MiniMax video generation

MiniMax submits async video tasks through an OpenAI-compatible endpoint, primarily via JSON. Currently only the model is supported.

POST /v1/videos

A successful submission returns id. Poll results with Task query. Total request body size must not exceed 64 MB. Prefer public URLs for large files instead of Base64.

Available models

Generation scenarios (auto-detected from request parameters)

Scenario	Parameters	Aspect ratio
Text-to-video	only	Default ; is not allowed
Image-to-video	or	Forced
Multimodal reference-to-video	/	Default

Image-to-video and multimodal reference-to-video are mutually exclusive and cannot be mixed.

Headers

Name	Type	Required	Description
	string	Yes	Request authentication. Use a Bearer Token, for example:
	string	Yes	Content type. Must be

Request body

Name	Type	Required	Description
	string	Yes	Model name
	string	Yes	Prompt. Required and non-empty for all generation scenarios. Max 7000 characters per prompt
	integer	No	Duration in seconds, 4–15, default . Values outside the range are clamped by the gateway
	string	No	Resolution: / , default
	string	No	Single input image as the first frame. Mutually exclusive with ; if both are sent, image takes precedence
	array<string>	No	Multiple reference images (image-to-video)
/	string	No	Multimodal reference assets (reference-to-video)

Request example

curl -X POST https://octopusx.ai/v1/videos \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "MiniMax-H3",
  "prompt": "A flower bud blooming in a time-lapse shot",
  "duration": 5,
  "size": "2K"
}'

Response example

{
  "id": "task_a1bxxxxxe5f6",
  "task_id": "task_a1bxxxxxe5f6",
  "object": "video",
  "model": "MiniMax-H3",
  "status": "queued",
  "progress": 0,
  "created_at": 1785125529
}

Response fields

Name	Type	Description
	string	Task ID for subsequent queries (public gateway ID)
	string	Legacy-compatible task ID (to be deprecated; same value as ). Prefer id for new integrations
	string	Always
	string	Model name used by the task
	string	Task status: / / /
	integer	Progress percentage 0–100
	integer	Creation timestamp

Other formats and modes

Official format (/v1/video/generations)

Create tasks with a multimodal array (organized like the Seedance official format). Query with GET /v1/video/generations/{task_id}.

Remix (/v1/videos/{video_id}/remix)

Remix a completed 768P video to 2K. By task ID: pass the source task ID as path parameter (recommended; no prompt required). By source video: pass in the body (must include all original inputs, and must match the original generation exactly). Only MiniMax-H3 768P output videos are supported, and the source task must be within the 7-day query window.

Related endpoints

• Video APIs overview
• MiniMax task query
• Error codes