Create Video Generation Task

Version: v1.0.0 | Last Updated: 2026-06-11
This platform fully supports the Doubao Seedance series video generation API. Requests and responses are transparently proxied, and all parameter semantics remain consistent with the official API.

Submit a video generation task. The task is processed asynchronously; a task id is returned immediately. Use the Query Task endpoint to poll for completion.

POST https://platform.dataeyes.ai/seedance/api/v3/contents/generations/tasks

Model Capabilities

Capability	2.0 / 2.0 Lite / 2.0 Fast	1.5 Pro	1.0 Pro / 1.0 Lite
Text-to-Video (T2V)	Yes	Yes	Yes
I2V (First Frame)	Yes	Yes	Yes
I2V (First + Last Frame)	No	Yes	Yes
Reference Video	Yes	No	No
Reference Audio	Yes	No	No
Reference Image	Yes	No	No
Audio Generation (`generate_audio`)	Yes	Yes	No
Draft Mode (`draft`)	No	Yes	No
Web Search (`tools`)	Yes	No	No
Priority Queue (`priority`)	Yes	No	No
Flex Service Tier	No	No	No
1080p Resolution	Yes (except 2.0 Fast)	Yes	Yes

Supported Content Combinations

Generation Mode	Required Content Types	Notes
Text-to-Video	`text`	Text prompt only
I2V (First Frame)	`text` + `image_url` (role: `first_frame`)	Single reference image as first frame
I2V (First + Last Frame)	`text` + `image_url` (role: `first_frame`) + `image_url` (role: `last_frame`)	1.5 Pro and 1.0 series only
Reference Video	`text` + `video_url` (role: `reference_video`)	2.0 series only
Reference Audio	`text` + `audio_url` (role: `reference_audio`)	2.0 series only
Reference Image	`text` + `image_url` (role: `reference_image`)	2.0 series only
Draft Upgrade	`draft_task`	1.5 Pro only; upgrades a completed draft task

Request Body Parameters

Top-Level Parameters

Parameter	Type	Required	Default	Description
`model`	string	Yes	-	Model Endpoint ID. See Model Overview.
`content`	object[]	Yes	-	Array of content objects defining the generation input. See Content Objects. Note: Seedance 2.0 models restrict face-related content in reference images; consult safety guidelines for details.
`callback_url`	string	No	-	Webhook URL for task completion notifications. When the task reaches a terminal status (`succeeded`, `failed`, `cancelled`, `expired`), the platform sends an HTTP POST request to this URL with the full task result as the JSON body. The callback is best-effort; implement idempotent handling on your server.
`return_last_frame`	boolean	No	`false`	When `true`, the completed task response includes a `last_frame_url` field containing the last frame of the generated video as an image URL. Useful for chaining I2V tasks or for thumbnail generation.
`service_tier`	string	No	`"default"`	Service tier for task execution. Values: `"default"`, `"flex"`. Seedance 2.0 models only support `"default"`. The `"flex"` tier offers lower cost with potentially longer queue times.
`execution_expires_after`	integer	No	`172800`	Maximum time (in seconds) a task may remain in the queue before being automatically expired. Range: [3600, 259200] (1 hour to 3 days). Default: 172800 (2 days). If the task has not started processing within this window, its status transitions to `expired`.
`generate_audio`	boolean	No	`true`	Generate synchronized audio with the video output. Supported by Seedance 2.0 and 1.5 Pro models only. When enabled, the output video contains a single audio channel. Set to `false` to produce silent video.
`draft`	boolean	No	`false`	Generate a fast draft-quality preview. Seedance 1.5 Pro only. Draft videos are rendered at 480p only, do not support last-frame return, and cannot use the offline/flex service tier. Draft tasks can later be upgraded to full quality by submitting a new task with a `draft_task` content object.
`tools`	object[]	No	-	Enable auxiliary tools for the generation process. Seedance 2.0 only. Currently supports web search: `[{"type": "web_search"}]`. When enabled, the model may search the web for reference material to improve generation quality. Tool usage is reported in `usage.tool_usage`.
`safety_identifier`	string	No	-	Safety policy identifier for content moderation. Used to apply specific safety filtering rules to the generation request.
`priority`	integer	No	-	Task priority within the queue. Seedance 2.0 only. Range: [0, 9], where `0` is highest priority and `9` is lowest. Tasks with the same priority are processed in FIFO (first-in, first-out) order. Tasks with higher priority (lower number) are dequeued before tasks with lower priority, regardless of submission time.

