MiniMax H3 Video Generation
Generate 4–15 second videos from text, up to nine images, or up to three audio references through one REST API, with unified async tasks, webhooks, billing, and CDN delivery.
One endpoint, three creative starting points
No action or separate paths: the API infers the mode from the supplied media.
Text to video
Generate clips directly from scene, motion, camera, and style descriptions.
Image to video
Use 1–9 images to control subjects, composition, characters, and visual style.
Audio guidance
Guide visual rhythm with up to three audio references, optionally combined with images and a prompt.
Async delivery
Return a task ID immediately and receive the final CDN video through polling or callback_url.
curl -X POST 'https://api.acedata.cloud/minimax/videos' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-H 'Content-Type: application/json' \
-d '{
"model": "minimax-h3",
"prompt": "A red fox running through a snowy forest at dawn, cinematic tracking shot",
"ratio": "16:9",
"duration": 6,
"async": true
}'Generate in a few lines
Parameters express the creative intent while the platform handles long-running execution, storage, and exact billing.
Describe or add media
Provide at least one of prompt, image_urls, or audio_urls.
Submit asynchronously
Set async to receive task_id immediately without holding a connection.
Retrieve the result
Use `/minimax/tasks` or a webhook to receive the AceDataCloud CDN video.
Common paths from media to video
One model covers creative exploration, still-image animation, and rhythm-driven clips.
Marketing clips
Turn product messages and shot lists into landscape ads or vertical social clips.
Character consistency
Use multiple images to constrain people, clothing, props, and scene style.
Rhythm visuals
Drive dance, camera motion, and visual changes with the rhythm of reference audio.
Launch in three steps
Use AceDataCloud authentication and tasks without building a separate media pipeline.
Get a token
Create an application and obtain a Bearer token.
Submit generation
Choose duration and ratio, then add text, images, or audio.
Deliver the video
Poll the task or receive a webhook and connect the CDN video to your product.
A production-ready unified video API
A stable contract for execution, errors, usage records, and file delivery.
| Capability | AceDataCloud MiniMax H3 | Some Others |
|---|---|---|
| Multimodal input | Text, 9 images, and 3 audio references | Multiple APIs to assemble |
| Long-running tasks | Polling and webhooks | Requires a custom task queue |
| Result files | Stored on the platform CDN | Inconsistent link lifetime |
| Failed-task billing | No charge on failure | Manual billing reconciliation |
Choose a workflow by input
The model is minimax-h3; the mode is inferred from input.
Prompt only
Generate directly from ideas, scripts, and shot descriptions.
Pass image_urls
Animate visual assets while preserving subject and style.
Pass audio_urls
Create dance, music visuals, and rhythm-driven clips.
async / callback
Ideal for batch and background jobs without a long HTTP connection.
Transparent per-second pricing
Every mode costs $0.25 per second, charged only on success.
4 seconds
- Fast creative validation
- Landscape or portrait
- Three input modes
8 seconds
- Complete short shot
- Multi-image and audio guidance
- Async task retrieval
15 seconds
- Maximum single output
- No charge on failure
- CDN result delivery
Frequently asked questions
Key details about modes, media, duration, and tasks.
Do I need to pass action?
No. audio_urls selects audio guidance; otherwise image_urls selects image-to-video; prompt-only requests use text-to-video.
Which ratios and durations are supported?
Ratios are 16:9 and 9:16. Duration must be an integer from 4 to 15 and defaults to 4 seconds.
How do I avoid long-connection timeouts?
Set async: true for an immediate task_id, then query `/minimax/tasks`, or provide callback_url.
Are failed tasks charged?
No. Only tasks that complete with a video result are charged by final duration.
More public video services
Combine other video capabilities based on model style and workflow needs.
Bring MiniMax H3 into your product
One token, one generation endpoint, and one task system for multimodal video experiences.