Create a Seedance 2.5 video (native format)

Supports text, image, video, and audio reference inputs and directly returns the native Volcengine Ark task ID.

To use the platform's unified task structure, see the platform-format create API.

To preview at low resolution before producing the final video, see the Draft video generation guide. Draft videos use 480p; final videos generated from a Draft use 1080p.

POST/api/v3/contents/generations/tasks

Create an asynchronous Seedance 2.5 video task through the native Volcengine Ark path. The core request parameters are the same as the compatible path, and a successful request returns a native-format task ID.

Request content, asset limits, task-type constraints, and output-format rules are the same as the compatible create API.

Authorization

bearerAuth

AuthorizationBearer <token>

Use Bearer your-api-key in the Authorization Header.

In: header

Request Body

application/json

Seedance 2.5 native video generation parameters.

TypeScript Definitions

Use the request body type in TypeScript.

model*string

Seedance 2.5 model ID, such as doubao-seedance-2-5-260628. For actual requests, use a model visible to the current API Key.

Length1 <= length

content*array<|>

Input items: text, images, video, audio, or a Draft task ID. Omni reference accepts up to 50 assets: 30 images, 10 videos, and 10 audio clips. Each video or audio clip must be 2–30 seconds; the total duration of each media type must not exceed 30 seconds. Editing requires a 4–30-second reference video. Audio-only reference is supported. First-frame and first-and-last-frame modes cannot be mixed with omni reference.

Reference assets in array order using @image1, @video1, and @audio1.

Items1 <= items

draft?boolean

Whether to generate a Draft preview. When true, use 480p. After success, reference content.draft_task.id within seven days to generate a 1080p final video. Omit this field or set it to false for final generation.

omni_reference_task_type?string

Omni reference task type. auto detects the task; reference generates from references; edit requires reference video, ratio=adaptive, and duration=-1; extend requires reference video and ratio=adaptive. An explicit type validates its constraints earlier. The prompt intent must still match the selected type.

Default"auto"

Value in

"auto"
"reference"
"edit"
"extend"

generate_audio?boolean

Whether to generate audio that matches the visuals. Enabling this option usually increases generation time and billing cost.

ratio?string

Video aspect ratio. Text-to-video and multimodal reference-to-video tasks can specify a ratio or use adaptive; first-frame, first-and-last-frame, video editing, and video extension tasks support only adaptive.

Default"adaptive"

Value in

"16:9"
"4:3"
"1:1"
"3:4"
"9:16"
"21:9"
"adaptive"

duration?|integer

Generation duration in seconds. Use a value from 4 to 30, or -1 to let the model decide automatically. Video editing tasks support only -1; video extension supports 4 to 30 or -1.

Default-1

resolution?string

Output video resolution. Supported values are 480p, 720p, and 1080p. Actual pixel dimensions depend on the resolution, aspect ratio, and current model capabilities.

Default"720p"

Value in

"480p"
"720p"
"1080p"

output_format?string

Output video format. mp4 is suitable for web and general distribution. mov uses H.264, YUV444P, and PCM and is suitable for post-production such as color grading, keying, and compositing.

Default"mp4"

Value in

"mp4"
"mov"

watermark?boolean

Whether to add a watermark. Availability depends on model permissions and service configuration.

seed?integer

Random seed from -1 to 4294967295. Using the same seed may improve reproducibility but does not guarantee identical results.

Range-1 <= value <= 4294967295

return_last_frame?boolean

Whether to return a watermark-free JPEG last frame for generating consecutive videos.

callback_url?string

Callback URL used after the task completes. If omitted, poll the retrieve API for task status.

Formaturi

safety_identifier?string

Stable unique identifier for the end user, used to help identify users who may violate usage policies. Send a hashed username, user ID, or email address and avoid submitting personal information.

Length1 <= length <= 64

execution_expires_after?integer

Task lifetime in seconds, measured from creation. A task that is still incomplete after this period enters the expired state.

Formatint32

Range3600 <= value <= 259200

Default172800

tools?array<>

Tools the model may call, such as web_search. Available tools and their parameters depend on the current account permissions.

service_tier?string

Service tier. Seedance 2.5 uses default online inference and does not support flex. Set queue priority with the separate priority parameter.

Default"default"

priority?integer

Queue priority within the same inference endpoint, from 0 to 9. Higher values take precedence; equal priorities follow submission order. Running tasks are not interrupted.

Range0 <= value <= 9

Default0

[key: string]?any

Response Body

`application/json`

curl -X POST "https://example.com/api/v3/contents/generations/tasks" \  -H "Content-Type: application/json" \  -d '{    "model": "MODEL_ID",    "content": [      {        "type": "text",        "text": "A small white rabbit rides a paper airplane through the clouds, with cinematic lighting and a smooth forward camera movement"      }    ],    "generate_audio": true,    "ratio": "16:9",    "duration": 10,    "resolution": "1080p",    "output_format": "mp4"  }'

{
  "id": "task_xxxxx"
}

Official Volcengine Ark documentation for creating video generation tasksReview the upstream protocol definition and the latest field changes for Seedance video generation tasks.

Create a Seedance 2.5 video (native format)

Authorization

Request Body

Response Body

200application/json

400

401

402

429

500

502

`application/json`