You need to enable JavaScript to run this app.
Lake AI Service

Lake AI Service

Copy page
Download PDF
Video generation
Enhanced and basic versions of video generation
Copy page
Download PDF
Enhanced and basic versions of video generation

The LAS enhanced/basic version video generation service is built around the Seedance series of models, integrating pre- and post-processing for video generation, and adding capabilities such as video format conversion, video segmentation, video restoration, subtitle generation, audio-video merging, and video super-resolution. These features extend the capability boundaries of video generation models, improve the quality and duration of generated videos, reduce token consumption costs, and help customers increase production efficiency.

Operator introduction

Application scenarios

As a general-purpose large video generation model, Seedance can be flexibly applied to a wide range of video generation scenarios. For example:

  • Marketing and promotional videos: product promotion, brand advertising shorts, event or holiday marketing content. Supports direct generation from text or images, suitable for high-frequency, batch material production.
  • Short video content production: Douyin/Video Account/Bilibili content, MCN batch production, rapid content creation for individual creators. Provides generation, editing, and extension capabilities, with an emphasis on style, camera movement, and character stability control.
  • Image-to-video / photo animation: converting static posters to dynamic content, dynamic interpretation of product or character images, animating old photos or illustrations. Supports text-to-video and image-to-video; some versions support control of the first and last frames.
  • Video editing and secondary creation: re-editing, video extension, stylization, and multi-material splicing. The 2.0 series explicitly supports video editing and extension, not just generation from scratch.
  • Education and knowledge expression: course explanation shorts, animated picture books or science popularization, teaching demonstration videos. Visualizes abstract knowledge or script content to improve communication efficiency.
  • Digital human / character consistency content: virtual anchors, virtual idols, digital humans for enterprise training. Ensures stable character features, object detail, and style restoration capabilities.
  • E-commerce and livestreaming materials: product display videos, virtual scene demonstrations, livestream room warm-up or clips. Enables large-scale, rapid, and low-cost material production, highly suited for e-commerce scenarios.

Effect demonstration


[Multimodal reference]
  • Enter prompt

    Based on reference images and reference videos, generate an advertisement video for LAS tea drink with Peking Opera makeup. The opening of the video basically matches reference image 1. In the middle section, the character walks to a desk, sees a pot of wine and a cup of LAS tea drink (reference image 2), shows slight hesitation, then chooses the LAS tea drink and drinks a sip with a straw. Close-up: after drinking, the character's previously dazed eyes instantly become bright, with a smile at the corner of their mouth, and says the line: "LAS tea drink, a sip of Beijing flavor. The story intoxicates, but you awaken." Then, at the end of the video, the character holds LAS in their left hand, raises their right hand's water sleeve slightly, and strikes the classic 'Please enjoy' pose. "

  • Reference image 1
    Image
  • Reference image 2
    Image
  • Generate video

Edit video – convert video style
  • Enter prompt

    Convert the reference video into a Chinese ink animation style video, changing only the video style to Chinese ink animation, including the character, environment, props, and camera movement, while keeping all other aspects unchanged.

  • Reference video
  • Generate video
Edit video – convert environment in video
  • Enter prompt

    Replace the environment in the reference video with the environment from the reference image, keeping the character, props, and camera movement unchanged and smooth.

  • Reference image
    Image
  • Reference video
  • Generate video
>> LAS provides you with Enhanced video editing, Audio and video merging, and other operators, offering more convenient and stable editing for replacement, audio merging, and other capabilities.

Extend video
  • Enter prompt

    The arched window in video 1 opens, and the camera enters the interior of the art gallery, then continues with video 2. Afterwards, the camera moves into the painting, then continues with video 3.

