Netflix has launched VOID, an advanced AI model designed to remove specific objects from videos while seamlessly restoring the surrounding scene. Reported by The Register on April 3, VOID, or Video Object and Interaction Deletion, utilizes a cutting-edge vision-language framework. It effectively deletes elements—such as vehicles in a collision—while reimagining the natural motion of remaining objects. For instance, when a vehicle is removed, VOID alters the scene to depict the intact vehicle continuing down the road. It also cleans up post-collision debris and replaces it with a pristine environment. Developed in collaboration with Sofia University, VOID excels at inpainting, a technique that reconstructs missing video segments. Available on Hugging Face, this innovative tool outperforms competitors like Runway and DiffuEraser, achieving a 64.8% preference rate among surveyed users. For creators and editors, VOID represents a significant advancement in digital content manipulation, enhancing storytelling capabilities in video production.
Source link
