Extend continues a video shot from an existing timeline boundary while preserving the original media. The generated continuation is placed on a separate generated track for comparison and safe removal.

Seedance 2.5 is the default forward-extension model. It uses 1.8-30 seconds of rendered timeline video as context and creates a 4-30 second continuation with optional native audio and lip-synced dialogue. Resolution supports Match Canvas, 480p, 720p, and 1080p; Advanced contains Bitrate and optional Seed. Its price includes both input context time and output duration.

LTX 2.3 remains available when extending before the reference is required. Kling models remain available for boundary-frame continuation. Review each model header for its exact limits and the visible credit estimate before generation starts.