Models
MiniMax H3 Max video model
TL;DR
Generate a 5-second 768P video in under 3 seconds with MiniMax H3 Max, using text or frame guidance for clips up to 15 seconds.
One model, two input modes
MiniMax H3 Max is built for faster-than-real-time generation: a 5-second 768P video renders in under 3 seconds. Fal reports roughly 15× higher average speed than models of comparable quality in its evaluation.
MiniMax H3 Max generates 768P video clips from 5 to 15 seconds. When no start or end frame is supplied, KrafLayer treats the request as Text-to-Video. When either frame is present, the same provider endpoint runs as Image-to-Video.
A start frame controls the opening image. An optional end frame can guide the final image. With image guidance, the output framing follows the input image rather than exposing a separate aspect-ratio selector.
Output storage and billing
MiniMax H3 Max costs 5 credits per generated second during the current limited-time 50% launch offer. A 5-second clip costs 25 credits and a 15-second clip costs 75 credits.
KrafLayer requests an asynchronous provider result. If the provider returns Base64 instead of a CDN URL, the server uploads the video to KrafLayer storage first. Supabase stores only the final video URL, never the raw Base64 payload.
Current KrafLayer defaults
- Resolution is fixed at 768P.
- Prompt expansion uses balanced mode.
- Safety checking is enabled.
- Audio is generated natively and remains always on. Because the provider does not expose an audio switch, the Audio control is shown as active and locked.
Related KrafLayer tools
- AI product image tools — Browse the full tool list for ecommerce image editing and product visual workflows.