CONCLUSION     The facial animation model proposed here offers cosmetic realism insofar as it generates faces that look like the original speaker. However, even without the many cosmetic improvements yet to be implemented, it has largely solved the problem of generating realistic motions from a small set of control parameters. These parameters have the significant additional advantage of being highly correlated with the other observable events associated with speech production; namely, the underlying physiological activity, the deformations of the vocal tract that to a large extent are responsible for the visible facial motions during speech, and finally the speech acoustics. No other model of facial motion can claim such a realistic grip on the production of audiovisual behavior. Of course, the model's communicative efficacy in human perception can be judged only by its effects on perceivers, a job currently underway.   View QuickTime movie with sound (782K).   Introduction | Audible-Visible Synthesis | Extensions to the Basic Model Issues of Realism | Conclusion | References