MiniMax H3 Prompt Writer for ComfyUI

MiniMax H3 Prompt Writer is a ComfyUI extension for writing prompts specifically for MiniMax H3.

Describe what you want in normal language, add image / video / audio references, and let a local or API model turn it into an editable H3 prompt.

This is a UI extension, not a workflow node.

It does not run MiniMax H3, modify your graph or queue a video. It writes the prompt, then you copy it into your existing H3 workflow.

GitHub:
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer

Features

Supports all current H3 input modes:

• T2VA

• I2VA

• FL2VA

• L2VA

• Reference

Reference mode supports up to 9 images, 3 videos and 3 audio references, with 12 files total.

You can tell Writer exactly what each reference should provide:

Use Picture 1 for the character, Picture 2 only for the clothes, and only the movement from Video 1. Put the character on a rainy street at night.

Writer handles the H3-specific prompt structure and reference syntax around that.

Other features:

• Editable generated prompts

• Refine an existing prompt with a short instruction

• Separate saved drafts for every H3 mode

• Insert Picture / Video / Audio tags directly into the text

• Video contact sheets with frame sampling controls

• Automatic context handling

• Local model unload and VRAM controls

• Custom system prompts

• Official MiniMax prompt-writing guides used during generation

• MiniMax Music 3 workspace

Prompt Model Providers

The model writing the prompt is separate from the MiniMax H3 model used in your video workflow.

There are currently four ways to run it.

Ollama

Probably the easiest local option for most people.

Gemma 4 is the recommended local model family. Qwen 3.6 was also tested through Ollama in all five H3 modes without requiring Qwen-specific changes.

Direct GGUF

Loads the prompt model directly inside ComfyUI through llama-cpp-python.

Direct GGUF is currently built and validated around Gemma 4 with its matching vision projector.

External llama.cpp

Connect Writer to your own llama-server.

Useful if you already use llama.cpp or want full control over model loading, GPU placement, context, KV cache and the runtime itself.

API Providers

Supports:

• Gemini

• OpenAI

• OpenRouter

• Custom OpenAI-compatible endpoints

Custom endpoints can also be used with local servers such as LM Studio.

Local providers keep the prepared media and prompt request on your machine.

Remote API providers receive the data required for the prompt request.

Provider guide:
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer/blob/main/docs/PROVIDERS.md

MiniMax Music 3

Prompt Writer also includes a separate workspace for MiniMax Music 3.

Describe the track you want in normal language and Writer builds the structured Music 3 caption.

Lyrics can be written manually, generated from the Music Brief, or rewritten using Refine.

Music 3 is separate from H3 video prompting, but uses the same provider system.

Basic Usage

1. Open H3 Prompt Writer

2. Go to Settings and choose your provider / model

3. Select the H3 mode

4. Add your references

5. Write the Creative Brief normally

6. Press Generate prompt

7. Edit it, use Refine, or copy it into your H3 workflow
   
   
   

You do not need to manually write H3 timestamps, sections or reference formatting.

Installation

Download the ZIP from this Civitai page.

Extract:

ComfyUI-MiniMaxH3-Prompt-Writer

into:

ComfyUI/custom_nodes/

The final path should look like:

ComfyUI/custom_nodes/ComfyUI-MiniMaxH3-Prompt-Writer/

Then restart ComfyUI.

You can also install the latest version directly from GitHub:

cd ComfyUI/custom_nodes
git clone https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer.git

Open it using the floating H3 Prompt Writer button


or:

Extensions > H3 Prompt Writer

Important: This is a UI extension. No H3 Prompt Writer node will appear in node search.

Full installation guide:
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer/blob/main/docs/INSTALLATION.md

Direct GGUF Setup

If you want to use Direct GGUF on Windows Portable with NVIDIA, open PowerShell or CMD in the ComfyUI Portable folder and run:

.\python_embeded\python.exe -m pip install --only-binary=:all: --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu130 "llama-cpp-python>=0.3.34,<0.4"

Then restart ComfyUI.

H3 Prompt Writer > Settings > Direct GGUF

Use Browse verified models to choose a supported GGUF + matching mmproj.

Place both files under:

ComfyUI/models/LLM/

For most users, Ollama is still the simpler recommended local setup.

Full Direct GGUF guide:
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer/blob/main/docs/DIRECT_GGUF.md

Notes

• Video references are shown to the prompt model as ordered contact sheets instead of the full encoded video.

• Prompt models do not listen to uploaded audio files. Describe what should be taken from the audio, such as voice, music, rhythm or soundtrack.

• The base extension has no additional Python requirements.

• Remote API providers receive the data needed for the request. Local providers keep it local.

• The project itself is released under the MIT License.

• MiniMax-provided guides and model files retain their upstream terms.

Links

GitHub
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer

Usage / Creative Brief examples
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer/blob/main/docs/USAGE.md

Provider guide
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer/blob/main/docs/PROVIDERS.md

Troubleshooting
https://github.com/duckyshell/ComfyUI-MiniMaxH3-Prompt-Writer/blob/main/docs/TROUBLESHOOTING.md