Seedance Video

Seedance Video uses an asynchronous video task API. Create a task first, then poll the retrieve endpoint for status and the final video URL.

Recommended integration

For new integrations, use the unified /v1/video/generations endpoint with the official-style top-level content array for text, images, videos, and audio.

Which Endpoint To Use

Scenario	Recommended endpoint	Notes
New integration	`/v1/video/generations`	Unified video endpoint. It returns OpenAI-style video task objects and works well when you use multiple video models
Existing official Seedance / Ark client	`/api/v3/contents/generations/tasks`	Official-compatible path and task object shape for lower-friction migration
Existing simple image-to-video form	`/v1/video/generations` with simplified fields	You can send `prompt`, `image_url`, `images`, `video_url`, or `audio_url`; Cubicspaces converts them into official-style `content`

Endpoints

Capability	Method	Path
Create video task (official-compatible)	`POST`	`/api/v3/contents/generations/tasks`
Retrieve video task (official-compatible)	`GET`	`/api/v3/contents/generations/tasks/{task_id}`
Create video task	`POST`	`/v1/video/generations`
Retrieve video task	`GET`	`/v1/video/generations/{task_id}`
Moderate images	`POST`	`/v1/images/moderations`

Supported Models

Model	Notes
`doubao-seedance-2.0`	Seedance 2.0 standard model
`doubao-seedance-2.0-fast`	Seedance 2.0 fast model; availability depends on account permissions and platform configuration

Supported Modes

Mode	How to call
Text to video	Send text items in the top-level `content` array
Image to video	Send image items in `content`; use `role` for first frame, last frame, or reference images
Video reference	Send video items in `content` with `role: "reference_video"`
Audio reference	Send audio items in `content` with `role: "reference_audio"` together with image or video media
Image moderation	Submit public image URLs to `/v1/images/moderations`, then use the returned `asset://` URL in `content`

All asset://{...} values in this page are placeholders. In real requests, replace them with the full asset:// URL returned by the image moderation endpoint.

Top-Level Parameters

Parameter	Required	Description
`model`	Yes	Seedance model name, such as `doubao-seedance-2.0`
`content`	Recommended	Multimodal content array for prompts, media inputs, media types, and roles
`duration`	No	Duration in integer seconds. Default is usually `5`; Seedance 2.0 commonly supports `4` to `15`
`ratio`	No	Output aspect ratio. Supported values include `adaptive`, `21:9`, `16:9`, `4:3`, `1:1`, `3:4`, `9:16`
`resolution`	No	Output resolution. Common values are `480p`, `720p`, and `1080p`; `doubao-seedance-2.0-fast` does not support `1080p`
`generate_audio`	No	Whether to generate synchronized audio. Set `false` for silent video
`watermark`	No	Whether to add a watermark
`seed`	No	Random seed, useful when you want a more repeatable style
`callback_url`	No	Task callback URL. You can also poll the retrieve endpoint instead

Compatible Request Shapes

Cubicspaces supports two request shapes. For long-term integrations, prefer the official-style content array. Existing clients can keep using simplified fields.

Official-style content

content is an ordered array. Text, images, videos, and audio are all represented as content items. In prompts, refer to media by order, such as "image 1", "video 1", or "audio 1".

json

{
  "model": "doubao-seedance-2.0",
  "content": [
    {
      "type": "text",
      "text": "Use image 1 as the product reference and video 1 as the camera-motion reference. Generate a 5-second product showcase."
    },
    {
      "type": "image_url",
      "role": "reference_image",
      "image_url": {
        "url": "asset://{product_asset_id}"
      }
    },
    {
      "type": "video_url",
      "role": "reference_video",
      "duration": 5,
      "video_url": {
        "url": "https://example.com/camera-motion.mp4"
      }
    }
  ],
  "duration": 5,
  "resolution": "720p",
  "ratio": "16:9"
}

Simplified fields

Simplified fields are useful for ordinary form-based integrations. Cubicspaces converts them into official-style content and appends prompt as a text item.

Simplified field	Conversion
`prompt`	A `text` item in `content`
`image_url`	One `image_url` content item
`images`	Multiple `image_url` content items; either string arrays or `{ "url": "...", "role": "..." }` arrays are accepted
`video_url`	One `video_url` content item
`videos`	Multiple `video_url` content items
`audio_url`	One `audio_url` content item
`audios`	Multiple `audio_url` content items
`resolution_name`, `size`	Used as `resolution` when `resolution` is empty
`seconds`	Used as `duration` when greater than 0

json

{
  "model": "doubao-seedance-2.0",
  "prompt": "Generate a smooth first-and-last-frame transition from image 1 to image 2.",
  "images": [
    { "url": "asset://{first_frame_asset_id}", "role": "first_frame" },
    { "url": "asset://{last_frame_asset_id}", "role": "last_frame" }
  ],
  "duration": 8,
  "resolution": "720p",
  "ratio": "16:9"
}

metadata passthrough

If you need an official field that is not available as a unified top-level field, put it in metadata. Cubicspaces merges metadata into the upstream request body. metadata.model is ignored; the top-level model and platform model mapping remain authoritative.

