MiniMax-H3 Video Generator
One endpoint for text, first/last-frame, and image/video/audio reference generation, with built-in music and sound effects. 768P, 4–15 s, billed per second.
MiniMax-H3 Video Generator
adaptive (follows the input's ratio) only works when the request includes an image or video; text-only and audio-only requests need a fixed ratio.
Gacha mode: submit several tasks with the same settings (each is billed)
- * Example durations. Duration is an integer from 4 to 15 s; cost = $0.03 × seconds.
- Reference images, videos and audio are not billed extra.
- Failed tasks are refunded automatically; resubmitting is billed again.
Billed per second. Failed tasks are refunded in full; submit errors are not charged.
No videos yet
What is MiniMax-H3
MiniMax-H3 (Hailuo 3.0) is MiniMax's video generation model. The generation mode is inferred from the media in content[] — no endpoint switching.
Text to video
Prompt only — 4–15 s videos with built-in music and sound effects.
First/last frame
First frame only, last frame only, or both — control where the shot starts and ends.
Reference to video
Mix up to 9 images, 3 videos and 3 audio clips, referenced in the prompt as <Picture 1> etc.
Why MiniMax-H3
Low price, per-second billing, refunds on failure.
Parameters & capabilities
MiniMax-H3 request parameters at a glance.
768P
This channel outputs 768P only.
4–15 seconds
Integer seconds, billed per second.
7 aspect ratios
21:9, 16:9, 4:3, 1:1, 3:4, 9:16, plus adaptive (follows the input).
First/last frame
Use a first frame, a last frame, or both.
Multimodal references
Combine image, video and audio references freely.
Default token
Uses your default-group, pay-as-you-go APIYI token — no dedicated token needed.
FAQ
Questions? Contact hi@apiyi.com.