Repolode
Submit
pixelaigc

pixelaigc/ComfyUI-MiniMax-H3-Workflow

A ComfyUI workflow for MiniMax H3 that generates videos from first and last frames, a first frame alone, or text prompts, with a Turbo LoRA and a resolution control panel.

View on GitHub
Stars
404
Forks
62
Open issues
0
Last commit
2 months ago

What is it?

This is a ComfyUI workflow JSON for generating video with the MiniMax H3 model. It accepts a first frame, an optional last frame, and a text prompt describing motion, camera and sound. Model loading, Turbo LoRA, sampling, resolution calculation and audio/video decoding are packed into one subgraph, while steps, target duration, seed, aspect ratio and megapixels are exposed on the outer control panel.

Setting only the first frame gives image-to-video, and bypassing both image inputs gives text-to-video. The output is 24 fps video with audio decoded separately and merged before saving through SaveVideo. Resolutions are rounded to multiples of 32, and frame counts are aligned to the model's required 17k+5 pattern.

The workflow depends on the ComfyUI-MiniMax-H3-Turbo custom node package for its LoRA loader, and on the MiniMax H3 model files from Hugging Face, which must be downloaded separately. Image files are not included in the JSON and must be uploaded again.

Key features

  • 01Supports first-and-last-frame, image-to-video and text-to-video modes.
  • 02Exposes model, LoRA, steps, duration, seed and resolution settings on one panel.
  • 03Calculates output resolution from aspect ratio and megapixels, aligned to multiples of 32.
  • 04Outputs 24 fps video with decoded audio merged into the file.
  • 05Includes an audio VAE and video VAE alongside the base model.

Use cases

  • Turning a start image and an end image into a short video with a described transition.
  • Animating a single still image from a text prompt.
  • Generating video from text alone without any input images.
  • Testing MiniMax H3 video generation inside ComfyUI with adjustable resolution and duration.

Quick start

  1. 1

    Install the required custom node pack via git clone into custom_nodes, then restart ComfyUI

    git clone https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo custom_nodes/ComfyUI-MiniMax-H3-Turbo
  2. 2

    Download the workflow JSON from the repository and drag it into the ComfyUI canvas

  3. 3

    Download the model files listed in the README into the matching ComfyUI models folders

  4. 4

    Upload first and last frame images in the two image loader nodes, enter a prompt, then run

Things to know

  • Requires downloading several large model files from Hugging Face before it can run.
  • The README provides no GPU, VRAM, or speed requirements, so hardware needs are unknown.
  • Needs a separate custom node pack installed; the workflow fails with missing nodes otherwise.
  • The repository has no declared license, so reuse terms are unclear.

In collections

AI video generationAI video & audio tools

Topics

#comfyui#video-generation#image-to-video#text-to-video#workflow#lora#minimax

FAQ

What is ComfyUI-MiniMax-H3-Workflow?
A ComfyUI workflow for MiniMax H3 that generates videos from first and last frames, a first frame alone, or text prompts, with a Turbo LoRA and a resolution control panel. This is a ComfyUI workflow JSON for generating video with the MiniMax H3 model. It accepts a first frame, an optional last frame, and a text prompt describing motion, camera and sound. Model loading, Turbo LoRA, sampling, resolution calculation and audio/video decoding are packed into one subgraph, while steps, target duration, seed, aspect ratio and megapixels are exposed on the outer control panel.
Is ComfyUI-MiniMax-H3-Workflow free to use? What license does it have?
ComfyUI-MiniMax-H3-Workflow has no declared license. Its code is public, but without a license you have no explicit permission to reuse it; ask the author before relying on it.
Is ComfyUI-MiniMax-H3-Workflow still maintained?
Updated within 6 months. The last commit to ComfyUI-MiniMax-H3-Workflow was on Sep 10, 2026.
What can I use ComfyUI-MiniMax-H3-Workflow for?
Turning a start image and an end image into a short video with a described transition. Animating a single still image from a text prompt. Generating video from text alone without any input images. Testing MiniMax H3 video generation inside ComfyUI with adjustable resolution and duration.
How do I install ComfyUI-MiniMax-H3-Workflow?
Install the required custom node pack via git clone into custom_nodes, then restart ComfyUI: git clone https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo custom_nodes/ComfyUI-MiniMax-H3-Turbo Download the workflow JSON from the repository and drag it into the ComfyUI canvas Download the model files listed in the README into the matching ComfyUI models folders Upload first and last frame images in the two image loader nodes, enter a prompt, then run
Where can I get ComfyUI-MiniMax-H3-Workflow?
The source code and installation instructions are on GitHub: https://github.com/pixelaigc/ComfyUI-MiniMax-H3-Workflow

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects