Create Context-IR Prompt Task

POST /v2/h3_context_ir

Generate a video prompt asynchronously. The result is text. Set request model=MiniMax-H3; the token and group must allow MiniMax-H3-Context-IR. Authenticate with a platform Bearer model API key.

Request parameters

Parameter	Type	Required	Description
model	string	Yes	Fixed MiniMax-H3
content	array<object>	Yes	At least one non-empty text item; maximum 7,000 characters per item; see media rules for roles and combinations
duration	integer	Yes	Target video duration, 4–15 seconds
ratio	string	Required for text-only	Text-only input requires a concrete ratio such as 16:9; adaptive is allowed with frame or reference inputs
callback_url	string	No	Customer callback URL; must pass challenge verification

Omit resolution and source_task_id. Context-IR does not generate a video. Your application can read the generated prompt and use it in a subsequent video request; video generation is a separate, separately billed task.

Request example

{
  "model": "MiniMax-H3",
  "duration": 4,
  "ratio": "16:9",
  "content": [
    {"type": "text", "text": "Expand this idea into a video prompt: a kitten stretches in the sunshine as the camera slowly moves closer"}
  ]
}

Response (HTTP 200)

{"task_id":"task_platform_ir"}

Save the platform task_id and query the task. On success, task.task_type=h3_context_ir and task.modality=text. Read task.content.prompt, not a video URL.

{
  "task": {
    "id": "task_platform_ir",
    "model": "MiniMax-H3",
    "status": "succeeded",
    "task_type": "h3_context_ir",
    "modality": "text",
    "content": {"prompt": "In warm afternoon sunlight, a kitten slowly stretches as the camera smoothly moves closer..."}
  }
}

Context-IR is settled using input/output tokens at the pricing configured in the console. Queries and customer callbacks do not expose upstream usage.