Ever wanted to put your own mouth motions on someone else’s face in a video?
Stanford University visiting scientist Matthias Niessner along with Justus Thies from the University of Erlangen-Nuremberg and their colleagues have developed a software that can manipulate videos by transferring facial expressions in real-time, reported the New York Times.
Dubbed as a “live facial re-enactment program,” the system can map pixels from faces in videos in order to transfer expressions. This technique allows the placing of one person’s smile, frown or raising of eyebrows to appear on live video of another person’s face. With the use of a high-grade camera, the 30-millisecond transfer differential is not noticeable to the naked eye.
In its latest iteration, the rendering of the target’s mouth is greatly improved, producing a seamless effect.
The only current limitation is the program’s inability to determine fingers and handle motion beyond a 30-degree angle from the neck.
While the technology’s objective is to improve dubbing in movies and aid in improving auto-translated Skype conversations between foreigners, the potential for other uses is undoubtedly great.
The team is scheduled to present a demo at the Computer Vision Conference CVPR in Las Vegas in June this year.