MiniMax-H3

MiniMax-H3 provides asynchronous multimodal video generation through AnyFast. Create a task with POST /minimax/v2/video_generation, then query it with GET /minimax/v2/query/video_generation/{task_id}.

For specific content requests, select the MiniMax-H3-nsfw model under the Direct resource group.

The /minimax prefix is part of the AnyFast route and is required in both requests.

Create video generation task
Query video generation task

Create task

POST /minimax/v2/video_generation

cURL

curl --request POST \
  --url https://www.anyfast.ai/minimax/v2/video_generation \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "MiniMax-H3",
    "content": [
      {
        "type": "text",
        "text": "A red paper airplane glides above the shoreline at sunrise as the camera follows from a low angle."
      }
    ],
    "resolution": "768P",
    "duration": 4,
    "ratio": "21:9",
    "aigc_watermark": false
  }'

200

{
  "task_id": "asyntask_example123"
}

Request headers

string

required

Bearer authentication in the form Bearer YOUR_API_KEY.

string

default:"application/json"

required

Must be application/json.

Request parameters

string

required

Case-sensitive model ID. Use MiniMax-H3; for specific content requests, use MiniMax-H3-nsfw under the Direct resource group.

object[]

required

Multimodal input parts. Include exactly one non-empty text item and any media required by the selected workflow.

string

required

text, image_url, video_url, or audio_url.

string

Prompt for a text part. Maximum 7,000 characters.

string

Purpose of a media part: first_frame, last_frame, reference_image, reference_video, or reference_audio.

object

Required when type is image_url.

string

Public HTTP or HTTPS image URL. Base64 is not supported.

object

Required when type is video_url.

string

Public HTTP or HTTPS reference-video URL.

object

Required when type is audio_url.

string

Public HTTP or HTTPS reference-audio URL.

string

required

768P or 2K. Values are case-sensitive.

integer

required

An integer from 4 through 15, in seconds.

string

adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16. See the workflow rules below.

string

Optional callback URL. It must return MiniMax’s challenge value within 3 seconds before status callbacks can be enabled.

boolean

default:"false"

Whether to add an AIGC watermark. Send a JSON boolean.

Workflow rules

Text to video: use one text part only. ratio is required and cannot be adaptive.
Image to video: add one first_frame or last_frame, or one of each. The output ratio follows the images and resolves to adaptive; another supplied ratio is ignored.
Reference to video: combine reference_image, reference_video, and reference_audio parts. Include at least one reference image or video; audio alone is not sufficient. ratio is optional and defaults to adaptive.
Do not mix first_frame or last_frame with any reference_* role.

Media constraints

The request body must not exceed 64 MB. Use public URLs rather than embedding file data.
Images support JPG, JPEG, PNG, WEBP, HEIC, and HEIF. Each image can be up to 30 MB, each dimension must be 256-5,760 px, and the aspect ratio must be 0.4-2.5. Use at most one first frame, one last frame, or nine reference images.
Reference videos support MP4 and MOV containers, H.264/H.265 video, and AAC/MP3 audio. Each file can be up to 50 MB and 2-15 seconds. Use at most three videos with no more than 15 seconds total. Each dimension must be 256-5,760 px, the aspect ratio must be 0.4-2.5, and the frame rate must be 23.976-60 fps.
Reference audio supports WAV and MP3. Each file can be up to 15 MB and 2-15 seconds. Use at most three audio files with no more than 15 seconds total.
Use at most 12 media files across images, videos, and audio.

A successful create response confirms task creation, not media validation or generation success. Remote media errors can appear later as a failed task. Each accepted create request is a separate task, so poll the task ID instead of retrying the POST.

Response

string

Video generation task ID. Use it with the query endpoint.

Error response

400

{
  "error": {
    "http_code": "400",
    "message": "resolution must be 768P or 2K",
    "type": "invalid_request_error"
  },
  "request_id": "request_example123",
  "type": "error"
}

string

Top-level response type. Error responses use error.

object

Error details.

string

HTTP status code represented as a string.

string

Human-readable error detail.

string

Error category, such as invalid_request_error or authorized_error.

string

Request ID for troubleshooting.

Query task

GET /minimax/v2/query/video_generation/{task_id}

cURL

curl --request GET \
  --url https://www.anyfast.ai/minimax/v2/query/video_generation/asyntask_example123 \
  --header 'Authorization: Bearer YOUR_API_KEY'

200

{
  "task": {
    "id": "asyntask_example123",
    "model": "MiniMax-H3",
    "status": "succeeded",
    "created_at": 1785936718,
    "updated_at": 1785936805,
    "content": {
      "url": "https://example.com/generated-video.mp4"
    },
    "resolution": "768P",
    "duration": 4,
    "usage": {
      "total_seconds": 4,
      "output_seconds": 4,
      "input_image_count": 0
    },
    "ratio": "21:9",
    "task_type": "generation"
  }
}

Request headers

string

required

Bearer authentication in the form Bearer YOUR_API_KEY.

Path parameters

string

required

Task ID returned by the create endpoint.

Response fields

object

Queried video generation task.

string

Task ID.

string

Model used for the task.

string

queued, running, succeeded, failed, or cancelled.

integer

Task creation time as a Unix timestamp.

integer

Most recent task update time as a Unix timestamp.

object

Result content. Usually present only after successful generation.

string

Generated video URL. Download it promptly because the URL may expire.

object

Failure detail. Present when the task fails.

string

Upstream error code.

string

Failure reason.

string

Requested output resolution.

integer

Requested output duration in seconds.

object

Usage metrics. Some zero-value fields may be omitted.

number

Sum of the reported input and output seconds.

number

Processed reference-video duration. May be omitted when no reference video is used.

number

Generated output duration used by the API.

integer

Number of input images.

string

Resolved output aspect ratio.

string

generation for this endpoint.

Continue polling only while the status is queued or running. A succeeded, failed, or cancelled status is terminal.

Overview

Native Compatible Endpoints

Model APIs

Create task

Request headers

Request parameters

Workflow rules

Media constraints

Response

Error response

Query task

Request headers

Path parameters

Response fields

​Create task

​Request headers

​Request parameters

​Workflow rules

​Media constraints

​Response

​Error response

​Query task

​Request headers

​Path parameters

​Response fields

Create task

Request headers

Request parameters

Workflow rules

Media constraints

Response

Error response

Query task

Request headers

Path parameters

Response fields