MiniMax-H3 video generation

Call MiniMax-H3 through APIMaster's OpenAI-compatible video API. Text-to-video requests run asynchronously.

Base URL: https://apimaster.ai/v1 is https://apimaster.ai/v1. Create MiniMax-H3 tasks with POST https://apimaster.ai/v1/video/generations.

Endpoint compatibility and validation

Both POST https://apimaster.ai/v1/video/generations and POST https://apimaster.ai/v1/videos/generations are supported. For the same model, request parameters and billing rules are identical; each endpoint retains its existing creation response format. Existing clients can keep their current path.

MiniMax-H3 retains the singular example on this page; size and resolution are aliases, as are ratio and aspect_ratio. Unsupported specifications or conflicting aliases return HTTP 400 rather than silently selecting a higher resolution. Specify resolution, aspect ratio, and duration explicitly.

Create a video task

POST https://apimaster.ai/v1/video/generations

Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

Field	Type	Required	Description
model	string	yes	Must be MiniMax-H3
prompt	string	conditional	Description of the video
duration	integer	no	4–15 seconds; defaults to 4
size	string	no	768P or 2K; defaults to 768P
ratio	string	no	Aspect ratio; defaults to 16:9
metadata	object	no	Extra parameters; explicit content overrides prompt/images, and top-level V2 fields override metadata equivalents

curl -s "https://apimaster.ai/v1/video/generations" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-H3",
    "prompt": "A calm ocean at sunrise, gentle waves, cinematic wide shot",
    "duration": 4,
    "size": "768P",
    "ratio": "16:9"
  }'

The response returns a task ID immediately:

{
  "id": "task_xxxxxxxx",
  "task_id": "task_xxxxxxxx",
  "object": "video",
  "model": "MiniMax-H3",
  "status": "queued",
  "progress": 0,
  "created_at": 1786436206
}

Poll task status

GET https://apimaster.ai/v1/videos/{task_id}

curl -s "https://apimaster.ai/v1/videos/task_xxxxxxxx" \
  -H "Authorization: Bearer YOUR_API_KEY"

Possible states are queued, in_progress, completed, and failed. A completed task includes the video URL:

{
  "id": "task_xxxxxxxx",
  "status": "completed",
  "progress": 100,
  "url": "https://apimaster.ai/v1/videos/task_xxxxxxxx/content"
}

Wait 10–20 seconds after submission, then poll every 3–5 seconds.

Download the video

GET https://apimaster.ai/v1/videos/{task_id}/content

curl -L "https://apimaster.ai/v1/videos/task_xxxxxxxx/content" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -o output.mp4

Video URLs may expire. Download and store the MP4 promptly after completion.

Limits and errors

• model must be MiniMax-H3.
• duration supports 4–15 seconds.
• size supports 768P and 2K.
• A disabled or insufficient-balance channel returns a routing or upstream balance error; check the account balance and channel status first.
• Do not download a task unless its status is completed.

Reference-to-video (r2v)

Send the official V2 content array to the APIMaster endpoint above. A non-empty text item is required; a top-level prompt is not required when content is supplied. reference_image guides the generated subject or style; it does not lock the first frame. Legacy images maps the first two images to first/last frames. Do not mix first/last frames with reference roles.

{
  "model": "MiniMax-H3",
  "content": [
    {"type": "text", "text": "Use the character in Image 1 as the subject, walking through a garden."},
    {"type": "image_url", "image_url": {"url": "https://YOUR_PUBLIC_IMAGE_URL/reference.jpg"}, "role": "reference_image"}
  ],
  "resolution": "768P",
  "duration": 4,
  "ratio": "16:9"
}

Keep using POST https://apimaster.ai/v1/video/generations; /v1/v2/video_generation is not a supported path. This endpoint accepts the V2 body, while creation responses and polling retain the APIMaster task format shown above. Top-level content takes precedence over metadata.content; explicit content is not merged with images.