content array, and returns the task id.POST https://studio.dataeyes.ai/seedance/api/v3/contents/generations/tasks{ "id": "cgt-20261111150751-sxe7w" }id to poll Retrieve a task. Task IDs are valid for 7 days (draft tasks: 14 days).| Parameter | Type | Required | Description |
|---|---|---|---|
model | string | Yes | Model name; see Overview · Models |
content | object[] | Yes | Input content; see "content" below |
resolution | string | No | 480p, 720p (default), 1080p. 2.0 fast / mini do not support 1080p |
ratio | string | No | 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive (default; chosen from the input) |
duration | integer | No | Duration in seconds. 2.5: 4–30; 2.0 series: 4–15; -1 lets the model choose (2.5 defaults to -1) |
generate_audio | boolean | No | Generate synchronized audio (voices, sound effects, music); default true. false outputs a silent video |
watermark | boolean | No | Add an "AI generated" watermark at the bottom right; default false |
return_last_frame | boolean | No | Return the last frame as an image (content.last_frame_url); default false. Useful as the first frame of the next clip |
draft | boolean | No | 2.5 only. Draft mode: generate a 480p draft to check the result, then generate the final video; see "Draft mode" below |
omni_reference_task_type | string | No | 2.5 only. Type of a multimodal reference task: auto (default), reference (generate from references), edit (edit a video), extend (extend a video). When set, parameters are validated at submission |
output_format | string | No | 2.5 only. mp4 (default) or mov |
tools | object[] | No | Tools, for example [{"type": "web_search"}] to enable web search |
priority | integer | No | Queue priority 0–9, default 0; higher runs first |
callback_url | string | No | Receives a POST whenever the task status changes, with the same body as the retrieve endpoint |
execution_expires_after | integer | No | Task timeout in seconds, 3600–259200, default 172800 (48 hours). The status becomes expired after timeout |
safety_identifier | string | No | End-user identifier (hashing recommended), up to 64 characters |
duration: -1, the actual duration is the duration returned by the retrieve endpoint.content is an array of objects distinguished by type. role sits alongside image_url / video_url / audio_url, not inside them.type | Format | role |
|---|---|---|
text | {"type": "text", "text": "prompt"} | — |
image_url | {"type": "image_url", "image_url": {"url": "..."}, "role": "first_frame"} | first_frame, last_frame, reference_image |
video_url | {"type": "video_url", "video_url": {"url": "..."}, "role": "reference_video"} | Always reference_video |
audio_url | {"type": "audio_url", "audio_url": {"url": "..."}, "role": "reference_audio"} | Always reference_audio |
draft_task | {"type": "draft_task", "draft_task": {"id": "draft task ID"}} | — |
url accepts a public URL, Base64 (data:image/png;base64,..., data:audio/wav;base64,...; not supported for video), or an asset ID (asset://...; see the Private Asset Library).| Mode | content |
|---|---|
| Image to video, first frame | Text (optional) + 1 first_frame image |
| Image to video, first and last frames | Text (optional) + 1 first_frame + 1 last_frame image |
| Multimodal reference | Text (optional) + any combination of reference images, videos, and audio. The 2.0 series needs at least 1 image or video; 2.5 accepts audio only |
ratio must be adaptive).| Input | Requirements | Count (2.5) | Count (2.0 series) |
|---|---|---|---|
| Image | jpeg, png, webp, bmp, tiff, gif, heic, heif; aspect ratio 0.4–2.5; width and height 300–6000 px; under 30 MB each | 1–30 reference images | 1–9 reference images |
| Video | mp4, mov (H.264 / H.265); aspect ratio 0.4–2.5; up to 200 MB each; 24–60 fps | Up to 10; 2–30 s each; 30 s total | Up to 3; 2–15 s each; 15 s total |
| Audio | wav, mp3; up to 15 MB each | Up to 10; 2–30 s each; 30 s total | Up to 3; 2–15 s each; 15 s total |
480p only):model and draft_task. Some parameters are carried over from the draft and cannot be passed again, such as resolution (passing it returns 400). Draft task IDs are valid for 14 days.| Error (excerpt) | Cause |
|---|---|
the parameter duration specified in the request is not valid | duration is outside the series' range |
InvalidParameter.TaskTypeConstraint ... only supported in the multimodal reference scenario | omni_reference_task_type cannot be set for text-to-video, first-frame, or first-and-last-frame tasks |
omni_reference_task_type ... allowed values: auto, reference, edit, extend | Invalid omni_reference_task_type value |
InputImageSensitiveContentDetected.PrivacyInformation | The input image may contain a real human face |
status is then failed, with the reason in error of the retrieve response.