Sofia University’s Institute of Computer Science, Artificial Intelligence and Technology (INSAIT) and streaming platform Netflix have developed a new artificial intelligence model that can remove objects from videos while realistically reconstructing the scene, according to a Bulgarian institute.
This model, called VOID (Video Object and Interaction Deletion), goes beyond standard tools that simply fill in deleted space. Analyze how objects interact in a scene and predict what will happen after an element is removed. For example, if a person holding an object is erased, the system will simulate the object falling or moving naturally and adjust shadows, movement, and other physical effects accordingly.
Unlike traditional video editing tools that focus on visual consistency, VOID has a built-in understanding of physical cause and effect, allowing you to recreate scenes as if they were originally filmed without objects being removed.
The technology is based on large-scale video models and trained on synthetic datasets designed to simulate real-world interactions, allowing it to produce more realistic and consistent results across frames.
VOID gives film and video producers greater creative flexibility, reduces the need for costly reshoots, and enables more advanced post-production editing.
Editor: Miglena Ivanova
Edited and posted in English by E. Radkova
