Do I need specific model files to run this locally?

Yes. Install the Qwen-Image ComfyUI extension and download its assets. Place qwen_image_vae.safetensors in models/vae and ensure the Qwen text encoder (Qwen 2.5 VL 7B) is installed per the extension’s instructions. Put the Gummy Animals LoRA in your LoRA/models directory so the generator node can load it.

How do I control how “gummy” the result looks?

Use the LoRA strength and prompt wording. Higher LoRA strength yields a stronger translucent, sugar-coated look. In the prompt, include descriptors like "translucent," "sour gummy," "sugar crystals," and "backlit." Reduce strength or remove candy-specific terms if the style overpowers the subject.

Can I use non-animal subjects?

The LoRA is tuned for animals and works best when the subject is an animal (e.g., dog, cat, horse). It may produce interesting results with other nouns, but consistency and anatomy are most reliable with animal prompts.

My images look muddy or not translucent. How can I fix that?

Try a brighter, high-key setup: add "on white background," "studio lighting," and "soft backlight" to the prompt. Increase LoRA strength slightly, raise steps for more detail, and keep guidance moderate to prevent oversaturation. Also try simpler compositions (single subject, centered) at 512–768px for crisper material cues.

Qwen Image: Gummy Animals LoRA

This workflow turns any animal prompt into a translucent, sugar-coated "sour gummy" character using the Qwen-Image model with a style LoRA. The core of the graph is a single Qwen-Image generation node (id: e5cfe5ba-2ae0-4bc4-869f-ab2228cb44d3) that handles text conditioning, sampling, and LoRA application, followed by a SaveImage node to write results to disk. Two annotations (Note and MarkdownNote) provide usage tips and a link to ComfyUI’s Subgraph guide so you can encapsulate and reuse the gummy-style block in your own projects.

Under the hood, the Qwen-Image node loads its VAE (qwen_image_vae.safetensors) and text encoder (Qwen 2.5 VL 7B), then injects the Gummy Animals LoRA into the model’s attention layers. This low‑rank adaptation steers the base model toward a consistent sour-candy aesthetic without retraining the full network. You control the look through the prompt (e.g., "a chubby corgi, translucent sour gummy, sugar crystals") and standard generation parameters like steps, guidance, resolution, seed, and the LoRA strength, allowing you to balance style intensity against subject fidelity. The SaveImage node captures final frames to your ComfyUI/output folder for quick iteration.

API

Use this workflow from code

Every Comfy workflow is a JSON graph. The payload below is this workflow, exactly as ComfyUI runs it — fetch it from the URL, keep it in version control, or load it in ComfyUI and run it node by node.

template_sugar_coated_gummy_style_qwen.json

Fetching workflow JSON…

Run it from TypeScript or Python with the Comfy SDK. The same code targets Comfy Cloud or a ComfyUI you host yourself — only the base URL changes.

// Install (beta)
npm i @comfyorg/sdk

// Run "Qwen Image: Gummy Animals LoRA" (TypeScript)
import { Comfy } from "@comfyorg/sdk";

const client = new Comfy({ apiKey: "comfyui-..." });

// This workflow, exported in API format (see note below)
const wf = await client.workflows.fromFile("template_sugar_coated_gummy_style_qwen_api.json");

const job = await client.run(wf);
await job.getOutputs("60")[0].toFile("output.png");  // SaveImage

The SDK takes a workflow in API format: open this workflow in ComfyUI and use File → Export Workflow (API).

Comfy Cloud API access requires a plan with an API key.

SDK docs Get an API key

FAQ

Frequently Asked Questions