MiniMax H3 is a video model from MiniMax that makes the video and its audio together: voices, music and sound effects in one pass, for clips up to 15 seconds. In Promptus it comes as three ready-made CosyTemplates that you install with Install Locally. Watch our MiniMax H3 install video.
| Developer | MiniMax |
| Output | Video with audio (voices, music, sound effects), up to 15 seconds |
| Workflows in Promptus | Text to Video, Image to Video, Reference to Video |
| Reference inputs | Up to 9 images, 3 videos and 3 audio clips (12 files max) |
| Video model files | 66.3 GB (bf16), 40.2 GB (pruned), 34.0 GB (int8), Comfy-Org repack |
| Text encoder | Qwen3-VL 32B: 51.5 GB (bf16) or 27.1 GB (int8) |
| ComfyUI | Native |
| Licence | MiniMax H3 Community License: no rights granted in the EU, UK, South Korea or US |
On an RTX 5090 Laptop GPU with 24 GB, MiniMax H3 took about 3.1 minutes per 5-second clip at 640×640 and 24 fps in our five-model image-to-video test, and it was the only one of the five to produce an audio track.
The MiniMax H3 Community License grants rights only outside the European Union, the United Kingdom, South Korea and the United States. Commercial products earning more than $20 million a year need MiniMax's written authorisation, and commercial products must show “MiniMax H3” in their interface. Read the licence for your own case.
Read our full guide: MiniMax H3 in ComfyUI: Install It Locally.
Join our distributed GPU compute network. Help us make AI accessible, scalable
and secure for designer, developers and start-ups.