The first group of latents
[-]
A_count: The number of latents in group A. This will be equal to split_index unless the latents input has length less than split_index
[-]
LATENT_B: The second group of latents
[-]
B_count: The number of latents in group B
[-]
Widgets:
[-]
split_index: The index of the first latent that will be in the second output groups.
Append copies of a mask to itself so it repeats
[-]
Inputs:
[-]
masks: The masks to be repeated
[-]
Outputs:
[-]
LATENT: The mask with repeats
[-]
count: The number of mask in the output. Equal to the length of the input mask * multiply_by
[-]
Widgets:
[-]
multiply_by: Controls the number of times the mask should repeat. 1, the default, means no change.
Return the number of masks in an input as an INT
[-]
Outputs:
[-]
count: The number of masks in the input
Combine two groups of masks into a single group of masks
[-]
Inputs:
[-]
mask_A: The first group of masks
[-]
mask_B: The first group of masks
[-]
Outputs:
[-]
MASK: The combined group of masks
[-]
count: The length of the combined group
[-]
Widgets:
[-]
merge_strategy: Determines what the output resolution will be if input resolutions don't match
[-]
match A: Always use the resolution for A
[-]
match B: Always use the resolution for B
[-]
match smaller: Pick the smaller resolution by area
[-]
match larger: Pick the larger resolution by area
[-]
scale_method: Determines what method to use if scaling is required
[-]
crop: When sizes don't match, should the resized image have it's aspect ratio changed, or be cropped to maintain aspect ratio
Keep only 1 mask for every interval
[-]
Outputs:
[-]
count: The number of mask in the input
[-]
Widgets:
[-]
select_every_nth: The interval from which one frame is kept. 1 means no frames are skipped.
[-]
skip_first_mask: A number of frames which that is skipped from the start. This applies before select_every_nth. As a result, multiple copies of the node can each have a different skip_first_frames to divide the mask into groups
Split a set of masks into two groups
[-]
Inputs:
[-]
mask: The masks to be split.
[-]
Outputs:
[-]
MASK_A: The first group of masks
[-]
A_count: The number of masks in group A. This will be equal to split_index unless the mask input has length less than split_index
[-]
MASK_B: The second group of masks
[-]
B_count: The number of masks in group B
[-]
Widgets:
[-]
split_index: The index of the first latent that will be in the second output groups.