MiniMax H3 Max
MiniMax H3 Max for high-quality asynchronous video generation at 480P or 768P.
Execution
Async video
Submit a task, then poll or use a callback
Duration
5-15s
Integer output duration
Resolution
480P / 768P
Mode-dependent fixed or adaptive aspect ratios
minimax-h3-maxPlayground
minimax-h3-maxClick or drag
Click or drag
Resolution (sent to backend)
Aspect Ratio (sent to backend)
No output yet
Pricing
| Plan | Context Window | Max Output | Input | Output | Cache Write | Cache Read |
|---|---|---|---|---|---|---|
MiniMax-H3-Max | — | - | $0.0553 per second | - | - | - |
480P · $0.05 · per second
768P · $0.08 · per second
Input Video 768P · $0.143 · per second
Input Image · $0.074 · each (past free)
README
- MiniMax H3 Max supports text-to-video, first-frame image-to-video, first-and-last-frame generation, and multimodal reference workflows.
- Multimodal reference mode accepts up to 9 reference images and 3 reference videos. Frame inputs and reference media cannot be mixed in one request.
- Text-only requests require a fixed aspect ratio. Frame modes always use the input image ratio, while reference mode supports an effective fixed ratio or adaptive source ratio.
API
MiniMax H3 Max for high-quality asynchronous video generation at 480P or 768P.
Model Code
minimax-h3-max
API
Playground coming soon
Related Models
MiniMax H3
MiniMax Hailuo-03 for high-quality asynchronous video generation with frame and multimodal references.
Seedance 2.5
ByteDance's Seedance 2.5 model for long-form multimodal video generation, editing, extension, and audio-driven creation.
Seedance 2
ByteDance's Seedance 2 video model for multimodal reference input, editing, extension, and asynchronous generation.
Seedance 2 Fast
ByteDance's faster Seedance 2 variant for multimodal asynchronous video generation at lower per-second rates.
Seedance 1.5 Pro
ByteDance's Seedance 1.5 Pro video model for higher-capability text and frame-guided video generation.
Seedance 1.0 Pro
ByteDance's Seedance 1.0 Pro video generation model for text-guided and frame-guided cinematic video tasks.
