Create a Seedance video - CometAPI Documentation

POST

videos

curl https://api.cometapi.com/v1/videos \
  -H "Authorization: Bearer $COMETAPI_KEY" \
  -F 'prompt="A slow cinematic camera push across a coastal landscape at sunrise"' \
  -F 'model="doubao-seedance-2-0"' \
  -F 'seconds="4"' \
  -F 'size="16:9"'

{
  "id": "task_abc123",
  "task_id": "task_abc123",
  "object": "video",
  "model": "doubao-seedance-2-0",
  "status": "queued",
  "progress": 0,
  "created_at": 1776681149
}

The size field accepts exact WxH values and ratio preset labels. Ratio preset labels map to the model-specific default WxH values in the table below. If you need a specific 720p or 1080p output, pass the exact WxH value instead of only a ratio preset. Exact WxH support remains model-dependent, so use a documented size entry for the target model when the rendered resolution must stay fixed.

Use reference images

Only the Seedance 2.0 models accept input_reference. Add one input_reference file for image-guided generation. For advanced multi-reference prompts, repeat the same multipart field in upload order. Seedance 2.0 is sensitive to how the prompt connects each reference image to the scene. Refer to the uploaded files by order, such as [Image 1], [Image 2], and [Image 3], then assign each image a clear role:

Use [Image 1] for the main subject, character, or product.
Use [Image 2] for a secondary subject, prop, or companion.
Use [Image 3] for the background, setting, lighting, or style.

Describe what should stay recognizable, what can change, the action, camera motion, visual style, and scene. A vague prompt can make the generated video look like the reference image was not used, even when the upload was accepted. This prompt structure is more reliable than only listing images:

Use the uploaded images in order: keep the boy wearing glasses and a blue T-shirt from [Image 1], add the corgi puppy from [Image 2], and use the lawn from [Image 3] as the setting. Create a 4-second 3D cartoon video with the boy and puppy sitting together, a slow camera push-in, soft daylight, and no extra characters.

Duration by model

If you omit seconds, CometAPI requests a 5-second clip. Send seconds as a string when you need a specific duration.

Model family	Supported `seconds`	Default	Boundary behavior
Seedance 2.0 and Seedance 2.0 Fast	`4`-`15`	`5`	Use integer seconds in range.
Seedance 1.5 Pro	`4`-`12`	`5`	Use integer seconds in range.
Seedance 1.0 Pro	`2`-`10`	`5`	Use integer seconds in range.

Size support by model

The table below follows the official Seedance resolution mapping and keeps the published size values in one place. For 1080p, the rightmost column lists the documented Seedance 1.5 Pro and Seedance 2.0 values. Seedance 2.0 Fast is not listed in that official 1080p table.

Resolution	Aspect ratio	Pixel values for Seedance 1.0 series	Pixel values for Seedance 1.5 Pro / Seedance 2.0 / Seedance 2.0 Fast
`480p`	`16:9`	`864x480`	`864x496`
	`4:3`	`736x544`	`752x560`
	`1:1`	`640x640`	`640x640`
	`3:4`	`544x736`	`560x752`
	`9:16`	`480x864`	`496x864`
	`21:9`	`960x416`	`992x432`
`720p`	`16:9`	`1248x704`	`1280x720`
	`4:3`	`1120x832`	`1112x834`
	`1:1`	`960x960`	`960x960`
	`3:4`	`832x1120`	`834x1112`
	`9:16`	`704x1248`	`720x1280`
	`21:9`	`1504x640`	`1470x630`
`1080p`	`16:9`	`1920x1088`	`1920x1080`
	`4:3`	`1664x1248`	`1664x1248`
	`1:1`	`1440x1440`	`1440x1440`
	`3:4`	`1248x1664`	`1248x1664`
	`9:16`	`1088x1920`	`1080x1920`
	`21:9`	`2176x928`	`2206x946`

How ratio presets differ from exact WxH

Use a ratio preset when the model-specific default output size is acceptable. For the doubao-seedance-1-5-pro family, ratio-only requests use the 720p row from the table above:

Ratio preset	Default rendered size
`16:9`	`1280x720`
`4:3`	`1112x834`
`1:1`	`960x960`
`3:4`	`834x1112`
`9:16`	`720x1280`
`21:9`	`1470x630`

Use exact WxH when the rendered resolution must stay fixed. Exact WxH remains model-dependent, so inspect the completed media before relying on an undocumented size.

