Netflix’s VOID AI is an open-source model that removes objects from video footage while realistically simulating physical interactions and scene changes. Rather than generating new content from scratch, VOID—which stands for Video Object and Interaction Deletion—edits existing footage by understanding how physics, shadows, reflections, and cause-and-effect dynamics change when something disappears. Developed by Netflix researchers including Saman Motamed and William Harvey, the system was open-sourced under Apache 2.0 on Hugging Face as of April 4, 2026, making it free for anyone to use.
Key Takeaways
- VOID removes objects from video while simulating realistic physics, shadows, and reflections—not just erasing pixels
- Built on Alibaba’s CogVideoX-Fun diffusion model, fine-tuned with a proprietary Quadmask system that encodes physical interactions
- User study showed VOID preferred 64.8% of the time versus competitors like Runway (18.4%)
- Completely free and open-source; no integration into Netflix products yet
- Handles complex interactions: removes a person jumping in a pool (no splash), deletes a vehicle in a collision (other vehicle undamaged)
How Netflix VOID AI Actually Works
Netflix VOID AI doesn’t work like the content-aware fill tool in Photoshop or the painstaking frame-by-frame labor of a VFX artist. Instead, it combines vision-language input—you provide a video and a text description of what to remove—with a technique called Quadmask. The Quadmask uses four values (0, 63, 127, 255) to encode not just the removal area but also the physically affected regions around it: shadows that disappear, reflections that vanish, and collision paths that change.
The system then runs a diffusion process over all temporal frames, using Alibaba’s CogVideoX-Fun-V1.5-5b-InP as its foundation. This generates plausible physics—gravity pulls falling objects, motion continues smoothly, and cause-and-effect chains resolve naturally. If you remove a person holding a briefcase, the briefcase drops. If you delete one car in a multi-vehicle collision, the other cars respond as if that vehicle was never there. The output is a convincingly edited video that looks like it was shot that way originally, not manipulated afterward.
Why Netflix VOID AI Beats Existing Video Editing Tools
Existing object-removal tools like Runway, Generative Omnimatte, and DiffuEraser excel at fixing surface-level artifacts—erasing shadows, removing reflections—but they fail when interactions matter. Remove a person from a scene where they’re colliding with another actor, and competitors leave collision debris, smoke, or awkward gaps. Netflix’s VOID AI models the dynamics that follow from removal, making it fundamentally different in scope.
In a user study of 25 participants across multiple scenarios, VOID was preferred 64.8% of the time, with Runway placing second at 18.4%. That’s a decisive margin—nearly two-thirds of users chose Netflix’s approach over established alternatives. The preference reflects VOID’s ability to handle the physics that other tools ignore: a person jumping out of a pool leaves no splash when removed; a vehicle in a crash scene disappears without leaving wreckage or affecting neighboring cars.
Open Source Means Filmmakers Can Edit Without Reshoots
Netflix’s decision to open-source VOID under Apache 2.0 is the strategic move here. The company isn’t selling a subscription service or locking the technology behind a studio paywall. Instead, it’s planted itself at the center of AI video editing by giving filmmakers and editors a free tool that eliminates reshoots for scene modifications. A director who realizes a boom mic is visible in a take, or an actor needs to be removed for a reshoot, can now edit the scene algorithmically rather than calling the crew back to set.
The accessibility cuts both ways. Making reality-rewriting this easy and free raises legitimate concerns about misinformation—deepfakes and manipulated videos become harder to detect when the editing looks this natural. But from a production standpoint, VOID solves a real problem: VFX artists spend weeks on invisible edits that now take minutes.
What Netflix VOID AI Still Doesn’t Do
VOID is not yet integrated into Netflix’s production pipelines or consumer products. It remains a research release, available to anyone who wants to download it and run it locally or on Hugging Face. The model works on video footage you provide; it doesn’t generate new scenes or create content from text alone. If you want to remove an object, you need the video first. That’s a meaningful limitation compared to generative AI tools like Runway, which can create entirely new footage, but it’s also why VOID’s output looks photorealistic—it’s editing reality, not inventing it.
Is Netflix VOID AI free to use?
Yes. Netflix VOID AI is completely free and open-source under the Apache 2.0 license. Anyone can download it from Hugging Face and use it without paying or subscribing. There are no regional restrictions or premium tiers.
Can I use Netflix VOID AI for commercial projects?
The Apache 2.0 license permits commercial use, but you should review the full license terms and consider the legal implications of editing video in your jurisdiction, especially regarding deepfakes and misrepresentation.
How does Netflix VOID AI compare to Runway?
Both remove objects from video, but they approach the problem differently. Runway generates new content; VOID edits existing footage while modeling physics and interactions. In user preference studies, VOID was chosen 64.8% of the time versus Runway’s 18.4%, largely because VOID handles complex dynamics like collisions and falling objects more convincingly.
Netflix VOID AI represents a meaningful shift in how video editing works. It’s not a generative tool that makes something from nothing—it’s a surgical editor that understands physics. For filmmakers and studios, that’s worth far more than another text-to-video generator. The fact that it’s free and open-source means the technology will shape how professional video gets made, for better and worse. If you’re editing video and need to remove an object without reshooting, VOID is now the tool to try first.
This article was written with AI assistance and editorially reviewed.
Source: Tom's Guide
