What does the "zhuanchang" trigger word do, and where should I put it?

It activates the ltx2.3-transition behavior in the model+LoRA combo. Place it near the end of your prompt so your core subject, scene, camera, and transformation details remain dominant. This helps the base description stay clear while the transition logic is applied.

How do I control how fast the transition happens?

Use frame count and FPS. More frames at the same FPS produce a longer, slower transition; fewer frames produce a faster one. You can also phrase timing cues in your prompt (e.g., "first half… then…"), but duration is primarily governed by total frames and playback speed.

When should I use a reference image with LoadImage?

Use a reference if you want to anchor color palette, texture, or a specific subject look. Connect it to the pipeline’s reference input. If you aim for a purely text-driven transition, skip the reference and rely on descriptive prompts.

The transition feels too strong or inconsistent. How can I improve stability?

Lower the LoRA strength to 0.4–0.6 for subtler shifts, fix the seed for repeatability, and simplify the prompt to fewer, clearer visual goals. Avoid conflicting descriptors; keep subject and camera consistent while describing how the scene or style evolves.

LTX-2.3 Style Transition - ComfyUI Workflow

This ComfyUI workflow builds a guided style and scene transition on top of the LTX-2.3 text-to-video model by applying the ltx2.3-transition LoRA. The core generation happens inside a custom LTX-2.3 pipeline node (beb19732-5803-4af6-b2d2-289692ce780b), which takes your prompt, applies the LoRA at a user-defined weight, and renders a sequence of frames that smoothly evolve from one style or scene description to another. If you provide a reference image with LoadImage, the pipeline can also pull color palettes or subject cues into the transition; if you skip it, the transition is guided purely by text.

Practical controls are surfaced where you need them: prompt fields (with support for the trigger word "zhuanchang"), frame count and FPS for timing, and LoRA strength to dial how assertively the transition behaves. MarkdownNote is embedded to keep prompting tips visible as you iterate, and SaveVideo encodes the final frames into a video file at your chosen frame rate and quality. The result is a reliable, loop-friendly way to blend visual elements—style, lighting, subject emphasis, or even scene composition—without hand-animating or post-processing crossfades.

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_ltx2_3_style_transition.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 "LTX-2.3 Style Transition" (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_ltx2_3_style_transition_api.json");

const asset = client.assets.fromFile("input.png");
wf.setInput("137", "image", asset);  // LoadImage

const job = await client.run(wf);
await job.getOutputs("68")[0].toFile("output.png");  // SaveVideo

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