Netflix just published its first-ever public AI model, and it's not a recommendation algorithm or a subtitle generator. It's called VOID - Video Object and Interaction Deletion - and it does exactly what the name says: removes objects from video footage while keeping the scene looking natural.
The model, now available on Hugging Face, tackles one of the harder problems in video editing. Removing a static object from a photo is relatively straightforward (Photoshop has done it for years). Removing an object from video is much harder because you need consistency across hundreds of frames. VOID goes a step further by also handling "interaction deletion" - removing not just the object itself but the traces it leaves behind, like shadows, reflections, or the way other objects react to it.
Netflix operating a massive internal AI operation is no secret. The company has published research papers for years. But releasing a usable model that anyone can download and run locally? That's new territory for them. It signals Netflix sees value in contributing to the open-source AI ecosystem rather than keeping everything behind closed doors.
For video editors and VFX artists, VOID addresses a real pain point. Current object removal in video tools like After Effects requires tedious frame-by-frame work or expensive plugins that produce inconsistent results. An open model that handles this well could save hours of manual cleanup on every project. The practical question is how well it performs on real-world footage with complex motion and lighting - research demos tend to show best-case scenarios.