Seedance 2.5 is an AI model for joint audio-video generation. It creates long narrative videos and accepts abundant image, video and audio references to maintain consistent characters, scenes and camera movements. Timestamp-based targeted local editing avoids full re-rendering. Its audiovisual joint architecture improves motion smoothness and synchronization for commercial video production.