Seedance

Seedance is ByteDance's AI video-generation model, currently at version 2.5. It is a video model rather than a chatbot: it natively generates narrative videos with multiple cohesive shots, keeping the main subject, visual style and atmosphere consistent. Seedance 2.5 generates synchronised audio in the same pass as the visuals, supports up to around 50 references (images, video and audio) to steer lighting, camera and performance, and can produce longer clips (up to about 30 seconds) with region-level editing. It supports multilingual creation and clean 4K output. It is available through ByteDance's Dreamina (CapCut) platform. Exact capabilities and limits change over time, so confirm current details on the official website.

Seedance generates video from text prompts and references. You can supply up to around 50 image, video and audio references to steer characters, lighting, camera and performance, and the model produces a multi-shot narrative clip with synchronised audio generated in the same pass. Region-level editing lets you change parts of a clip without regenerating everything, and longer clips (up to about 30 seconds) keep the subject and style consistent.

• Cinematic storytelling

  Produce multi-shot narrative clips with consistent characters and synchronised sound in one generation.

• Product ads

  Use reference images to keep a product on-brand across a short commercial clip.

• Music and performance video

  Generate clips with lip-synced subjects and audio references for music-style content.

• Longer narrative sequences

  Extend stories up to around 30 seconds with stable continuity across shots.

• Creators
• Filmmakers
• Marketers
• Agencies

• Multi-shot narrative video generation
• Synchronised audio generated with the video
• Up to ~50 references (image, video, audio)
• Reference-to-video (R2V) control
• Region-level editing
• Longer clips up to ~30 seconds
• Character and style consistency
• Multilingual creation
• Up to 4K output

Platforms: Web · API: Not available