Submit Task (Doubao / Seedance Format)

Doubao video models (the Seedance series, e.g. seedance-1-0-pro-250528, doubao-seedance-2-0-260128, doubao-seedance-2-0-mini-260615 — see the Model Plaza) reuse the generic task format's endpoint — no separate integration needed (for the request structure overview see Create Video Generation Task). This page documents the Doubao channel's parameter mapping and channel differences; for querying see Query & Download.

Endpoint

Creation and query reuse the generic task format endpoints: POST /v1/video/generations and GET /v1/video/generations/{task_id}.

Request Body

The gateway converts the request into the upstream's {model, content[], ...parameters} structure and forwards it: prompt becomes the text item in the upstream content array; image / images become image items. All other generation parameters are passed through metadata and merged into the upstream request body by field name (unrecognized keys are dropped).

ParameterLocationTypeNotes
prompttopstringRequired, video description
modeltopstringSeedance model name; the gateway routes by model
image / imagestopstring / string[]First-frame / reference image for image-to-video (URL or Base64); converted into upstream content image items automatically
secondstopstringDuration (seconds). ⚠️ The Doubao channel reads this field; a numeric top-level duration has no effect
metadata.durationmetadatanumberDuration (seconds), equivalent to top-level seconds; seconds wins when both are present
metadata.resolutionmetadatastringResolution tier: 480p / 720p / 1080p (per model support). Billing ratio is computed from this tier
metadata.ratiometadatastringAspect ratio: 16:9 / 9:16 / 1:1 / 4:3 / 3:4 (per model support). ⚠️ Top-level size has no effect
metadata.seedmetadataintegerRandom seed
metadata.generate_audiometadatabooleanGenerate audio (Seedance 2.0 series); the output carries an audio track
metadata.watermarkmetadatabooleanWatermark switch
metadata.return_last_framemetadatabooleanReturn the last-frame image; available at data.data.content.last_frame_url on success
metadata.camera_fixedmetadatabooleanLock the camera: true disables camera movement (fixed shot), default false. Verified to work on t2v — true / false produce visibly different camera movement
metadata.callback_urlmetadatastringUpstream status callback URL (publicly reachable, accepts POST). Pushed once per status change (queued / running / succeeded…) with the upstream task object as the JSON body — use status to detect the terminal state
metadata.contentmetadataarrayAdvanced: a full content item array; may contain video_url / audio_url items for video reference / continuation (text items are stripped by the gateway, prompt is appended separately)
metadata.frames / metadata.service_tiermetadata—❌ Not supported by the upstream for text-to-video (t2v); submission returns 400 (the error carries the upstream Request id). Text-to-video frame rate is fixed at 24fps. Whether image-to-video supports them is per the upstream documentation
WARNING

Top-level duration, size, mode, width, height, fps, seed, n, response_format, user are silently dropped on the Doubao channel (no error, no effect): use seconds for duration, and metadata.ratio + metadata.resolution for the frame. Verified by test: the same request with top-level duration: 8 + seconds: "5" produced a 5-second video; top-level seed: 999 + metadata.seed: 42 echoed 42.

NOTE

The upstream validates parameters per "model × mode": unsupported combinations return 400 at submission (the message carries the model name, the mode, and a Request id) — just remove the offending parameter and retry. Tested (2026-09):

  • doubao-seedance-2-0-260128 t2v: resolution / ratio / seed / duration / generate_audio / return_last_frame / watermark / camera_fixed all take effect (1080p output and audio track verified; camera_fixed re-tested 2026-10 — t2v submission returns 200 and true / false show clearly different camera movement); frames / service_tier are not supported
  • doubao-seedance-2-0-mini-260615 i2v: ratio / generate_audio take effect (720p output with audio track); resolution is not supported — when omitted the output is fixed at 720p
  • Text-to-video frame rate is fixed at 24fps across models; other "model × mode" combinations are untested — defer to upstream errors
WARNING

Image-to-video (i2v) image / images do not accept photos of real people: The upstream compliance check rejects them outright with 400 InputImageSensitiveContentDetected.PrivacyInformation (no task created, no billing). Use illustration / anime / product / scenery images for the first frame.

Request Examples

cURL
Python
Node.js
# Text-to-video
curl -X POST https://api-model.shuntianchuang.cn/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{
    "model": "doubao-seedance-2-0-260128",
    "prompt": "An orange cat chasing butterflies on a lawn, cinematic shot, slow zoom in",
    "seconds": "5",
    "metadata": {
      "resolution": "1080p",
      "ratio": "16:9",
      "seed": 42,
      "generate_audio": true,
      "return_last_frame": true
    }
  }'
# Image-to-video: image takes the first-frame image URL or Base64
curl -X POST https://api-model.shuntianchuang.cn/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{
    "model": "doubao-seedance-2-0-260128",
    "prompt": "Make the cat in the frame run, camera slowly zooming in",
    "image": "https://example.com/first-frame.jpg",
    "seconds": "5",
    "metadata": {
      "resolution": "1080p",
      "ratio": "16:9",
      "generate_audio": true
    }
  }'

Response Body

A successful submission returns a task object; the task_id is used for querying, and status starts as queued.

Response Example

{
  "id": "task_xxxxxxxx",
  "object": "video",
  "task_id": "task_xxxxxxxx",
  "status": "queued",
  "progress": 0,
  "model": "doubao-seedance-2-0-260128",
  "created_at": 1769473000
}

For status query and download see Query & Download.