New Software Tech Allows Believable Manipulation of Faces in Videos
As if it wasn't hard enough already to believe what you see and hear on the Internet, new software is being developed that allows you to manipulate the faces and mouth movements of people in existing videos.
Imagine, if you will, a broadcast of a prominent world leader giving a live address being watched by millions. This one-time dove has suddenly become a hawk, announcing war and totalitarian policies. Little would you know that the leader's speech has been altered by software that can do real-time manipulation of the face and mouth by a voice actor saying the words.
Sound far-fetched? Not with some new "Face2Face" technology being developed by researchers from the University of Erlangen-Nuremberg, Max-Planck-Institude for Informatics, and Stanford University. The technical video shows the "target actor" - in this case the politician of your choice - rendered with a neutral expression. Then the source actor could say any words or offer any expressions of their choice. Right now, the technology works on existing YouTube videos.
This is the same team that did real-time expression transfer [http://graphics.stanford.edu/~niessner/thies2015realtime.html] just last year. The project is looking at real-world applications that would allow a translator to give a guest speaker the appearance of speaking in the native tongue of an audience.
Obviously, it is still a work in progress, but the potential is fascinating, if not downright scary. We all know that programs like this could easily be used for more nefarious means, right?
Source: Stanford University [http://www.graphics.stanford.edu/~niessner/thies2016face.html]
Permalink
As if it wasn't hard enough already to believe what you see and hear on the Internet, new software is being developed that allows you to manipulate the faces and mouth movements of people in existing videos.
Imagine, if you will, a broadcast of a prominent world leader giving a live address being watched by millions. This one-time dove has suddenly become a hawk, announcing war and totalitarian policies. Little would you know that the leader's speech has been altered by software that can do real-time manipulation of the face and mouth by a voice actor saying the words.
Sound far-fetched? Not with some new "Face2Face" technology being developed by researchers from the University of Erlangen-Nuremberg, Max-Planck-Institude for Informatics, and Stanford University. The technical video shows the "target actor" - in this case the politician of your choice - rendered with a neutral expression. Then the source actor could say any words or offer any expressions of their choice. Right now, the technology works on existing YouTube videos.
This is the same team that did real-time expression transfer [http://graphics.stanford.edu/~niessner/thies2015realtime.html] just last year. The project is looking at real-world applications that would allow a translator to give a guest speaker the appearance of speaking in the native tongue of an audience.
Obviously, it is still a work in progress, but the potential is fascinating, if not downright scary. We all know that programs like this could easily be used for more nefarious means, right?
Source: Stanford University [http://www.graphics.stanford.edu/~niessner/thies2016face.html]
Permalink