Seedance 3.0 generates video from text, images, video clips, and audio references. Text-to-video creates scenes with motion, lighting, and sound, while image-to-video animates photos or drawings and can use an end frame. Multi-reference inputs help preserve people, products, styling, and settings. The system creates dialogue with lip sync, music, ambient audio, and synchronized effects. Multi-shot prompts can include a beginning, middle, and ending while maintaining cast consistency. Users can extend clips, follow motion and camera references, or provide a simple 3D clay model for composition control. Local editing changes selected outfits, characters, products, backgrounds, or timestamps while preserving the rest. Outputs support 4K generation, multiple aspect ratios, downloadable MP4 files, and private asset storage.