Video model · MiniMax · served by Fal
Home / Models / MiniMax H3 Max
MiniMax H3 Max
5 seconds of video. Under 3 seconds.
A globally leading ultra-fast video model: MiniMax H3 Max turns a 5-second 768P idea into finished video in under 3 seconds, with native synchronized audio.
Model configuration
- Generation speed
- 5s video in under 3s
- Resolution
- 768P
- Duration
- 5–15 seconds
- Generation cost
- 5 credits per second
Input modes
- Text-to-Video
- Image-to-Video
- First-to-last-frame guidance
Features
- Faster-than-real-time generation: a 5-second 768P clip renders in under 3 seconds
- Around 15× faster on average than models of comparable quality in Fal's evaluation
- Prompt expansion tuned for balanced speed and detail
- Start-frame and optional end-frame guidance
- Input-image aspect ratio carried into guided video output
- Safety checking enabled for every generation
- Limited-time 50% launch pricing from the provider
How KrafLayer handles this model
With no start or end frame, KrafLayer records the task as Text-to-Video.
With either frame present, KrafLayer records the task as Image-to-Video.
Provider output is copied to KrafLayer storage before the generation record is completed; raw Base64 video data is never stored in Supabase.
Audio is generated natively and is always on. KrafLayer shows the Audio control as active and locked because the provider exposes no off switch.