Back to models
Bytedance
Video modelNewSeedance 2.5
Seedance 2.5 is a joint audio and video generation model from ByteDance's Seed team for producing more complete video works. It focuses on longer narratives, precise use of references, and coordination between sound and visuals. Creators can plan characters, settings, pacing, and audio, then add images, video, or sound material to guide visual style and performance. Official materials describe longer clips with sound in a single generation and support for extending video, along with a more detailed understanding of reference video's intent, composition, and cinematic language. This makes it suitable for brand stories, narrative shorts, promotional videos, and content that communicates a message across multiple shots. If you already have a storyboard or finished footage, explain the reference's role in the prompt, such as borrowing camera movement, character actions, or the sound atmosphere, instead of asking for an exact copy. Image-to-video can include an optional ending image to help plan the closing. Review character continuity, audio-visual pacing, and brand details before final delivery.
- Extended Storytelling
- Precise Reference Guidance
Key features
Longer clips with audio
Supports 4–30-second output with optional simultaneous sound generation.
Image, video, and audio references
Use separate materials to guide subject appearance, camera movement, and sound atmosphere.
Camera and visual control
Describe camera movement, lighting, and character actions through prompts; image-to-video also accepts an ending image.
