ComfyUI custom nodes for the MiniMax H3 API, delivered through Muapi.
The pack follows the same focused workflow pattern as seedance2.5-comfyui: add one API Key node, connect it to a generation node, then connect video_url to Save Video and first_frame to ComfyUI's Preview Image node.
▶ Watch: How to Use MiniMax H3 (Hailuo) Uncensored in ComfyUI — Native 2K Video with Sound
| Node | What it does |
|---|---|
| MiniMax H3 API Key | Stores one Muapi API key for the workflow. |
| MiniMax H3 Text-to-Video | Generates a 2K clip from a text prompt. |
| MiniMax H3 Image-to-Video | Animates one ComfyUI image or remote image URL. |
| MiniMax H3 Reference-to-Video | Uses image, video, and optional audio references to guide a 2K clip. |
| MiniMax H3 Save Video | Downloads a generated URL to ComfyUI/output and returns decoded frames. |
Copy this directory into ComfyUI/custom_nodes/minimax-h3-comfyui, then install the Python dependencies in the same environment ComfyUI uses:
pip install -r ComfyUI/custom_nodes/minimax-h3-comfyui/requirements.txtRestart ComfyUI. The example workflows can be loaded with File → Load.
Included examples are:
MiniMaxH3_T2V_Example.jsonMiniMaxH3_I2V_Example.jsonMiniMaxH3_ReferenceToVideo_Example.json
Use the 🔑 MiniMax H3 API Key node and connect its output to a generation node. Alternatively, leave the node field empty and configure one of these:
export MUAPI_API_KEY="your_muapi_key"or configure the Muapi CLI:
muapi auth configure --api-key YOUR_KEYThe nodes also read ~/.muapi/config.json when it contains an api_key field. Keys are only used server-side for the request and are not written to generated workflow files unless you explicitly save them in a node field.
- Add 🔑 MiniMax H3 API Key.
- Add 🎬 MiniMax H3 Text-to-Video.
- Choose an aspect ratio,
2kresolution, and a duration from 5–15 seconds. - Connect
video_urlto 🎬 MiniMax H3 Save Video. - Connect
first_frameto Preview Image.
Connect Load Image → image on 🎬 MiniMax H3 Image-to-Video. The node uploads the ComfyUI IMAGE tensor and sends the resulting URL as image_url. You can use the image_url field instead when the source is already hosted or is a local file path.
The reference node provides five image slots, three video slots, and two audio slots. Image slots accept either a connected IMAGE or a URL/path field. Video and audio slots accept a URL/path field or a file selected from ComfyUI/input; a URL/path overrides the dropdown. At least one image or video reference is required. Audio references are optional and cannot be used by themselves.
Describe how each asset should influence the shot in the prompt, for example:
Use the product image as the subject, borrow the movement from the reference clip,
and use the audio reference for the atmosphere. Create a slow cinematic reveal.
The nodes call the current Muapi endpoints documented by the MiniMax H3 SDK:
| Workflow | Endpoint |
|---|---|
| Text-to-video | POST /api/v1/minimax-h3-text-to-video |
| Image-to-video | POST /api/v1/minimax-h3-image-to-video |
| Reference-to-video | POST /api/v1/minimax-h3-reference-to-video |
| Poll task | GET /api/v1/predictions/{request_id}/result |
| Upload local media | POST /api/v1/upload_file |
Text-to-video and reference-to-video expose the supported aspect ratios 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. All generation nodes expose the current 2k resolution and integer durations from 5 through 15 seconds.
All requests use the x-api-key header. Generation is asynchronous: the nodes submit a task, poll until it completes, and return the final media URL plus a decoded first frame.
Python 3.8+, requests, Pillow, numpy, and opencv-python. ComfyUI supplies the installed PyTorch runtime used for IMAGE tensors. For in-ComfyUI video playback, install ComfyUI-VideoHelperSuite.
- MiniMax-H3-API — Python SDK for the same Muapi API.
- seedance2.5-comfyui — related focused ComfyUI nodes.
- MiniMax H3 API guide — model capabilities and API overview.
MIT