query_generation_task

Poll the status of an asynchronous generation task created by generate_video. When the task reaches Success, the server automatically retrieves the download URL for the generated video.

Parameters

Parameter	Type	Required	Default	Description
`task_id`	string	✓	—	Task ID returned by `generate_video`
`api`	string	—	`auto`	Which API the task was created with: `v1` (V1 models), `v2` (MiniMax-H3), or `auto` (default) — tries V2 first and falls back to V1 when the task is not found.

Task Statuses

Status	Meaning	What to do
`Preparing`	Task is being prepared	Wait and poll again
`Queueing`	Task is queued	Wait and poll again
`Processing`	Video is being generated	Wait and poll again
`Success`	Video is ready	Download URL is provided in response
`Fail`	Generation failed	Try again with a different prompt

Usage Examples

Check status:

“Check the status of video generation task 106916112212032.”

Poll until ready:

“Is my video ready? The task ID is 176843862716480.”

Two-Step Workflow

Step 1: generate_video(prompt="A golden retriever playing fetch on a beach")
        → { task_id: "106916112212032" }

Step 2 (repeat until Success):
        query_generation_task(task_id="106916112212032")
        → { status: "Processing" }   ← wait 30–60 seconds and retry
        → { status: "Success", downloadUrl: "https://..." }

Response

Returns a text part with:

Task ID — the task being polled
Status — current task status
API — the API version used (v1 or v2)
Download URL — the URL to access the generated video (on Success only)
Resolution — video dimensions (on Success only)
Error — the underlying error message (on Fail only)

Notes

Video generation typically takes 1–5 minutes. Poll every 30–60 seconds.
With the default api: 'auto', V2 (MiniMax-H3) tasks are queried first, with automatic fallback to V1 when the task is not found — you don’t need to track which API created the task.
If you know the task came from a V1 model, you can pass api: 'v1' to skip the V2 attempt.
The download URL is retrieved automatically when the status is Success — no separate download step is needed.
On Fail, retry generate_video with a revised prompt or different model.