I’ve dedicated a lot of time to testing every new AI video tool on the market, from OpenAI’s Sora to the most recent updates from Runway. Typically, the pitch is the same: "Type a prompt, get a movie. " However, Netflix has quietly launched a research model named VOID, and it’s taking a totally different approach. Instead of generating new worlds or scenes from nothing, it modifies the footage you’ve already shot — and it’s so effective you might stop trusting “real” videos altogether. What is Netflix VOID? VOID stands for Video Object and Interaction Deletion. At first glance, it resembles an advanced version of the "Magic Eraser" on devices like the Pixel 8 or Galaxy S24, where you select an object and it disappears. Article continues below But here’s the revolutionary part: VOID comprehends physics and causality. Unlike typical editing tools that just fill in the gap with background textures, VOID rewrites the scene’s logic to reflect the absence of the object. Several examples on GitHub showcase its capabilities: The Guitar Test: In a demo, a person holding a guitar is removed. In conventional tools, the guitar would simply vanish or float unnaturally. VOID understands the guitar is no longer supported and generates frames that show it falling naturally to the ground. The Crash Test: Removing one car from a head-on collision doesn’t leave behind any residual fire or smoke. VOID “re-imagines” the trajectory of the remaining car as if the accident never occurred — transforming a wreck into a calm drive on an empty road. Why this marks the "end of the reshoot" For Netflix, this represents a huge potential cost saver for the film industry. Remember the notorious "Game of Thrones" Starbucks cup mistake?

Typically, fixing such an error involves painstaking frame-by-frame digital editing. With VOID, a producer can just delete the unwanted item and allow the AI to realistically simulate what should happen next — whether that means splashes of water, settling dust, or nothing at all. And it’s not limited to minor fixes. Instead of a massive crew returning for a reshoot, AI could fix mistakes after filming wraps. It could even alter plot details by removing important objects and recalculating the scene so everything still looks authentic. Can you try it yourself? Perhaps the most surprising part is that Netflix has open-sourced VOID. The model is currently available on Hugging Face, licensed under Apache 2. 0. But don’t expect to run it on a typical laptop. VOID is resource-intensive, requiring a GPU with at least 40GB of VRAM (such as an NVIDIA A100 or H100) for smooth inference. It’s built on a 5-billion parameter version of CogVideoX and employs a proprietary “quadmask” system that guides the AI on which parts of the physics to recalculate. The takeaway The “visual receipt” used to be the ultimate proof, but that’s changing fast. Netflix has created a tool that can rewrite actual footage so flawlessly it appears completely real. Meanwhile, AI-generated content is becoming more convincing than ever — flooding the internet with visuals that feel authentic but are not. This points to a future where seeing is no longer believing. Welcome to the era of editable reality. Follow Tom's Guide on Google News and add us as a preferred source to keep receiving our latest news, analysis, and reviews directly in your feed. More from Tom's Guide