Named Speech2Face, the neural network was trained by scientists on millions of videos that showed over 100,000 different people talking.








Source: AI Listened to People's Voices. Then It Generated Their Faces.