MuoGen · AI video model
MiniMax H3 on MuoGen
MiniMax H3 offers 768P and 2K video settings with image, video and audio references in the MuoGen catalog.
Plan a 4–15 second scene with multiple reference types
Review MiniMax H3 when your scene fits within 4–15 seconds and you need several reference types. Its output duration range differs from its allowed reference duration range; do not use the reference limit as the output duration.
Which settings are available?
| Model identifier | minimax-h3 |
|---|---|
| Workflow | Video creation |
| Resolutions | 2K, 768P |
| Initial resolution | 768P |
| Catalog video duration | 4–15 seconds |
| Catalog aspect ratios | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptive |
| Accepted references | image: 30 maximum video: 10 maximum audio: 10 maximum |
| Maximum size per reference | image: 30 MB video: 200 MB audio: 15 MB |
| Maximum prompt length | 7000 characters |
| Audio on/off control | No separate control |
| Duration per video or audio reference | 2–30 seconds |
| Total reference duration | 30 seconds of video; separately, the same limit for audio |
These capabilities describe the settings exposed by MuoGen, not a guarantee of visual quality. Check the model and its settings in the workspace before preparing a request. Catalog checked on .
How do I prepare the prompt and references?
Describe the subject and action before atmosphere or effects. If a video reference illustrates camera motion, say so explicitly. Audio references should have a clear role, such as providing an intended sound direction.
Example brief to adapt
A cyclist passes through a quiet city street at sunrise. Side tracking shot, steady movement and warm light. Use the video reference for camera motion and the image reference for the street setting.
This example is a preparation brief, not a generated result or a promise of output. Use only references you have the right to upload.
What limitations should I plan for?
The catalog does not expose a separate audio on/off control for this model. Compare the actual settings in the workspace before planning a silent deliverable, and check the result’s sound after generation becomes available.
Can I generate with this model today?
The catalog and settings are available to explore. Paid generation in the public browser workspaces is not yet open. When enabled, it will require a verified account and paid credits. Preparing a prompt or reading this guide does not start a generation.
Explore settings for MiniMax H3
Compare and prepare your project
AI video preparation guide · MuoGen model catalog · Availability and frequently asked questions
- Seedance 2.5: Plan a video with image, video and audio references
- Seedance 2.5 Edit: Describe changes to an existing video
- Seedance 2.0: Prepare a short video with resolution and reference choices
Updated: