Netflix open-sources VOID — AI that removes objects AND their physics
VOID deletes objects from video along with every physical interaction they caused — not just shadows, but gravity too.
Netflix just dropped VOID (Video Object and Interaction Deletion) as an open-weight model on Hugging Face. Unlike typical video inpainting tools that erase an object and patch the background, VOID understands physical causality — remove a person holding a guitar and the guitar falls. Remove hands gripping a door and the door swings free.
Built on top of CogVideoX and fine-tuned with interaction-aware mask conditioning, VOID runs as two sequential transformer passes: one for base inpainting, one for temporal consistency. It's licensed Apache 2.0, meaning commercial use is fair game.
In human preference tests, VOID was chosen over Runway 64.8% vs 18.4% — a wide margin that signals this isn't just a research demo.
Why it matters: physics-aware video editing unlocks a new tier of post-production tooling — and Netflix just handed it to everyone for free.
Sources
Primary: the company, paper or repository
- Netflix VOID on GitHub github.com
- VOID: Video Object and Interaction Deletion (arXiv) arxiv.org
Independent coverage
- Now even Netflix has its own video AI — The Register theregister.com
Written by an AI pipeline from the sources above. Methodology · Report an error
Feed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.