  • Reference video 1
  • Reference video 2
  • Reference video 3
  • Generate video

Supported models

Model series
Supported models

Seedance 2.x series models

  • Video generation - basic version:
    • doubao-seedance-2-0-fast-260128
    • doubao-seedance-2-0-mini-260615
  • Video generation - enhanced version:
    • doubao-seedance-2-0-260128

Model capabilities

Function
Subfeature / parameter

Seedance 2.0

Seedance 2.0 Fast

Seedance 2.0 Mini

Text-to-video

"content.tpye": "text"

Image-to-video

Image-to-video - first frame

Image-to-video - first and last frames

Multimodal reference

Image reference

Video reference

  • Combined reference
    • Image + audio
    • Image + video
    • Video + audio
    • Image + video + audio

Edit video

Not applicable

Extend video

Not applicable

Generate video with audio

"generate_audio": "true"

Supported regions

  • Beijing: cn-beijing
  • Shanghai: cn-shanghai

Operator performance

Sub-item
Performance impact description

Maximum RPM

  • Non-4k resolution: 600
  • 4k resolution: 15

Maximum concurrency

  • Non-4k resolution: 10
  • 4k resolution: 1

Input and output requirements

Input requirements

Subcategory

Detailed requirements

Input data modality

  • Text
  • Image
  • Video
  • Audio cannot be input independently; at least one reference video or image must be included.

The following combinations are supported:

  • Text
  • Text + image
  • Text + video
  • Text + image + audio
  • Text + image + video
  • Text + video + audio
  • Text + image + video + audio

tip

When using different Seedance models, the supported input data modalities may vary. For details, refer to the "Model capabilities" section above.

General requirements

Seedance 2.0 series models do not support direct upload of reference images or videos containing real human faces. To facilitate creators' use of portraits, you can apply to enable the LAS material library feature (allowlist feature), upload authorized real-person reference images or videos to the material library, and input them to the model using the material library ID.

Input format: image

  • Formats: jpeg, png, webp, bmp, tiff, gif. Among them, Seedance 1.5 Pro and Seedance 2.0 series models additionally support heic and heif.
  • Aspect ratio (width/height): [0.4, 2.5]
  • Width and height (px): [300, 6000]
  • Size: Each image must be less than 30 MB. Request body size must not exceed 64 MB.
  • Number of images:
    • Image-to-video - first frame: 1 image
    • Image-to-video - first and last frames: 2 images
    • Seedance 2.0 series multimodal reference video generation: 1–9 images

Input format: video

  • Format: mp4, mov.
  • Resolution: 480p, 720p, 1080p, 4k
  • Duration: Each video must be between 2 and 15 seconds. Up to 3 reference videos can be uploaded, and the total duration of all videos must not exceed 15 seconds.
  • Dimensions:
    • Aspect ratio (width/height): [0.4, 2.5]
    • Width and height (px): [300, 6000]
    • Total pixel count: [640×640=409600, 3326×2494=8295044], that is, the product of width and height must be within the range [409600, 8295044].
  • Size: Each video must not exceed 200 MB.
  • Frame rate (FPS): [24, 60]

Input format: audio

  • Format: wav, mp3
  • Duration: Each audio file must be between 2 and 15 seconds. Up to 3 reference audio files can be uploaded, and the total duration of all audio files must not exceed 15 seconds.
  • Size: Each audio file must not exceed 15 MB, and the request body must not exceed 64 MB.

Input path requirements

Provide input data to the operator via the content request parameter. The following input methods are supported for images, audio, and video:

  • Public URL: A publicly accessible video URL in the format http/https.
    • Public URLs that require login status or additional header authentication are not supported; temporary URLs must remain valid during task execution.
  • Base64 encoding (supported for audio/images): Convert the local file to a Base64-encoded string, then submit it to the large model. Follow the format: data:image/<image or audio format>;base64,<Base64 encoding>. Note that <image or audio format> must be lowercase, for example, data:image/png;base64,{base64_image}. Do not use Base64 encoding for large files.
  • Asset ID (allowlist feature): After enabling the LAS asset library management feature, you can upload assets to the LAS asset library and obtain the corresponding asset ID from the LAS asset library management page. When calling the video generation operator, you can use the asset ID as a reference input for video generation, following the format: asset://<ASSET_ID>.

