ID-V2V: Identity-Preserving Video Restylization

Researchers introduced ID-V2V, a video restylization framework that changes scene, lighting, and style while preserving a person’s identity.

It keeps facial expressions, eye gaze, and lip sync consistent across the video by using keyframe-based control signals.

The method avoids paired training data by reframing identity preservation as a relighting problem.

ID-V2V could make AI video editing more practical for film, media production, and other human-centered workflows.