MiniMax-H3 video generation API
Hailuo H3 multimodal video: 4 to 15 seconds, 768P or 2K, six aspect ratios, billed per second through an OpenAI Videos compatible endpoint.
The MiniMax-H3 video model
MiniMax-H3 (Hailuo H3) is MiniMax's 2026 multimodal video generation model. It handles text-to-video, image-to-video and generation guided by reference images, reference videos and reference audio. Clips run from 4 to 15 seconds at 768P or 2K, in 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16.
MiniMax-H3 API pricing (USD)
| Item | Rule | Price |
|---|---|---|
| Generated video, 768P | Per output second | $0.0746 / second |
| Generated video, 2K | Per output second | $0.1194 / second |
| Reference images | First 5 free | $0.0299 each beyond that |
| Reference video | Per input second at the output resolution's rate | 768P $0.0746 / second, 2K $0.1194 / second |
| Reference audio | Free | $0 |
A 5-second 768P clip costs about $0.37 and a 2K clip about $0.60. The estimate is reserved when the task is submitted, settled on the actual output length, and failed tasks are not charged.
Quick start (OpenAI Videos compatible endpoint)
curl https://yiduochan.com/v1/videos \
-H "Authorization: Bearer $YIDUOCHAN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "MiniMax-H3",
"prompt": "An orange cat stretching on a sunlit windowsill, photorealistic, slow push-in",
"seconds": 5,
"size": "768P"
}'
# poll the task; the video URL is returned once it completes
curl https://yiduochan.com/v1/videos/{video_id} \
-H "Authorization: Bearer $YIDUOCHAN_API_KEY"Request parameters
| Parameter | Description |
|---|---|
model | Always MiniMax-H3 |
prompt | Description of the video, Chinese or English |
seconds | Duration, an integer from 4 to 15, default 5 |
size | 768P or 2K, default 768P |
input_reference | First-frame image URL for image-to-video |
metadata.ratio | Aspect ratio: adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 |
metadata.reference_video / metadata.reference_audio | Reference video and audio URLs for multimodal guidance |
Typical uses
- Product showcase clips and short-form social video at scale.
- Image-to-video from a product or character shot that keeps the subject consistent.
- New footage that follows the camera motion and pacing of a reference video.
FAQ
How is the MiniMax-H3 video API billed?
Per output second: $0.0746 at 768P and $0.1194 at 2K. The first 5 reference images are free and $0.0299 each after that; a reference video is billed per input second at the output resolution's rate.
How long can a MiniMax-H3 video be?
Between 4 and 15 seconds per task, at 768P or 2K, with 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 or an adaptive ratio taken from the input media.
Is generation synchronous?
No. The API returns a task id immediately; poll the retrieval endpoint and the video URL appears when the task completes. Failed tasks are not charged.
Is image-to-video supported?
Yes. Pass a first-frame image URL, or supply several reference images plus reference video and audio for multimodal generation.