MIT researchers have developed an open-source pipeline for generating realistic animated avatars. The avatar received at its output is marked with a watermark to distinguish it from the original content.
Avatars are proposed to be used in tasks to improve educational content and to preserve confidentiality in conversations without erasing nonverbal signals.
Other apps may include avatars who help conduct therapy to alleviate the growing shortage of mental health professionals and reach the approximately 44 percent of Americans with mental health problems who never receive counseling. The technology can also be used to anonymize faces on video while preserving facial expressions and emotions, which can be useful for sessions where people want to share confidential information, such as information about health and injuries, or for informants and witness testimony.
The pipeline consists of a combination of generative-adversarial neural networks for animating facial expressions and gestures, as well as for generating voice.
