Abstract: In this work we present a method for human face animation which allows us to generate animations for a novel person given just a single mesh of their face. These animations can be of arbitrary text and may include emotional expressions. We build a multilinear model from data which encapsulates the variation in dynamic face motions over changes in identity, expression, and over different texts. We then describe a synthesis method consisting of a phoneme planning and a blending stage which uses this model as a base and attempts to preserve both face shape and dynamics given a novel text and an emotion at each point in time.