What Dreamactor V2 is good at — and what it's not

Dreamactor V2 is ByteDance's motion-driven character model: provide a character image and a reference video, and it drives the image with the reference's full performance — pose, motion, and expression transferred onto your character. Unlike a lip-sync model, which only moves the mouth, Dreamactor V2 reenacts the whole performance, which makes it the tool for animating a character image into a dancer, a presenter, or an actor following a recorded clip.

The inputs are simple — a character still and a driving video — with an optional prompt to steer the result. Pricing is about 8 credits per second of generated performance.

Where it sits: it's closely related to Kling's motion-control tiers (Kling V2.6 and V3 Pro), which also transfer motion from a reference video; the choice comes down to which results suit your character. For driving only the mouth from audio, Kling Avatar V2 is the lip-sync tool. For text-only or general video, Kling V3 and Seedance V2.0. Dreamactor V2's lane is full-performance character animation from a reference video.

A reasonable mental model: default to Dreamactor V2 when you want a character image to perform the full motion of a reference video, and use a lip-sync model when only the mouth needs to move.