Virtual Talking Heads and audiovisual articulatory synthesis

Our approach to audiovisual articulatory synthesis involves the development of Virtual Talking Heads that integrate the articulatory, aerodynamic and acoustic phenomena underlying speech production. Specifically, these Talking Heads are faithful clones of the speakers whose data the various models are based on. Our contribution presents some of the results achieved at ICP in this domain: 3D orofacial linear articulatory models, made possible by recent progresses in medical imaging and video processing techniques; aerodynamic and acoustic models, and basic glottis / oral constriction coordination principles; development of various strategies to determine the articulatory control parameters (coarticulation models vs. simple concatenative strategies); text-to-audiovisual speech synthesis. We finally make some suggestions for future developments.