AI Deepfake Technology Brings the Mona Lisa to Life
A new artificial intelligence system uses advanced neural networks to animate the Mona Lisa and other famous portraits into realistic, moving videos.
Artificial intelligence gives the famous Mona Lisa a startling new range of expressions in a recently shared deepfake video. A convolutional neural network analyzes and processes images similarly to a human brain, allowing the system to generate realistic video sequences from a single painted frame.
Researchers train the algorithm to understand the general shapes of facial features and how they move relative to each other. For the Mona Lisa animations, the AI learns facial movements from three different human subjects, which lends distinct and recognizable "personalities" to each living portrait.
The engineers also apply this technology to photographs of historical icons like Albert Einstein and Marilyn Monroe. Because human heads are highly complex and people easily notice minor mistakes in 3D models, creating these deepfakes presents a significant challenge that often triggers an uncanny valley effect for viewers.