Generation Parameters

These parameters control the output video characteristics. They can be specified either as top-level request body fields (recommended) or embedded within a --params JSON block appended to the text prompt.

Parameter Upgrade Note
The recommended approach is to pass generation parameters as top-level fields in the request body (the "new way"). The legacy approach of embedding parameters in the text prompt via a --params {...} suffix (the "old way") is still supported for backward compatibility but is discouraged for new integrations.
New way (recommended):
{
  "model": "seedance-2.0",
  "content": [{"type": "text", "text": "A cat walking on the beach"}],
  "resolution": "720p",
  "ratio": "16:9",
  "duration": 5
}
Old way (legacy, still supported):
{
  "model": "seedance-2.0",
  "content": [{"type": "text", "text": "A cat walking on the beach --params {\"resolution\":\"720p\",\"ratio\":\"16:9\",\"duration\":5}"}]
}
When both are present, the top-level body parameters take precedence.

Parameter	Type	Required	Default	Description
`resolution`	string	No	Varies by model	Output video resolution. Values: `"480p"`, `"720p"`, `"1080p"`. Default is `"720p"` for most models. Seedance 2.0 Fast does not support `"1080p"`. See Resolution & Pixel Dimensions for exact pixel sizes.
`ratio`	string	No	`"16:9"`	Output video aspect ratio. Values: `"16:9"`, `"4:3"`, `"1:1"`, `"3:4"`, `"9:16"`, `"21:9"`, `"adaptive"`. The `"adaptive"` ratio automatically matches the aspect ratio of the input reference image. Adaptive rules: only available when a reference image is provided; the system selects the closest standard ratio to the input image's aspect ratio. If no reference image is present, `"adaptive"` falls back to `"16:9"`.
`duration`	integer	No	`5`	Output video duration in seconds. Supported ranges vary by model: Seedance 2.0 series supports `[5, 10]`; Seedance 1.5 Pro supports `[5, 10]`; Seedance 1.0 series supports `[5, 10]`. Set to `-1` for automatic duration selection by the model. Mutually exclusive with `frames`.
`frames`	integer	No	-	Output video frame count. Not supported by Seedance 2.0 or 1.5 models. Range: [29, 289]. The value must follow the formula `25 + 4n` (where n is a non-negative integer), i.e., valid values are 29, 33, 37, 41, ..., 285, 289. Mutually exclusive with `duration`.
`seed`	integer	No	Random	Random seed for reproducible generation. Range: [-1, 2^32 - 1]. Use `-1` or omit for random seed selection. The actual seed used is returned in the task query response.
`camera_fixed`	boolean	No	`false`	Lock the camera position during video generation. When `true`, the camera remains stationary throughout the video. Not supported by Seedance 2.0 models.
`watermark`	boolean	No	-	Add a watermark to the output video. When `true`, a platform watermark is overlaid on the generated video.

Content Objects

The content array defines the input materials for the generation task. Each object has a type field that determines its structure.

Important: The role field is specified at the content object level (i.e., content[].role), NOT nested inside image_url, video_url, or audio_url.

Text Content

Field	Type	Required	Description
`type`	string	Yes	Must be `"text"`.
`text`	string	Yes	The text prompt describing the desired video. Maximum length varies by model.

{
  "type": "text",
  "text": "A golden retriever running through a sunflower field at sunset, cinematic lighting"
}

Image Content

Field	Type	Required	Description
`type`	string	Yes	Must be `"image_url"`.
`image_url.url`	string	Yes	URL of the reference image. Supports HTTPS URLs and base64 data URIs (`data:image/...;base64,...`).
`role`	string	Yes	Role of this image in the generation. Values: `"first_frame"`, `"last_frame"`, `"reference_image"`. `first_frame`: use as the opening frame of the video. `last_frame`: use as the closing frame (1.5 Pro and 1.0 series only). `reference_image`: use as a style/subject reference (2.0 series only).

{
  "type": "image_url",
  "image_url": {
    "url": "https://example.com/reference.jpg"
  },
  "role": "first_frame"
}

Video Content

Field	Type	Required	Description
`type`	string	Yes	Must be `"video_url"`.
`video_url.url`	string	Yes	URL of the reference video. Supports HTTPS URLs and base64 data URIs.
`role`	string	Yes	Must be `"reference_video"`. Seedance 2.0 series only.

