MiniMax-H3-Flex
MiniMax-H3-Flex is a faster MiniMax H3: same inputs, same price, and in our tests each video finishes in about 70% of the time hailuo_h3 takes. It follows motion and action in the prompt more closely (spins, jumps, a move between the first and last frame), while keeping the same text-to-video, first / last frame and audio-driven lip sync features.
Pass image_url for a first frame, optionally with image_end_url for a last frame. For reference-driven generation use the hailuo_h3_refs model instead.
Pricing — billed per second, by resolution:
| Resolution | Tokens / sec | USD / sec |
|---|---|---|
| 480p | 1.5 | $0.03 |
| 720p | 3 | $0.06 |
| 2K | 4.5 | $0.09 |
All resolutions support 5–15 seconds.
Audio-driven lip sync — pass audio_url (optionally with audio_start) and the characters sing or speak along with your audio. The model generates only the picture; the video’s soundtrack is your audio clip, and the video lasts exactly duration seconds as long as the audio covers it. It works with text-to-video, a first frame, or first + last frames. Billed the same as without audio.
Returns an img_uuid; poll GET /api/v1/jobs/detail (or use custom_callback_url) to get the generated video.
Headers
Body
Model name. Must be minimax-h3-flex.
minimax-h3-flex Text prompt describing the video. Required. Max 7000 characters.
7000Optional first-frame image. JPG/PNG/WEBP/AVIF, ≤10MB, shortest side ≥300px. When provided, runs image-to-video and the output ratio follows the image.
Optional last-frame image (same limits as image_url). Must be used together with image_url.
Output resolution. Also sets the price — see the pricing table above. 2K takes longer than 720p.
480p, 720p, 2K Video duration in seconds. Any integer from 5 to 15.
5 <= x <= 15Aspect ratio. Text-to-video only; ignored when image_url is provided (the ratio then follows the input image).
21:9, 16:9, 4:3, 1:1, 3:4, 9:16 Optional. Drives the video with this audio: the characters lip-sync to it, and the output's soundtrack is the audio clip from audio_start to audio_start + duration. Pass the whole song or speech track; it does not need to be pre-cut. MP3/WAV/M4A/AAC/FLAC/OGG, ≤50MB. Works with text-to-video, image-to-video and first/last frame. No extra charge.
Optional. Where to start reading audio_url, in seconds (decimals allowed). Must be used together with audio_url. The audio left after audio_start must cover duration: a shortfall of less than one second is accepted (the video then ends with the audio); anything more is rejected before tokens are charged.
x >= 0Optional. A publicly accessible HTTPS URL. When the task status changes (processing / success / failed), GoEnhance sends a POST request to this URL. The request body is identical to the response of GET /api/v1/jobs/detail. If your server does not respond with HTTP 200, the notification is retried up to 3 times, with a 3-second timeout per attempt.
"https://your-server.com/goenhance/callback"