A new AI system can replicate realistic human avatars without the use of cameras or motion capture technology. The technology analyses audio and textual data to create 3D models that closely resemble the person's appearance and mouth movements as they talk.