Video Generation

The MiniMax H3 (Hailuo-03) endpoint supports text-to-video, first-frame image-to-video, first-and-last-frame video, and multi-reference video generation through POST /v1/videos. The interface is async: on success it returns a video task object with status=queued; poll the Query Result endpoint with the returned id until status becomes completed, then retrieve the final video through the content endpoint.

Billing model

This endpoint is billed under the model specified by the model field (MiniMax-H3), deducted from your account balance. Add the model on the channel first (select channel type MiniMax).

Supported model and parameters

Model: MiniMax-H3 (Hailuo-03);
Text: every request needs non-empty text, supplied as top-level prompt or a type=text item in metadata.content;
Duration seconds: an integer from 4 to 15 (default 5);
Resolution: set size, top-level resolution, or metadata.resolution to 768P or 2K (default 768P). Pixel dimensions such as 1280x720 are not accepted as size;
Aspect ratio: use metadata.ratio with 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive. Text-only requests default to 16:9; requests with an image or video default to adaptive;
Other integrated fields: metadata.callback_url and metadata.aigc_watermark.

Supported input modes

Text-to-video: provide only prompt;
First-frame guidance: provide input_reference / image, one images entry, or metadata.first_frame_image;
First-and-last-frame guidance: provide two images entries (first, then last), or use metadata.first_frame_image and metadata.last_frame_image;
Multi-reference generation: use metadata.content with reference_image, reference_video, and reference_audio items. Up to 9 input images, 3 reference videos, and 3 reference audio items are accepted;
Frame images cannot be mixed with reference media. metadata.content is the complete multimodal content array; if it already contains a text item, top-level prompt is not appended.

Authorization

BearerAuth

AuthorizationBearer <token>

使用 Bearer Token 认证

In: header

Request Body

model*string

模型名称

prompt?string

文本提示词。若 metadata.content 已含非空 text 项，可省略

seconds?integer

视频时长（秒），默认 5

Default5

Range4 <= value <= 15

size?string

输出分辨率，不是像素尺寸

Default"768P"

Value in"768P" | "2K"

resolution?string

size 的等价分辨率字段

Value in"768P" | "2K"

input_reference?string

首帧图片 URL

image?string

首帧图片 URL

images?array<string>

帧图片 URL。第一个为首帧，第二个为尾帧

Itemsitems <= 2

metadata?

Response Body

application/json

curl -X POST "https://www.omwai.xyz/v1/videos" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-H3",
    "prompt": "A calico cat playing piano on a stage",
    "seconds": 5,
    "size": "768P",
    "metadata": {
      "ratio": "16:9"
    }
  }'

curl -X POST "https://www.omwai.xyz/v1/videos" \  -H "Content-Type: application/json" \  -d '{    "model": "MiniMax-H3",    "prompt": "A calico cat playing piano on a stage",    "seconds": 5,    "size": "768P",    "metadata": {      "ratio": "16:9"    }  }'

{
  "id": "string",
  "object": "video",
  "model": "MiniMax-H3",
  "status": "queued",
  "progress": 100,
  "created_at": 0,
  "completed_at": 0
}

Empty

Async flow

POST /v1/videos submits and returns an id with status=queued;
Use that id as task_id when polling GET /v1/videos/{task_id};
When status=completed, retrieve the video via GET /v1/videos/{task_id}/content.

Video Generation

Authorization

Request Body

Response Body

200application/json

400

application/json