Modeling and Synthesis of Realistic Visual Speech in 3D - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 287

complete visemes and single independent components. Having the system work
on the basis of the same keyframes (i.e., in a Viseme Space) helps to make the
interaction with the animator more intuitive, as the animator and the animation
tool “speak” the same language.
Results
As a first example, we show some frames out of an animation sequence, where
the person is animated using his own visemes. This person was one of our
examples. Four frames are shown in Figure 16. Although it is difficult to
demonstrate the realism of animation on paper, the different face expressions
look natural, and so does the corresponding video sequence.
A second example shows the transition between two faces (see Figure 17). In
this case, the visemes of the man are simply cloned to get those for the boy (Noh
Figure 16. One of the example faces uses its own visemes.
Figure 17. To see the effect of purely cloned visemes, a specific experiment
was performed. The man's visemes are kept throughout the sequence and
are cloned onto the mixed face.