Skip to main content
Doubao Seedance 2.5 Flash is ByteDance’s faster variant of Seedance 2.5. It supports text-to-video, first-frame and first/last-frame generation, multimodal references, video editing, and video extension. POST https://api.mixroute.ai/v1/video/generations Keep model, prompt, and asset at the top level. Put the provider fields in metadata. For text-only requests, use asset: false and include a text item in metadata.content that matches prompt. Use asset: true when sending reference media.
Flash does not support Draft. Omit metadata.draft and do not include a draft_task item. For Draft previews, use the standard Doubao Seedance 2.5 route.

Parameters and limits

Input modes

  • For frame conditioning, use first_frame or first_frame and last_frame. Set metadata.ratio to adaptive. Do not mix frame roles with reference_* roles.
  • For multimodal input, use reference_image, reference_video, or reference_audio. Per request: up to 10 images, 5 video clips, and 5 audio clips. Audio-only references are supported. The total duration of reference videos and the total duration of reference audio must each be at most 30 seconds.
  • For video editing, include a reference_video of 4–30 seconds, set metadata.ratio to adaptive, and set metadata.duration to -1. Video extension also requires a reference_video and adaptive ratio.

Request examples

Set MIXROUTE_API_KEY to your MixRoute API key. For the editing example, replace the example URL with an accessible source video.

Text-to-video

Edit a video

Query the task

Replace MIXROUTE_TASK_ID with the task ID returned by submission. Wait for success before retrieving the video URL. See Query Video Task for statuses and response fields.
See the Doubao API for shared parameters and Seedance media limits for input formats and sizes. For current pricing, see the Model Marketplace.