Documentation

Video Generation

The Playground supports asynchronous video tasks with MiniMax, Doubao, the Doubao Extended Protocol, and Alibaba Cloud video models. Board selects the matching protocol, options, and price estimate from the chosen model.

Create a video

  1. Open Playground and choose Video.
  2. Select an available video model and API key.
  3. Enter a prompt and optionally upload reference images or, for Doubao multi_ref_vid, MP4/MOV reference videos.
  4. Set the duration, ratio, resolution, and seed.
  5. Review the estimate and click Create.

MiniMax, Doubao, and Alibaba prompts can contain up to 7,000 characters; the Doubao Extended Protocol accepts up to 20,000. Doubao supports text-to-video, first-frame, first/last-frame, image-reference, and mixed image/video-reference modes. After submission, completed videos can be previewed or downloaded.

Parameters

ParameterMiniMaxDoubaoDoubao Extended ProtocolAlibaba
Resolution768PModel-dependent: 480p, 720p, 1080p, 4k768P480P, 720P, 1080P
Duration4-15 seconds4-15 seconds4-15 secondsModel-dependent, 2-30 seconds
Ratio16:9, 4:3, 1:116:9, 4:3, 3:4, 9:16, 1:1, 21:9, adaptive16:9, 9:16, 1:116:9, 9:16, 3:4, 4:3, 4:5, 5:4, 1:1
Images and videosUp to 1 imageUp to 4 images; up to 4 MP4/MOV videos in multi_ref_vidUp to 4 imagesText, first-frame, or up to 10 reference images, depending on model
Seed42 by default; editable42 by default; editable42 by default; editable42 by default; editable
AudioDepends on the modelGenerated by defaultAlways generated nativelyEnabled by default; metadata.generate_audio accepts an override

Doubao shows the resolutions supported by the selected model. Seedance mini and fast default to 480p; the standard model defaults to 720p. The price estimate reports unsupported model and resolution combinations before submission. Switching models refreshes the estimate.

Image input

Supported image formats are JPEG, PNG, WebP, HEIC, and HEIF, with a maximum size of 30 MB per image. Doubao reference videos accept MP4/MOV up to 64 MB each. MiniMax accepts one image; Doubao and the Doubao Extended Protocol accept up to four images; Alibaba supports text, first-frame, and model-dependent reference-image modes. Board sends media through the selected protocol and does not expose the originals as public URLs.

Price and balance

The estimate follows the selected protocol's billing unit and configured platform pricing. Native Doubao Seedance reserves an estimated Token amount and settles against actual output tokens after completion, returning or charging the difference. Other video protocols currently reserve duration-based amounts. A task is not created when the balance is insufficient.

If a video task fails or is cancelled, the platform settles the refund according to the task state. Usage records can appear with a short delay while asynchronous settlement is synchronized.

Task states

StatusMeaning
QueuedThe request was accepted and is waiting upstream processing
RunningThe upstream service is generating the video
SucceededThe video is stored by Board and can be previewed or downloaded
FailedUpstream generation or result transfer failed; see the task details
CancelledThe task was cancelled and settled according to the billing state

Video generation is asynchronous. After the upstream reports completion, Board still transfers the result into private storage. A task can therefore remain Running briefly after upstream generation has finished.

Protocol differences

The four protocols use the same Playground workflow with different limits. Board selects the protocol from the chosen model; users do not enter /minimax, /doubao, /marslab, or /aliyun paths manually. Playground exposes ordinary parameters; advanced vela fields for the Doubao Extended Protocol are API-only.

If a model is missing from the selector, it may not advertise video support, may not be available to the current organization, or its upstream may be unavailable.