GlobalAiOpcGlobalAiOpc
wechatwechat
Sign in
GlobalAiOpc

Seedance2.0 Special Price - Create Task

Create Seedance2.0 special-price video tasks

POST/kyyReactApiServer/v1/seedance-special/videos

Seedance2.0 Special Price Create Task

Use a full model code to create a video task. The model code includes model type, resolution, and whether video reference is used.

认证

获取 Key
cURL
Authorization: Bearer {{key}}

Request Parameters

modelstringrequired
Model name. When using a _with_video_ref model, content must include a video_url item with role=reference_video.
  • sd_2.0_special_720p
  • sd_2.0_special_1080p
  • sd_2.0_special_2k
  • sd_2.0_special_4k
  • sd_2.0_special_720p_with_video_ref
  • sd_2.0_special_1080p_with_video_ref
  • sd_2.0_special_2k_with_video_ref
  • sd_2.0_special_4k_with_video_ref
  • sd_2.0_fast_special_720p
  • sd_2.0_fast_special_720p_with_video_ref
sd_2.0_fast_special only supports 720p.
resolutionstring
Optional. The full model code already includes resolution. If provided, it must match the resolution in model.
ratiostring
Aspect ratio. Default: 16:9. Supported values: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptive.
durationinteger
Duration in seconds. Default: 5. Range: 4-15.
generate_audioboolean
Whether to generate audio. Default: true.
return_last_frameboolean
Whether to return the last-frame image. Default: false. When set to true and the task generates a last frame, read last_frame_url from the query API.
toolsarray
Tool configuration. Default: not sent. Currently only the Seedance 2.0 web search tool is exposed: {"type":"web_search"}.
tools[].typestring
Tool type. Currently supported value: web_search. Other tool types are not supported.
seedinteger
Random seed. Default: -1, which lets the model service choose a random seed. A fixed value improves reproducibility, but generation is still probabilistic.
contentarrayrequired
Content array. One type=text prompt is required.
content[].typestringrequired
text, image_url, video_url, or audio_url.
content[].textstring
Required when content[].type=text. The text prompt.
content[].rolestring
first_frame, last_frame, reference_image, reference_video, or reference_audio.
content[].image_url.urlstring
Image URL. Public URL or uploaded asset reference: assetId://{assetId}. Base64 input is not supported.
content[].video_url.urlstring
Video URL. Public URL or uploaded asset reference: assetId://{assetId}.
content[].audio_url.urlstring
Audio URL. Public URL or uploaded asset reference: assetId://{assetId}. Base64 input is not supported.
For real-person or avatar media, upload the asset first and confirm that its status is ACTIVE through the asset detail API before referencing it with assetId://{assetId} and creating a task. Do not create a task while the asset is in any other status.

Unsupported Fields

The edition-specific API currently does not pass through callback_url, priority, safety_identifier, service_tier, execution_expires_after, frames, camera_fixed, watermark, or draft. Do not depend on these fields in requests. If they are exposed later, this documentation will state it explicitly.

Media And Scenario Limits

  • Text prompt: Chinese prompts should be within 500 characters, and English prompts within 1000 words.
  • Image input: jpeg, png, webp, bmp, tiff, gif, heic, and heif are supported. Each image must be under 30 MB, with aspect ratio (0.4, 2.5) and width/height (300px, 6000px).
  • Image counts: first-frame mode uses 1 image; first-and-last-frame mode uses 2 images with first_frame and last_frame; multi-modal reference mode uses 1-9 reference_image items.
  • Video input: mp4 and mov are supported. Each video must be [2, 15] seconds, up to 3 reference videos, with total video duration no more than 15 seconds; each video must be under 50 MB, with FPS [24, 60].
  • Audio input: wav and mp3 are supported. Each audio item must be [2, 15] seconds, up to 3 reference audio items, with total audio duration no more than 15 seconds; each audio item must be under 15 MB.
  • Scenario exclusivity: first-frame, first-and-last-frame, and multi-modal reference scenarios cannot be mixed. Audio cannot be provided alone; include at least one reference image or video.
  • This system does not support image/audio base64 input. Use a public URL or an assetId returned by the asset upload API.

Response Parameters

idstring
Video task ID used to query task status and result.
objectstring
Object type, fixed as video.
createdinteger
Task creation timestamp.
modelstring
Model used by this request.
statusstring
Task status, usually queued or processing after creation.

Status

  • queued: the task has been created and is waiting to be processed.
  • processing: the task is being processed.
  • completed: the task succeeded and video_url is available from the query API.
  • failed: the task failed and error is available from the query API.

Request Body Example

Real-person reference
{
  "model": "sd_2.0_fast_special_720p",
  "ratio": "16:9",
  "duration": 4,
  "return_last_frame": true,
  "generate_audio": true,
  "content": [
    {
      "type": "text",
      "text": "A natural talking-head commercial, the person looks at the camera with a confident expression"
    },
    {
      "type": "image_url",
      "role": "reference_image",
      "image_url": {
        "url": "assetId://asset-20260722164336-p4mms"
      }
    }
  ]
}
Multi-modal reference
{
  "model": "sd_2.0_fast_special_720p_with_video_ref",
  "ratio": "9:16",
  "duration": 5,
  "return_last_frame": true,
  "generate_audio": true,
  "tools": [
    {
      "type": "web_search"
    }
  ],
  "content": [
    {
      "type": "text",
      "text": "A first-person fruit tea commercial using video 1 as the composition reference and audio 1 as background music"
    },
    {
      "type": "audio_url",
      "role": "reference_audio",
      "audio_url": {
        "url": "https://example.com/reference.mp3"
      }
    },
    {
      "type": "image_url",
      "role": "reference_image",
      "image_url": {
        "url": "https://example.com/reference.jpg"
      }
    },
    {
      "type": "video_url",
      "role": "reference_video",
      "video_url": {
        "url": "https://example.com/reference.mp4"
      }
    }
  ]
}