INSAIT and Netflix Develop Revolutionary AI Model VOID for Object Removal from Video

08.04.2026 | Technologies

INSAIT and Netflix have unveiled VOID - an AI model that realistically removes objects from video and recreates the scenes. The technology offers greater freedom to video creators and saves resources. VOID is open-source.

Pexels

The Institute for Computer Science, Artificial Intelligence and Technology (INSAIT) at Sofia University "St. Kliment Ohridski" and "Netflix" have developed a new AI model - VOID, which removes objects from videos and realistically recreates the changes in the scene.

VOID understands how objects interact, simulating falling or movement after an object is removed. It uses a special quadmask approach to distinguish objects, interaction zones, and the background, preserving the logic of the frame.

Simulated scenes generated through Blender are used for training. VOID achieves better visual consistency and realistic object behavior. The model is open source.

INSAIT is a partner with ETH Zurich and EPFL and is part of Sofia University. The ideologist is Prof. Martin Vechev.