Video Regeneration
Use `POST /v1/video/remix` to regenerate a generated 768P video to 2K.
https://zx1.deepwl.net/v1/video/remixcurl -X POST https://zx1.deepwl.net/v1/video/remix \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '
{
"model": "MiniMax-H3",
"task_id": "task_a1b2c3d4e5f6"
}'{
"id": "task_a1b2c3d4e5f6",
"task_id": "task_a1b2c3d4e5f6",
"object": "video",
"model": "MiniMax-H3",
"status": "queued",
"progress": 0,
"created_at": 1785125529
}Regenerates a generated 768P video to 2K; the output resolution is fixed at 2K. The OpenAI format counterpart is Video Regeneration; this API passes the source task via the request body task_id.
Two ways to provide the source video
| Method | Parameter | prompt |
|---|---|---|
| By task ID (recommended) | Request body task_id | Not required |
| By source video | metadata.base_video | Required, must exactly match the original generation |
When using the source video method, you must attach all inputs used to generate the 768P video (prompt, images, reference video/audio) unchanged; any mismatch may fail to achieve the expected result.
- The gateway converts
task_idto the upstream task ID automatically; it is not open for callers to specify or return. - Only videos matching the MiniMax-H3 768P output specification can be regenerated. The task ID method requires an upstream whitelist, and the source task must be within the 7-day query window.
Request Headers
Authorizationstring必填Authentication header. Uses a Bearer token, e.g. Bearer YOUR_API_KEY.
Content-Typestring必填Request content type, must be application/json.
Request Body
modelstring必填Model name. Currently only supports MiniMax-H3.
task_idstring必填The gateway task ID of the source video (the id returned by the creation API). The gateway converts it to the upstream task ID automatically; mutually exclusive with metadata.base_video.
promptstringPrompt.
- Not required when regenerating by task ID.
- Required when regenerating via
metadata.base_video, and must exactly match the finalpromptactually sent to the model when generating the 768P source video.
durationintegerRegenerated video duration in seconds, used for billing pre-deduction; should match the source video duration.
imagestringThe first-frame image of the original task. When using the base_video method and the original task was image-to-video, this must be attached unchanged.
imagesarrayThe image inputs of the original task (first frame, last frame). Must match the generation when using the base_video method.
metadataobjectExtended configuration.
base_videostring768P source video URL (regenerate by source video). Must meet the MiniMax-H3 768P output specification: with audio track, 24 fps, width and height divisible by 32, area up to 768×1344, total frames 107–362.
base_video_durationintegerSource video duration in seconds, used for billing pre-deduction; must be passed by the caller.
reference_videostringThe reference video used by the original task. Must match the generation when using the base_video method.
reference_audiostringThe reference audio used by the original task. Must match the generation when using the base_video method.
aigc_watermarkbooleanWhether to add an AIGC watermark, default false.
Request Example
Response Example
After submission, poll the new returned id via Query Task. Response fields are identical to the creation API; see MiniMax Video Generation.
Error Codes
| Status Code | Description |
|---|---|
| 400 | Invalid parameters: no source video provided, or missing prompt when using base_video |
| 401 | Authentication failed |
| 402 | Insufficient quota |
| 429 | Rate limited |
| 500 | Internal server error |