Output requirements

Breakdown

Detailed requirements

Output data type

  • Video
  • Restriction: Currently, frame rate adjustment is not supported.

doubao-seedance-2-0-260128
(audio and video co-generation)

  • Resolution: 480p, 720p, 1080p, 4k (10-bit depth)
  • Frame rate: 24 fps
  • Duration: 4–15 seconds
  • Video format: mp4

doubao-seedance-2-0-fast-260128
(audio and video co-generation)

  • Resolution: 480p, 720p
  • Frame rate: 24 fps
  • Duration: 4–15 seconds
  • Video format: mp4

doubao-seedance-2-0-mini-260615
(audio and video co-generation)

  • Resolution: 480p, 720p
  • Frame rate: 24 fps
  • Duration: 4–15 seconds
  • Video format: mp4

Output path: API

After the operator completes processing, the pre-signed link for the generated video is returned directly through the video generation task query interface. The link is valid for 24 hours.

Billing information

tip

Estimated cost: Price calculator.

  • Billing standards

    Breakdown items
    Billing standards description

    Billing item

    Duration of input/output video.

    Billing type

    Usage-based billing, unit: CNY/second, billed hourly based on actual billing usage.

    Unit price

    Depends on the algorithm version you select (Basic Edition, Enhanced Edition).

    warning

    • Total cost paid: In addition to the unit price, it also depends on the duration of the input/output video and the resolution of the input/output video.
    • Only tasks that successfully generate videos will incur charges. Scenarios where video generation fails due to review or other operational reasons will not be charged.
  • Billing details
    Billing estimate: Total cost ≈ Unit price * Billing usage = Unit price * (Video duration * Duration conversion coefficient)

    warning

    The above formula is for cost estimation only. Actual billing is based on the bill.

    Version breakdown
    Unit price

    Model type

    Scenario breakdown
    Billing item: Video duration
    Duration conversion coefficient: Related to video resolution

    Basic Edition

    1.6 CNY/second

    seedance 2.0 fast

    No input video

    Output video duration

    • 480P:0.465
    • 720P:1

    With input video

    [ max(input video duration, output video duration*2/3) + output video duration ] /2

    • 480P:0.554
    • 720P:1.220

    seedance 2.0 mini

    No input video

    Output video duration

    • 480P:0.290625
    • 720P:0.625

    With input video

    [ max(input video duration, output video duration*2/3) + output video duration ] /2

    • 480P:0.176875
    • 720P:0.38125

    Enhanced Edition

    2 CNY/second

    seedance 2.0

    No input video

    Output video duration

    • 480P:0.465
    • 720P:1
    • 1080P:2.504
    • 4k:3.19

    With input video

    [ max(input video duration, output video duration*2/3) + output video duration ] /2

    • 480P:0.580
    • 720P:1.220
    • 1080P:3.040
    • 4k:5.057
  • Billing example

    • Example scenario: Using video generation - Enhanced Edition for video generation, the input includes a reference video (duration 5 seconds), the output video duration is 7 seconds, and the resolution is 1080p.
    • Cost details:
      Total cost ≈ Unit price * Billing usage = 2 CNY/second * [ max(5 seconds, 7 seconds*2/3) + 7 seconds ]/2 * 3.040 = 36.48 CNY

Notes and prerequisites

Details

Caution and prerequisites

Costs

Before calling an operator, you need to understand the model invocation costs associated with using the operator. For details, see Large model invocation billing.

Authentication (API Key)

Before calling an operator, you need to generate an API Key for operator invocation. It is recommended to configure the API Key as an environment variable to ensure safer operator calls. For details, see Obtain and configure API Key.

