Netflix has open sourced an AI framework that can remove objects from videos and automatically adjust the physical impact those objects have on the rest of the scene. This system is called VOID, which stands for “Video Object and Interaction Deletion.” What makes this special is that it not only erases objects from the scene, but also handles the downstream physics effects, such as collisions, that the removed object caused in the first place.
VOID is built on Alibaba’s CogVideoX video diffusion model and fine-tuned with synthetic data from Google’s Kubric and Adobe’s HUMOTO for interaction detection. Google’s Gemini 3 Pro analyzes the scene to identify affected areas, while Meta’s SAM2 handles segmenting objects that need to be removed. An optional second pass uses optical flow to correct shape distortions.
The project was developed by Netflix researchers in collaboration with INSAIT Sofia University. Code, papers, and demos are available on GitHub, arXiv, and Hugging Face. This system is shipped under the Apache 2.0 license, so it can be used commercially.
AI News Without the Hype – Curated by Humans
as The Decoder Subscriberyou can read without ads. Weekly AI Newsletterexclusive “AI Radar” Frontier Report 6 times a yearaccess comments, and Complete archive.
Subscribe now
