HappyHorse Video Integration

Call HappyHorse through the ModelSell generic video API or the DashScope-compatible API and poll asynchronous tasks.

HappyHorse is an asynchronous video task API. ModelSell supports both its generic video format and a DashScope-compatible native format. Both formats use the same HappyHorse task capability.

Supported Models

Model	Use case	Media field
`happyhorse-1.0-t2v`	Text to video	No media required
`happyhorse-1.0-i2v`	Image to video	`image`, or `first_frame` in the native format
`happyhorse-1.0-r2v`	Multi-image reference video	`images[]`, or `reference_image` in the native format
`happyhorse-1.0-video-edit`	Video editing	`metadata.video_url`, optionally with `images[]` as references

Choose an API Format

Format	Create task	Poll task
ModelSell generic	`POST /v1/video/generations`	`GET /v1/video/generations/{task_id}`
DashScope-compatible native	`POST /api/v1/services/aigc/video-generation/video-synthesis`	`GET /api/v1/tasks/{task_id}`

Authenticate every request with your ModelSell API key:

Authorization: Bearer $MODELSELL_API_KEY

After creating a task, store the public id or task_id returned by ModelSell. Use this public task ID for polling, not the upstream task ID.

Generic Format

The generic API accepts JSON. ModelSell converts media fields to HappyHorse input.media entries according to the selected model:

Generic field	HappyHorse upstream field
`prompt`	`input.prompt`
`size`	`parameters.resolution`
`seconds` or `duration`	`parameters.duration`
`image`	A `first_frame` item in `input.media`; only the first image is used
`images[]`	`reference_image` items in `input.media`
`metadata.video_url`	A `video` item in `input.media`
`metadata.ratio`	`parameters.ratio`
`metadata.watermark`	`parameters.watermark`
`metadata.seed`	`parameters.seed`
`metadata.audio_setting`	`parameters.audio_setting`

Text to Video

curl -X POST "$MODELSELL_BASE_URL/v1/video/generations" \
  -H "Authorization: Bearer $MODELSELL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-t2v",
    "prompt": "A miniature city lights up at night as the camera slowly moves forward",
    "size": "720P",
    "seconds": "5",
    "metadata": {
      "ratio": "16:9",
      "watermark": false,
      "seed": 7
    }
  }'

Image to Video

{
  "model": "happyhorse-1.0-i2v",
  "prompt": "Make the cat run forward across the grass",
  "image": "https://example.com/first-frame.png",
  "size": "720P",
  "duration": 5
}

Multi-Image Reference Video

{
  "model": "happyhorse-1.0-r2v",
  "prompt": "The person in [Image 1] picks up the fan from [Image 2]",
  "images": [
    "https://example.com/character.jpg",
    "https://example.com/fan.jpg"
  ],
  "duration": 5,
  "metadata": {
    "ratio": "4:3"
  }
}

ModelSell converts images[] to reference_image items in the same order. Prompts can refer to them as [Image 1], [Image 2], and so on.

Video Editing

{
  "model": "happyhorse-1.0-video-edit",
  "prompt": "Dress the character in the sweater from the reference image",
  "images": [
    "https://example.com/sweater.webp"
  ],
  "duration": 5,
  "metadata": {
    "video_url": "https://example.com/input.mp4",
    "audio_setting": "origin"
  }
}

happyhorse-1.0-video-edit requires at least one publicly reachable video in metadata.video_url. Without it, the upstream API returns At least one video item is required in media list.

Poll a Generic Task

curl "$MODELSELL_BASE_URL/v1/video/generations/task_xxxxxxxxxxxxxxxx" \
  -H "Authorization: Bearer $MODELSELL_API_KEY"

ModelSell maps PENDING to queued, RUNNING to processing, SUCCEEDED to succeeded, and FAILED or CANCELED to failed. On success, read the final URL from metadata.url; on failure, inspect the error information.

{
  "id": "task_xxxxxxxxxxxxxxxx",
  "task_id": "task_xxxxxxxxxxxxxxxx",
  "object": "video.generation",
  "status": "succeeded",
  "model": "happyhorse-1.0-r2v",
  "seconds": "5",
  "metadata": {
    "url": "https://example.com/happyhorse-result.mp4",
    "request_id": "request_xxxxxxxxxxxxxxxx",
    "usage": {
      "duration": 5,
      "ratio": "16:9"
    }
  }
}

DashScope-Compatible Native Format

Existing DashScope HappyHorse clients can switch their base URL and API key to ModelSell. ModelSell adds the asynchronous upstream request header automatically.

curl -X POST "$MODELSELL_BASE_URL/api/v1/services/aigc/video-generation/video-synthesis" \
  -H "Authorization: Bearer $MODELSELL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-i2v",
    "input": {
      "prompt": "A cat runs across the grass",
      "media": [
        {
          "type": "first_frame",
          "url": "https://example.com/first-frame.png"
        }
      ]
    },
    "parameters": {
      "resolution": "720P",
      "duration": 5,
      "ratio": "16:9",
      "watermark": false,
      "seed": 7
    }
  }'

The native response preserves DashScope fields such as output.task_status, while output.task_id contains the public ModelSell task ID:

{
  "request_id": "request_xxxxxxxxxxxxxxxx",
  "output": {
    "task_id": "task_xxxxxxxxxxxxxxxx",
    "task_status": "PENDING"
  }
}

Poll with the same public task ID:

curl "$MODELSELL_BASE_URL/api/v1/tasks/task_xxxxxxxxxxxxxxxx" \
  -H "Authorization: Bearer $MODELSELL_API_KEY"

On success, output.video_url in the native response contains the final video URL:

{
  "request_id": "request_xxxxxxxxxxxxxxxx",
  "output": {
    "task_id": "task_xxxxxxxxxxxxxxxx",
    "task_status": "SUCCEEDED",
    "video_url": "https://example.com/happyhorse-result.mp4"
  },
  "usage": {
    "duration": 5,
    "ratio": "16:9"
  }
}

Video generation is asynchronous. Poll every 5 to 10 seconds. Every reference image and video URL must be directly reachable by the upstream service.