Seedance 2.0

Seedance 2.0 is a video generation model from ByteDance's Seed team, built around joint audio and video generation and supporting creative input from text, images, video, and audio. Users can describe a scene directly or provide character images, movement clips, camera examples, or sound material, explaining in the prompt how each reference should influence the result. This approach suits short videos that need to preserve brand visuals, reproduce a filming style, or develop connected storylines. Official materials emphasize stability in complex movements and interactions between multiple people, responsiveness to generation and editing instructions, and coordination of sound and visuals across multiple shots. Dialogue, ambient sound, and background music can be incorporated into the same creative process. The text-to-video route also accepts image, video, and audio references, while image-to-video uses a starting image to control the opening, allowing users to choose a route based on their available material. For greater control, describe the subject, setting, movement, camera, and audio requirements separately and avoid conflicting references. Review detail consistency and sound quality manually before deciding whether to use the result in finished content.

• Multimodal References
• Joint Audio and Video Generation