‘What if you could manipulate the facial features of a historical figure, a politician, or a CEO realistically and convincingly using nothing but a webcam and an illustrated or photographic still image?’
A new technique known as facial reenactment enables any image of a face to be put in motion by a third-party. VENTURE BEAT reports the new image can portray any emotion or action, or what SYNCED calls ‘any humanly possible face-based task.’
- The source of the movement is known as the ‘driver face.’ The original image is the ‘target face.’
- The technique is known as ‘few shot’ modelling, using machine learning to extract and then extrapolate the facial features.
- The target face can be only a single frame. Previous state-of-the-art systems required more frames to act as training data, for instance a few seconds (video typically has 30 frames per second.)
- Potential uses include motion tracking for film/tv production.
- Potential abuses include more plentiful deep fakes.
The new system is called MarioNETte. It was developed by a Seoul research lab, Hyperconnect.
Researchers train AI to map a person’s facial movements to any target headshot
VENTURE BEAT | November 27, 2019 | by Kyle Wiggers
- MarioNETte: Few-Shot Identity Preservation in Facial Reenactment
SYNCED | November 26, 2019 | ‘Starting with a dynamic driver face, researchers can manipulate any target face — from today’s celebrities to historical figures, including any age, ethnicity or gender — to perform any humanly possible face-based task.’