29 credits per second
No videos yet
Enter a prompt and click Generate Video to create your first AI-generated video.
MiniMax H3 is a multimodal video model on Nano Banana Lab that generates clips from text, from start/end frames, or from up to nine reference images — with optional video and audio references on top. It renders 4 to 15 second clips at 768P or 2K, priced per second from 29 credits.
How it works
Pick the mode that matches what you already have.
Describe the scene in up to 7,000 characters — the longest prompt window of any video model here.
Upload a start frame (and optionally an end frame) and MiniMax animates between them.
Attach up to 9 images plus optional video and audio references to steer identity, style, and motion.
Options & pricing
Rates per second of video.
| Option | Values | Cost |
|---|---|---|
| Resolution | 768P / 2K | 29 / 46 credits per second |
| Duration | 4–15 seconds | rate × seconds |
| References | up to 9 images + video + audio | included |
| Prompt length | up to 7,000 characters | included |
Example: a 4-second 768P clip costs 116 credits. Failed generations are refunded.
Its reference system: up to nine images plus video and audio references in one request, alongside classic text and frame modes. That makes it the strongest choice when a specific person, product, or style must persist in the clip.
768P and 2K — the highest video resolution on this site. Duration runs from 4 to 15 seconds.
Per second: 29 credits at 768P, 46 at 2K. The button always shows the total for your current duration and resolution.
21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 in text mode; frame modes follow your uploaded start frame.
Switch to the References tab above, drop in your images, and see how tightly H3 can follow them.
Generate with MiniMax H3