Seedance 2.5 is an integrated AI model for joint audio-video generation. It creates long narrative clips and accepts up to 50 multimodal references including images, videos and audio to stabilize characters, scenes and camera movements. It supports timestamp-based local editing without full re-rendering. Co-processed audio and visuals deliver smoother motion and synchronized audiovisual effects for commercials, short dramas and industrial creative production.