Video Generation

POSThttps://api.tokenbyte.ai/v1/videos

Use the unified video task endpoint with Wonder-Ultra, Wonder-Pro, or Wonder-Standard. Create an API key in the console and select the SeeDance key group.

Before you start

Image, video, and audio inputs must use URLs accessible by the service. Put all reference media in medias.

`job_type`	Purpose
`text_to_video`	Text to video
`image_to_video`	Image to video
`first_last_frame`	First-and-last-frame video
`reference_to_video`	Multimodal reference video
`video_extend`	Video extension; `Wonder-Ultra` only
`video_edit`	Video editing; `Wonder-Ultra` only

Text to video

curl --request POST 'https://api.tokenbyte.ai/v1/videos' \
  --header 'Authorization: Bearer sk-replace-with-your-key' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "Wonder-Standard",
    "job_type": "text_to_video",
    "prompt": "An orange cat walks through a neon street after the rain, cinematic camera movement, smooth natural motion",
    "duration": 5,
    "resolution": "720P",
    "aspect_ratio": "16:9",
    "n": 1,
    "job_parameters": {
      "EnableAudio": false,
      "Watermark": false
    }
  }'

Image to video

curl --request POST 'https://api.tokenbyte.ai/v1/videos' \
  --header 'Authorization: Bearer sk-replace-with-your-key' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "Wonder-Standard",
    "job_type": "image_to_video",
    "prompt": "The person turns naturally toward the camera with smooth, cinematic movement",
    "medias": [
      {
        "Type": "image",
        "Url": "https://example.com/reference.jpg"
      }
    ],
    "duration": 5,
    "resolution": "720P",
    "aspect_ratio": "16:9",
    "n": 1,
    "job_parameters": {
      "EnableAudio": false,
      "Watermark": false
    }
  }'

Reference media to video

curl --request POST 'https://api.tokenbyte.ai/v1/videos' \
  --header 'Authorization: Bearer sk-replace-with-your-key' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "Wonder-Pro",
    "job_type": "reference_to_video",
    "prompt": "Keep the referenced person consistent and reproduce the action, expression, rhythm, and camera movement",
    "medias": [
      {
        "Type": "image",
        "Url": "https://example.com/person.jpg"
      },
      {
        "Type": "video",
        "Url": "https://example.com/action.mp4"
      }
    ],
    "duration": 10,
    "resolution": "720P",
    "aspect_ratio": "adaptive",
    "n": 1,
    "job_parameters": {
      "EnableAudio": true,
      "Watermark": false
    }
  }'

Check a task and download the video

Save the id returned by the create request. Download the content after the task reaches completed.

curl --request GET 'https://api.tokenbyte.ai/v1/videos/TASK_ID' \
  --header 'Authorization: Bearer sk-replace-with-your-key'

curl --location --request GET \
  'https://api.tokenbyte.ai/v1/videos/TASK_ID/content' \
  --header 'Authorization: Bearer sk-replace-with-your-key' \
  --output result.mp4

Successful submission response:

{
  "id": "task_public-id",
  "object": "video",
  "model": "Wonder-Standard",
  "status": "queued",
  "progress": 0,
  "created_at": 1788429073
}

Request parameters

Parameter	Type	Required	Description
`model`	string	Yes	`Wonder-Ultra`, `Wonder-Pro`, or `Wonder-Standard`.
`job_type`	string	No	Explicitly recommended; inferred from the presence of media when omitted.
`prompt`	string	Yes	Video generation prompt.
`scene`	string	No	Reserved scene type; default `general`.
`medias`	object[]	Depends on task	Media inputs; see below.
`job_parameters`	object	No	Audio, watermark, output format, and other feature options.
`resolution`	string	No	`720P`, `1080P`, or `4K`; default `720P`. Actual support depends on the model.
`aspect_ratio`	string	No	`16:9`, `9:16`, `4:3`, `3:4`, `1:1`, `21:9`, or `adaptive`; default `16:9`.
`duration`	number	No	Default 5 seconds; usually 4–15 seconds, up to 30 for `Wonder-Ultra`.
`n`	integer	No	Output count; currently only `1` is supported.

image_to_video requires one image; first_last_frame requires two. Wonder-Pro and Wonder-Standard accept up to 15 reference items. Wonder-Ultra accepts up to 50: 30 images, 10 videos, and 10 audio files.

Each media object accepts Type (image, video, or audio) plus either Url or MediaId. The lowercase forms type, url, and media_id are also supported.

Media containing real faces

Wonder models do not accept reference images or videos containing real faces directly by URL. Register them with Yike ImportMedia first and use the returned MediaId. This endpoint does not perform the registration for you.

Media	Formats	Size and dimension limits
Image	jpeg, jpg, png, webp	Under 20 MB; short edge at least 400 px and long edge at most 8000 px
Video	mp4, mov	Up to 100 MB; width and height 300–4096 px; total pixels 409600–927408
Audio	mp3, wav	Up to 15 MB

`job_parameters`

Field	Type	Description
`EnableAudio`	boolean	Whether to produce audio.
`Watermark`	boolean	Whether to add an “AI generated” watermark.
`OutputFormat`	string	`mov` or `mp4`; `Wonder-Ultra` only.
`ReturnLastFrame`	boolean	Whether to return the last frame.