You’ll never believe what you watch someone say in a video again.
A research team has created software that allows them to control the face of anyone in any video. Using advanced facial recognition, it looks at about 15 seconds of any face in a video and creates a 3D model of that face in real time. The result is incredibly realistic, as you can see (ahem) in the video above.
The project is a collaboration between computer scientists at Stanford University, the Max Planck Institute, and the University of Erlangen-Nuremberg in Germany. The same team produced a similar study last year, but that iteration required data from special 3D cameras. Their new system works with any camera and any recorded video.
The result is a weird cross between Snapchat’s “face swap” feature and the “gibberish” scene from Bruce Almighty. It’s an effect that Hollywood studios spend millions to achieve. But it might be about to get a whole lot more common.