Seedance 2.0

About Seedance 2.0

Seedance 2.0 is ByteDance's text- and image-to-video model. It can cut a single generation into several shots and hold the same character, wardrobe and lighting across them, which is why so many of the clips here look edited rather than generated. It also takes reference images for characters, products and environments, and it follows explicit camera language (dolly, orbit, handheld, rack focus) more reliably than most models in its class.

The prompts on this page skew toward short cinematic sequences: action beats, fashion and product films, trailer-style openings, and walk-and-talk shots where the subject moves through a space. Prompts that describe motion in physical terms (weight, speed, what the camera does when) tend to hold up better than prompts that only describe a mood. Single static portraits are underusing the model; the strong examples give it something to move.

Prompting tips

• Structure prompts as a shot list: for each shot give subject, action, camera move, and lighting in that order.
• Attach reference images for anything that must stay consistent (face, outfit, product); describing it in text alone drifts.
• State one camera move per shot. Stacking several moves in one sentence usually produces a wobble instead of a choice.