BaseURL

Before calling an operator, you need to determine the BaseURL for operator invocation based on the region where your current LAS service is deployed. This is used to configure the path parameter values for operator calls.
For details, see Obtain the Base URL. The Examples below are for reference only; when making actual calls, replace the path values with those corresponding to your region.

API call

Generate

POST https://operator.las.ap-southeast-1.volces.com/api/v1/contents/generations/tasks

API description

Use the seedance model to generate videos.

Request parameters

Parameter
Type
Required
Example value
Description
model
string
Yes
doubao-seedance-2-0-260128
The model ID you need to call. Supported:
  • doubao-seedance-2-0-260128
  • doubao-seedance-2-0-mini-260615
  • doubao-seedance-2-0-fast-260128
content
object[]
Yes
Content provided to the model for generation. Supports text, image, video, and audio.
callback_url
string
Optional
`http://xxx/callback`
Callback notification address for the result of this generation task. When the task status changes, ModelArk will send a POST request to this address. Callback statuses include:
  • queued: queued
  • running: running
  • succeeded: succeeded
  • failed: failed
  • expired: expired
return_last_frame
boolean
Optional
false
Whether to return the last frame image of the video. Default is false.
  • true: Returns the last frame image, which can be used to stitch consecutive videos
  • false: Does not return the last frame image
execution_expires_after
integer
Optional
172800
Task timeout threshold, in seconds. Default is 172800 seconds (48 hours), value range is 3600 to 259200.
generate_audio
boolean
Optional
true
Controls whether the generated video includes sound synchronized with the visuals. Default is true.
  • true: Outputs video with synchronized audio
  • false: Outputs silent video
resolution
string
Optional
720p
Video resolution. Optional values:
  • 480p
  • 720p: Seedance 2.x series defaults to 720p
  • 1080p
  • 4k
Among them:
  • 1080p: Not supported in reference image scenarios; Seedance 2.0 Fast and Seedance 2.0 Mini are not supported.
  • 4k: Supported only by Seedance 2.0.
ratio
string
Optional
16:9
Aspect ratio for the generated video. Supported:
  • 16:9
  • 4:3
  • 1:1
  • 3:4
  • 9:16
  • 21:9
  • adaptive: Automatically selects the most suitable aspect ratio based on the input.

adaptive adaptation rules:
  • Text-to-video: Intelligently selects the most suitable aspect ratio based on the input prompt.
  • First frame / first and last frame video generation: Automatically selects the closest aspect ratio based on the ratio of the uploaded first frame image.
  • Multimodal reference video generation: Determine based on the user's prompt intent. If it is first-frame video generation, video editing, or video extension, select the aspect ratio closest to that image or video; otherwise, select the aspect ratio closest to the first media file provided (priority: "video" is higher than "image").
duration
integer
Optional
5
Either duration or frames can be specified; frames has higher priority than duration. To generate a video with an integer number of seconds, it is recommended to specify duration. Supported range: 4 to 15 seconds.
watermark
boolean
Optional
false
Whether the generated video contains a watermark.
  • false: No watermark
  • true: Contains watermark

Response parameters

Parameter

Type

Note

id

string

Task ID, retained for 7 days only (calculated from the created at timestamp); after expiration, it will be automatically deleted.

Example

Request example

curl --location 'https://operator.las.ap-southeast-1.volces.com/api/v1/contents/generations/tasks' \
--header 'Authorization: $LAS_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
    "model": "doubao-seedance-2-0-260128",
    "content": [
        {
            "type": "text",
            "text": "请把@视频1的背景替换成@图片1"
        },
        {
            "type": "image_url",
            "image_url": {
                "url": "image_url"
            },
            "role": "reference_image"
        },
        {
            "type": "video_url",
            "video_url": {
                "url": "video_url"
            },
            "role": "reference_video"
        }
    ],
    "generate_audio":false,
    "seed":42,
    "watermark": false
}'

Response example

{
    "id": "lsd-89b8ea3232b8f1205702"
}

Task

GET https://operator.las.ap-southeast-1.volces.com/api/v1/contents/generations/tasks/{id}

API description

Query the status of the video generation task.