json

{
  "model": "doubao-seedance-2.0",
  "prompt": "Generate a fixed-camera product showcase video.",
  "metadata": {
    "camera_fixed": true,
    "service_tier": "default",
    "return_last_frame": true
  },
  "duration": 5,
  "resolution": "720p"
}

Examples

Text-to-Video

bash

curl -X POST https://cubicspaces.cloud/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "doubao-seedance-2.0",
    "content": [
      {
        "type": "text",
        "text": "A cinematic aerial shot of a futuristic cubic city at sunrise"
      }
    ],
    "duration": 5,
    "resolution": "720p",
    "ratio": "16:9",
    "generate_audio": false,
    "watermark": false
  }'

Official-Compatible Endpoint

If your client already uses the official task API shape, call the official-compatible path directly. After creation succeeds, use the response id to retrieve the task.

bash

curl -X POST https://cubicspaces.cloud/api/v3/contents/generations/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "doubao-seedance-2.0",
    "content": [
      {
        "type": "text",
        "text": "A cinematic aerial shot of a futuristic cubic city at sunrise"
      }
    ],
    "duration": 5,
    "resolution": "720p",
    "ratio": "16:9",
    "generate_audio": false
  }'

Image-to-Video

bash

curl -X POST https://cubicspaces.cloud/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "doubao-seedance-2.0",
    "content": [
      {
        "type": "text",
        "text": "Use image 1 as the product appearance reference. Generate a clean 5-second studio showcase while preserving the subject."
      },
      {
        "type": "image_url",
        "role": "reference_image",
        "image_url": {
          "url": "asset://{image_asset_id}"
        }
      }
    ],
    "duration": 5,
    "resolution": "720p",
    "ratio": "1:1",
    "watermark": false
  }'

First And Last Frame

json

{
  "model": "doubao-seedance-2.0",
  "content": [
    {
      "type": "text",
      "text": "Generate a smooth transition video from image 1 to image 2."
    },
    {
      "type": "image_url",
      "role": "first_frame",
      "image_url": {
        "url": "asset://{first_frame_asset_id}"
      }
    },
    {
      "type": "image_url",
      "role": "last_frame",
      "image_url": {
        "url": "asset://{last_frame_asset_id}"
      }
    }
  ],
  "duration": 8,
  "resolution": "720p",
  "ratio": "16:9"
}

Video Reference Input

json

{
  "model": "doubao-seedance-2.0",
  "content": [
    {
      "type": "text",
      "text": "Use video 1 as the camera-motion and pacing reference, then generate a new scene in the same style."
    },
    {
      "type": "video_url",
      "role": "reference_video",
      "video_url": {
        "url": "https://example.com/reference.mp4"
      }
    }
  ],
  "duration": 5,
  "resolution": "720p",
  "ratio": "16:9"
}

Image and Audio Reference

json

{
  "model": "doubao-seedance-2.0",
  "content": [
    {
      "type": "text",
      "text": "Use image 1 and audio 1 to create a product showcase video."
    },
    {
      "type": "image_url",
      "role": "reference_image",
      "image_url": {
        "url": "asset://{image_asset_id}"
      }
    },
    {
      "type": "audio_url",
      "role": "reference_audio",
      "audio_url": {
        "url": "https://example.com/reference.wav"
      }
    }
  ],
  "duration": 5,
  "ratio": "1:1",
  "generate_audio": true
}

content Items

Field	Required	Description
`content[].type`	Yes	`text`, `image_url`, `video_url`, or `audio_url`
`content[].text`	Conditional	Used when `type` is `text`
`content[].image_url.url`	Conditional	Public image URL, or an approved image `asset://<asset ID>`
`content[].video_url.url`	Conditional	Public video URL. Cubicspaces checks that the URL is downloadable
`content[].audio_url.url`	Conditional	Public audio URL, or an upstream-supported audio asset URL
`content[].role`	Conditional	Media role; see the combination rules below
`content[].duration`	No	Reference duration for this media item, commonly used for video or audio references; unit is seconds

Media Combination Rules

Scenario	Recommended input	role requirements
Text-to-video	Text-only `content`	No media role needed
Single image-to-video	One `image_url` content item	`first_frame`, or omit the role
First-and-last-frame video	Two `image_url` content items	First image is `first_frame`; last image is `last_frame`
Multimodal reference video	Send reference images/videos/audios in `content`	Images use `reference_image`, videos use `reference_video`, audio uses `reference_audio`

Limits and guidance:

At most 9 reference images. First-and-last-frame requests should contain exactly 2 images.
At most 3 reference videos. Each video can be up to 15 seconds, and the total reference-video duration must not exceed 15 seconds.
At most 3 reference audio files. Each audio file can be up to 15 seconds, and the total reference-audio duration must not exceed 15 seconds.
Audio cannot be the only media input; provide at least one image or video as well.
Do not mix first-frame / first-and-last-frame roles with reference_image, reference_video, or reference_audio.
In prompts, refer to media by order, such as "image 1", "video 1", or "audio 1". Do not write Asset IDs directly in the prompt.

