Getting Started

Getting started

Requirements

• A current ComfyUI build containing native Add Guide for MiniMax H3 from ComfyUI PR #15439.
• The H3 model, text encoder, video VAE, audio VAE, and any LoRA selected by your chosen workflow. Model files are not bundled.
• ffmpeg on PATH is preferred. Review and assembly can fall back to ComfyUI's bundled PyAV.

Install version 0.5

From ComfyUI/custom_nodes:

git clone --branch feature/0.5-workflow-ux \
  https://github.com/ethanfel/ComfyUI-MiniMaxH3-Contex-Loop.git

Restart ComfyUI and reload the browser.

Your first run

1. Download T2V Normal or choose another example in the workflow chooser.
2. Drag the JSON onto the ComfyUI canvas.
3. Resolve any red model selections. The example model files are not bundled.
4. Give MiniMax H3 Contex Loop Plan a unique run_name.
5. Edit the scene prompts in Plan or Scene Prompt Editor.
6. Keep Guide and Generated audio for the simplest first test.
7. Queue the graph. Preflight should complete before H3 model execution.
8. At Review Gate, approve the scene, retry it with an edited prompt, reroll its seed, or approve and stop.
9. Assemble the completed or partial manifest.

Existing MP4 files are not overwritten. A collision produces _001, _002, and so on.

Good first settings

Setting	Start with
Canvas	960 × 544 or the example default
Transition	Guide
Audio	Generated final audio; generated continuity on
Encode / anchor	video / head
Timing	Keep the example's frame counts

H3 dimensions must be multiples of 32. Scene lengths are normalized upward to the 17k+5 frame grid.

Example assets

Some workflows need bundled images, masks, or footage copied into ComfyUI/input/. Example assets lists the exact files, sources, and licenses.

Next: Choose a workflow, write a Plan, or understand continuity.