Video

HappyHorse Videos

Call HappyHorse through the unified video API for text-to-video, image-to-video, reference-to-video and video editing.

POST /v1/video/generations

HappyHorse is called through the unified video API; tasks are submitted and queried the same way as for other video models. You do not pick a generation mode — the gateway chooses it from the inputs you send:

Inputs	Mode
`prompt` only	Text-to-video
One image (`image`, or `images` with a single entry)	Image-to-video, the image is the first frame
2–9 images (`images`)	Reference-to-video
A video (`input_reference`), plus 0–9 reference images	Video editing

Authentication

Send your key in the Authorization: Bearer sk-xxxx header.

Request body

Content type: application/json.

Parameter	Type	Required	Description
`model`	string	Yes	Model ID, for example `happyhorse-1.0`.
`prompt`	string	Yes	Text description of the video.
`image`	string	No	First-frame image URL.
`images`	string[]	No	Image URLs, up to 9. A single image is used as the first frame; several are used as reference images.
`input_reference`	string	No	URL of the video to edit. When present the task is a video edit and `images` are reference images.
`duration`	integer	No	Video length in seconds, `3`–`15`, default `5`. Not used for video editing — the output matches the source length.
`size`	string	No	Resolution: `720P` or `1080P`, default `1080P`. `720p`, `1280x720` and similar forms are also accepted.
`ratio`	string	No	Aspect ratio: `16:9`, `9:16`, `1:1`, `4:3`, `3:4`. `aspect_ratio` is also accepted.
`metadata`	object	No	Extra parameters passed to the provider, such as `seed` (integer) and `watermark` (boolean, default `false` — no watermark).

The content array format shown on the model details page also works: text is the prompt, image_url is an image ("role": "first_frame" marks the first frame), video_url is the video to edit, and resolution replaces size.

Image and video URLs must be publicly reachable over HTTPS; URLs that need a login or have expired signatures do not work. A duration, resolution or aspect ratio outside the ranges above is rejected with 400 and is not charged.

Request examples

Text-to-video

curl https://rutoken.ai/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "model": "happyhorse-1.0",
    "prompt": "An orange cat runs across a sunny meadow, tracking shot",
    "duration": 5,
    "size": "720P",
    "ratio": "16:9"
  }'

Reference-to-video

curl https://rutoken.ai/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "model": "happyhorse-1.0",
    "prompt": "The person from reference 1 picks up the fan from reference 2, slow push-in",
    "images": [
      "https://example.com/person.png",
      "https://example.com/fan.png"
    ],
    "duration": 5,
    "size": "1080P",
    "ratio": "9:16"
  }'

Video editing

curl https://rutoken.ai/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "model": "happyhorse-1.0",
    "prompt": "Restyle the video as a watercolor painting, keep the motion",
    "input_reference": "https://example.com/input.mp4",
    "images": ["https://example.com/style.png"],
    "size": "720P"
  }'

The response is a task object; query the task with its id. See the unified video API for the response format, task statuses and polling.

Billing

HappyHorse is billed per second of video, with different prices for 720P and 1080P; see the model marketplace for prices.

Text-, image- and reference-to-video: billed by duration.
Video editing: the provider bills input video length + output video length, and the output is as long as the source, so a 5-second edit is billed as about 10 seconds. Submission pre-charges twice the source length (30 seconds if the source length cannot be read), and the task is settled on the actual seconds reported by the provider when it finishes.

If a task fails, the charged quota is refunded automatically. Provider video URLs expire, so download the video soon after the task succeeds.