Authorizations

Authorization

string

header

required

Bearer token authentication. Use your CometAPI key.

Body

multipart/form-data

prompt

string

required

Text prompt that describes the video. Required. When using reference images, describe what each uploaded image should control, such as the subject from [Image 1], the secondary object from [Image 2], and the background or style from [Image 3]. Also state the action, camera motion, visual style, and scene.

Example:

"A slow cinematic camera push across a coastal landscape at sunrise."

model

enum<string>

required

Seedance model id. Only the two 2.0 models accept input_reference.

Available options:

doubao-seedance-2-0,

doubao-seedance-2-0-fast,

doubao-seedance-1-5-pro,

doubao-seedance-1-0-pro

Example:

"doubao-seedance-2-0"

seconds

integer

default:5

Video duration in seconds. The accepted range depends on the model: doubao-seedance-2-0 and doubao-seedance-2-0-fast accept 4 to 15, doubao-seedance-1-5-pro accepts 4 to 12, and doubao-seedance-1-0-pro accepts 2 to 10. The default is 5 for every model.

Required range: 2 <= x <= 15

Example:

5

size

string

Output size. Use either a ratio preset label such as 16:9, 4:3, 1:1, 3:4, 9:16, or 21:9, or an exact WxH value such as 1280x720. Ratio preset labels map to model-specific default render sizes. If you need a fixed 720p or 1080p output, pass the exact WxH value instead of only a ratio preset. Exact WxH support remains model-dependent, and an undocumented or unsupported raw WxH value can normalize to another size or fail.

Pattern: ^(16:9|4:3|1:1|3:4|9:16|21:9|[1-9]\d{2,3}x[1-9]\d{2,3})$

Examples:

"16:9"

"1280x720"

"1920x1080"

"4:3"

"1112x834"

"1:1"

"960x960"

"3:4"

"834x1112"

"9:16"

"720x1280"

"1080x1920"

"21:9"

"1470x630"

input_reference

file

Optional reference image uploaded as a multipart file. Repeat this field to send multiple reference images; the order of repeated fields is the order you should reference in the prompt as [Image 1], [Image 2], and so on. Use JPEG or PNG files for best compatibility. Only doubao-seedance-2-0 and doubao-seedance-2-0-fast accept this field; sending it with a 1.0 Pro or 1.5 Pro model returns HTTP 400.

Response

Task created. Save the returned id and poll GET /v1/videos/{id}.

string

required

Task id. Use it as the path parameter for GET /v1/videos/{id}.

object

string

required

Object type, always video.

model

string

required

Echo of the requested model id.

status

enum<string>

required

Initial task status. Newly created tasks are returned as queued.

Available options:

queued,

in_progress,

completed,

failed,

error

progress

integer

required

Completion percentage. 0 at creation.

Required range: 0 <= x <= 100

created_at

integer

required

Task creation time as a Unix timestamp in seconds.

task_id

string

Alias of id returned for compatibility. The value matches id.

Which video models can I use with CometAPI?

Retrieve a Seedance video

curl https://api.cometapi.com/v1/videos \
  -H "Authorization: Bearer $COMETAPI_KEY" \
  -F 'prompt="A slow cinematic camera push across a coastal landscape at sunrise"' \
  -F 'model="doubao-seedance-2-0"' \
  -F 'seconds="4"' \
  -F 'size="16:9"'

{
  "id": "task_abc123",
  "task_id": "task_abc123",
  "object": "video",
  "model": "doubao-seedance-2-0",
  "status": "queued",
  "progress": 0,
  "created_at": 1776681149
}

​Use reference images

​Duration by model

​Size support by model

​How ratio presets differ from exact WxH

Authorizations

Body

Response

Use reference images

Duration by model

Size support by model

How ratio presets differ from exact WxH