Meet DIRFA, a program developed by #NTUsg scientists that can generate realistic 3D videos of people talking from just their passport photo and an audio clip.
The 'DIverse yet Realistic Facial Animations' program is #AI-based and trained on over one million audiovisual clips from 6,000 people from an #opensource dataset. DIRFA improves on existing programs that typically struggle with pose variations and emotional control.
DIRFA could be used to create more realistic virtual assistants and chatbots, improving online user experiences across many industries. It could also help people with speech or facial disabilities communicate through expressive avatars or digital representations.