Seedance 2.5
seedance-2.5-text-to-videoCommercial useSeedance 2.5 is a multimodal video model for prompt-only generation or combined image, video, and audio references. It creates 480P, 720P, or 1080P clips up to 30 seconds with synchronized audio, adaptive aspect ratios, and optional last-frame output through the unified 97AI.PRO asynchronous API.
prompt is required. Optionally add reference_image_urls, reference_video_urls, or reference_audio_urls (videos/audios are always direct URLs). Or use first_frame_url + last_frame_url instead — but not together with reference_* fields. Playground starts empty by default.
First-frame image URL or asset://{assetId}. Mutually exclusive with reference_* fields.
Last-frame image URL or asset://{assetId}. Requires first_frame_url.
Portrait-library route: ingest hint for first_frame_url. Default skip (direct URL). Use virtual/real to upload portraits to the asset library.
Portrait-library route: ingest hint for last_frame_url. Default skip (direct URL). Use virtual/real to upload portraits to the asset library.
Text prompt describing the video to generate (max 30000 characters).
Reference image URLs (max 30). Mutually exclusive with first/last frame fields.
Reference video URLs (max 10; total duration ≤30s).
Reference audio URLs (max 10; total duration ≤30s).
Generate synchronized audio with the video.
Return the generated video's last frame. Cannot be true when draft=true.
Output video resolution.
Output aspect ratio.
Video duration from 4 to 30 seconds, or -1 for automatic selection.
Video output format.
Enable online search for the generation.
Enable content safety checking.