Netflix Eyeline research model that restyles video from an edited keyframe while preserving face identity and performance.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
ID-V2V is a video-to-video generative framework from Eyeline Labs (Netflix) for identity-preserving video restylization. Given a source video and an edited keyframe, it propagates the scene, lighting and style changes across the clip while keeping the subject's facial likeness and performance, including expressions, eye gaze and lip synchronization.
The method separates edit-driven synthesis from source-grounded identity preservation. Relit facial regions and facial normal maps constrain the face, while the edited keyframe and depth sequences guide the rest of the scene. Training pairs are built from a single video by relighting face regions, which avoids the need for scarce paired data. The project page says it supports single- and multi-subject scenarios, tolerates imperfect keyframes (the output re-aligns to the source from the second frame), and handles relighting as a special case.
The project page is labelled "for demonstration and inspiration purposes only". A YouTube creator describes it as open-source with output up to 720p and a main model of about 80 GB, but no code, weights or paper links were found in the provided material, so release status is unverified.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Open-source video editing AI that changes the style of a scene (background, lighting, clothing) from one edited keyframe while preserving character identity and motion. Outputs up to 720p; the main model is about 80 GB.
Closest neighbours by meaning, not just by category.
Model
Research model for video relighting that jointly generates relit video and per-frame environment maps, no camera poses needed.
Model · Viggle
Geometry-guided video-to-video model from Viggle that re-renders existing footage along a new camera path and timing.
Model
Research video generation framework that keeps scenes consistent after edits using RGB and depth context memory banks.
Model · Runway
Runway's video-to-video model for editing and transforming existing footage with text prompts.
Research · Google
Research method for free-view rendering and relighting of full-body dynamic humans from four sparse RGB views.
Model
Research method for re-rendering an existing video along a new camera trajectory while preserving subject motion.