The researchers use AI to capture individual movements from a video input and convert them into digital avatars.