Skip to main content
Video models are asynchronous by default. They may support different combinations of:
  • Text input.
  • Image reference input.
  • Video reference input.
  • Audio reference input.
  • Duration choices.
  • Resolution choices.
  • Aspect ratio choices.
APIAny.AI validates public request parameters before execution, so unsupported model options fail clearly.

Available models

Seedance 2.5 video edit

Send seedance-2.5-video-edit to /v1/videos/generations with a non-empty editing prompt and video_urls. The first video is the edit target; additional videos provide references. Images can guide clothing colors, lighting and background styling, identified in the prompt as @image1. Media URLs must be directly accessible. Maximum file sizes are 200 MB per video, 30 MB per image, and 15 MB per audio clip. Media duration, codec, and other file requirements are checked during model execution.
Omitting duration uses the same fixed 30-second charge as specifying -1, including shorter source clips. The total charge does not change with the result’s actual duration. At the current standard 480p price with content_filter: true, this is 12,600 APIAny.AI credits. Check the live model catalog for current prices before submitting a task. For response-loss recovery, an edit request can include an optional X-Client-Request-Id header containing a fresh UUID v4. If the submission response is lost, query /v1/tasks/req_edit_<uuid> with the same account’s API key to find the original task. The lookup returns the real task_id when the task is available and is isolated to your account and this edit model. A 404 can mean the task has not yet been saved; check again later. This header provides correlation, not POST idempotency: do not resubmit an uncertain generation request or reuse its UUID for another edit.