We present a new method for generating a dynamic, concatenative, unit of visual speech that can generate realistic visual speech animation. We redefine visemes as temporal units that describe distinctive speech movements ...