{
  "type": "video_url",
  "video_url": {
    "url": "https://example.com/reference.mp4"
  },
  "role": "reference_video"
}

Audio Content

Field	Type	Required	Description
`type`	string	Yes	Must be `"audio_url"`.
`audio_url.url`	string	Yes	URL of the reference audio. Supports HTTPS URLs and base64 data URIs.
`role`	string	Yes	Must be `"reference_audio"`. Seedance 2.0 series only.

{
  "type": "audio_url",
  "audio_url": {
    "url": "https://example.com/reference.mp3"
  },
  "role": "reference_audio"
}

Draft Task Content

Field	Type	Required	Description
`type`	string	Yes	Must be `"draft_task"`.
`draft_task.id`	string	Yes	The task ID of a previously completed draft task. Seedance 1.5 Pro only. The referenced draft task must have `status: "succeeded"` and `draft: true`.

{
  "type": "draft_task",
  "draft_task": {
    "id": "draft_task_abc123"
  }
}

Input Requirements

Image Requirements

Property	Requirement
Formats	JPEG, PNG, WebP
Maximum file size	10 MB
Minimum resolution	128 x 128 pixels
Maximum resolution	4096 x 4096 pixels
Color space	RGB (CMYK images may produce unexpected results)

Video Requirements

Property	Requirement
Formats	MP4, MOV, WebM
Maximum file size	100 MB
Maximum duration	60 seconds
Minimum resolution	128 x 128 pixels
Maximum resolution	4096 x 4096 pixels
Codecs	H.264, H.265, VP9

Audio Requirements

Property	Requirement
Formats	MP3, WAV, AAC, FLAC
Maximum file size	20 MB
Maximum duration	60 seconds
Sample rate	16 kHz - 48 kHz recommended

Resolution & Pixel Dimensions

The exact output pixel dimensions depend on the model series, resolution, and aspect ratio. The table below lists the actual width x height (in pixels) for each combination.

Resolution	Ratio	Pixels (1.0 Series)	Pixels (1.5 Pro / 2.0 Series)
480p	16:9	864 x 480	864 x 496
480p	4:3	736 x 544	752 x 560
480p	1:1	640 x 640	640 x 640
480p	3:4	544 x 736	560 x 752
480p	9:16	480 x 864	496 x 864
480p	21:9	960 x 416	992 x 432
720p	16:9	1248 x 704	1280 x 720
720p	4:3	1120 x 832	1112 x 834
720p	1:1	960 x 960	960 x 960
720p	3:4	832 x 1120	834 x 1112
720p	9:16	704 x 1248	720 x 1280
720p	21:9	1504 x 640	1470 x 630
1080p	16:9	1920 x 1088	1920 x 1080
1080p	4:3	1664 x 1248	1664 x 1248
1080p	1:1	1440 x 1440	1440 x 1440
1080p	3:4	1248 x 1664	1248 x 1664
1080p	9:16	1088 x 1920	1080 x 1920
1080p	21:9	2176 x 928	2206 x 946

Note: Pixel dimensions may not exactly match standard broadcast resolutions due to model architecture constraints (e.g., divisibility requirements). The values above are the actual output dimensions produced by each model series.

Request Examples

Text-to-Video (T2V)

Image-to-Video (First Frame)

Image-to-Video (First + Last Frame)

Response

A successful request returns a JSON object containing only the task ID. The full task result (including video URL, usage, and metadata) is available via the Query Task endpoint.

{
  "id": "task_abc123def456"
}

Field	Type	Description
`id`	string	Unique task identifier. Valid for 7 days from creation. Use this ID to query task status, retrieve results, or cancel/delete the task.

After receiving the task ID, poll the Query Task endpoint until the task reaches a terminal status (succeeded, failed, cancelled, or expired). See Overview - Asynchronous Workflow for the recommended polling pattern.

01-Create-Video-Generation-Task

Create Video Generation Task#

Model Capabilities#

Supported Content Combinations#

Request Body Parameters#

Top-Level Parameters#

Generation Parameters#

Content Objects#

Text Content#

Image Content#

Video Content#

Audio Content#

Draft Task Content#

Input Requirements#

Image Requirements#

Video Requirements#

Audio Requirements#

Resolution & Pixel Dimensions#

Request Examples#

Text-to-Video (T2V)#

Image-to-Video (First Frame)#

Image-to-Video (First + Last Frame)#

Response#