Media File Limits

Media	Limits
Image	Common formats include `jpeg`, `jpg`, `png`, `webp`, `bmp`, `tiff`, and `gif`; each image should be below 30 MB; recommended aspect ratio is between `0.4` and `2.5`; recommended width and height are between `300` and `6000` px
Video	`mp4` or `mov`; recommended resolution is `480p` or `720p`; the standard model may use `1080p` when supported by the account and model, while the fast model does not support `1080p`; each video should be below 50 MB; recommended frame rate is `4` to `60` FPS; the public URL must download video bytes directly, not an HTML login page
Audio	`mp3` or `wav`; each audio file should be below 15 MB

The current public API does not support base64 or inline binary media. Upload media to a publicly reachable URL first. For images, you can also use the moderation endpoint and pass the returned asset_url.

Moderate Images

When Seedance uses real-person or library-managed image assets, submit public image URLs to the image moderation endpoint first. After approval and ingestion, use the returned items[].asset_url as content[].image_url.url.

Submit all images used by the same Seedance generation request together in one image moderation request under the images array. Do not split images from the same generation request across multiple moderation calls; separate moderation batches can lead to inconsistent asset review batches or asset bindings.

http

POST /v1/images/moderations
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

Request Parameters

Parameter	Required	Description
`model`	Yes	Used to select Seedance capability. Use `doubao-seedance-2.0`
`images`	Yes	Image URL array; all images used by the same generation request must be submitted in the same batch
`asset_type`	No	Asset type, defaults to `Image`; keep the default for image moderation

Images must be public http or https URLs. Base64 and inline binary content are not supported.

Request Example

bash

curl -X POST https://cubicspaces.cloud/v1/images/moderations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "doubao-seedance-2.0",
    "images": [
      "https://example.com/person.png",
      "https://example.com/product.png"
    ]
  }'

Response Example

json

{
  "code": "success",
  "message": "",
  "data": {
    "object": "asset_moderation",
    "status": "approved",
    "review_batch_id": "review-batch-id",
    "task_id": "moderation-task-id",
    "items": [
      {
        "source_url": "https://example.com/person.png",
        "asset_url": "asset://{person_asset_id}",
        "submit_review_status": 1,
        "passed": true
      },
      {
        "source_url": "https://example.com/product.png",
        "asset_url": "asset://{product_asset_id}",
        "submit_review_status": 1,
        "passed": true
      }
    ]
  }
}

When passed is true and submit_review_status is 1, the image is approved.

Response and Retrieval

A successful create request returns a standard video task object:

json

{
  "id": "video_xxx",
  "task_id": "video_xxx",
  "object": "video",
  "model": "doubao-seedance-2.0",
  "status": "queued",
  "progress": 0,
  "created_at": 1770000000
}

Retrieve the task:

bash

curl https://cubicspaces.cloud/v1/video/generations/video_xxx \
  -H "Authorization: Bearer YOUR_API_KEY"

If you submitted the task through the official-compatible path, you can also use the response id to retrieve it through the official-compatible path:

bash

curl https://cubicspaces.cloud/api/v3/contents/generations/tasks/video_xxx \
  -H "Authorization: Bearer YOUR_API_KEY"

When the task is completed, read the video URL from data.result_url. If the upstream provider returns token usage, the response also includes data.usage.completion_tokens and data.usage.total_tokens.

Status Values

Status	Meaning
`queued`	Queued
`running`	Generating
`succeeded`	Completed
`failed`	Failed; inspect `error.message`

Notes

Use ratio: "adaptive" when you are unsure, or omit ratio and let upstream defaults apply.
Video reference URLs must be directly downloadable by the platform server. Do not pass URLs that require login, return HTML pages, or block server-side fetching.
Image references can use public URLs or asset://<asset ID> returned by the image moderation endpoint.
Actual model, resolution, and tool availability can vary by account permissions and platform configuration.

Seedance Video ​

Which Endpoint To Use ​

Endpoints ​

Supported Models ​

Supported Modes ​

Top-Level Parameters ​

Compatible Request Shapes ​

Official-style content ​

Simplified fields ​

metadata passthrough ​

Examples ​

Text-to-Video ​

Official-Compatible Endpoint ​

Image-to-Video ​

First And Last Frame ​

Video Reference Input ​

Image and Audio Reference ​

content Items ​

Media Combination Rules ​

Media File Limits ​

Moderate Images ​

Request Parameters ​

Request Example ​

Response Example ​

Response and Retrieval ​

Status Values ​

Notes ​

Seedance Video

Which Endpoint To Use

Endpoints

Supported Models

Supported Modes

Top-Level Parameters

Compatible Request Shapes

Official-style content

Simplified fields

metadata passthrough

Examples

Text-to-Video

Official-Compatible Endpoint

Image-to-Video

First And Last Frame

Video Reference Input

Image and Audio Reference

content Items

Media Combination Rules

Media File Limits

Moderate Images

Request Parameters

Request Example

Response Example

Response and Retrieval

Status Values

Notes