Request parameters

Parameter

Type

Required

Example value

Note

id

string

Yes

cgt-2025******-****

The ID of the video generation task you need to query.
This parameter is passed as a URL path segment.

Response parameters

Parameter
Type
Example value
Description
id
string
cgt-2025******-****
Video generation task ID
model
string
Model name and version used for the task, model name-version
status
string
succeeded
Model status and related information:
  • queued: In queue.
  • running: Task is running.
  • cancelled: Task cancelled. Cancelled status will be automatically deleted after 24 hours (only tasks in queue can be cancelled).
  • succeeded: Task succeeded.
  • failed: Task failed.
  • expired: Task expired.
error
error
Error message. Returns null if the task succeeds; returns error data if the task fails. For details on error messages, see .
code
string
Error code.
message
string
Error message.
created_at
integer
1764059744
Creation time
updated_at
integer
1764059744
Update time
content
content
Output content of the video generation task.
video_url
string
URL of the generated video, in mp4 format. To ensure information security, the generated video will be deleted after 24 hours; please save it promptly. It is recommended to configure the data subscription feature provided by Volcano Engine TOS to automatically save your model inference output to your own TOS bucket for long-term backup or further processing.
last_frame_url
string
URL of the last frame image of the video. Valid for 24 hours; please save it promptly.
Note: This parameter is returned when "return_last_frame": true is set during video generation task creation.
file_url
string
URL of the video generation result file (such as flv or other non-mp4 formats).
seed
integer
1345
Seed integer value used for this request.
resolution
string
1080p
Resolution of the generated video.
ratio
string
9:16
Aspect ratio of the generated video.
duration
integer
10
Duration of the generated video, unit: seconds.
Note: Only one of the parameters duration or frames will be returned. When frames is not specified when creating a video generation task, duration will be returned.
frames
integer
Number of frames in the generated video.
Note: Only one of the parameters duration or frames will be returned. When frames is specified when creating a video generation task, frames will be returned.
framespersecond
integer
30
Frame rate of the generated video.
generate_audio
boolean
false
Whether the generated video contains audio synchronized with the visuals.
  • true: The video output by the model contains synchronized audio.
  • false: The video output by the model is silent.
execution_expires_after
integer
Task timeout threshold, unit: seconds.
usage
token_usage
Video generation token usage
completion_tokens
integer
13987
Number of tokens in the output content.
total_tokens
integer
15000
Total number of tokens consumed by this request. The video generation model does not count input tokens; input tokens are 0, so total_tokens=completion_tokens.

Example

Request example

curl --location "https://operator.las.ap-southeast-1.volces.com/api/v1/contents/generations/tasks/lsd-89b8ea3232b8f1205702" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer $LAS_API_KEY"

Response example

{
  "id": "lsd-89b8ea3232b8f1205702",
  "model": "doubao-seedance-2-0-260128",
  "status": "succeeded",
  "content": {
    "video_url": " `https://ark-acg-cn-beijing.tos-cn-beijing.volces.com/doubao-seedance-2-0/02178419371805600000000000000000000f****21.mp4?X-Tos-Algorithm=*****d&X-Tos-SignedHeaders=host` "
  },
  "usage": {
    "completion_tokens": 648900,
    "total_tokens": 648900
  },
  "created_at": 1784193717,
  "updated_at": 1784194096,
  "seed": 84841,
  "resolution": "720p",
  "ratio": "16:9",
  "duration": 15,
  "framespersecond": 24,
  "service_tier": "default",
  "execution_expires_after": 172800,
  "generate_audio": true,
  "draft": false,
  "priority": 0
}

Error code

HttpCode

Error code

Error message

Note

401

ApiKey.Invalid

The api key is invalid.

API is invalid

Last updated: 2026.08.04 11:00:54