Methods and apparatus for facial animation using motion capture data are described herein. A mathematic solution based on minimizing a metric reduces the number of motion capture markers needed to accurately translate motion capture data to facial animation. A set of motion capture markers and their placement on an actor are defined and a set of virtual shapes having virtual markers are defined. The movement of the virtual markers are modeled based on an anatomical model. An initial facial capture is correlated to a corresponding virtual reference shape. For each subsequent facial capture, a delta vector is computed and a matrix solution determined based on the delta marker, initial positions, and set of virtual shapes. The solution can minimize a metric such as mean squared distance. The solution can be manually modified or edited using a user interface or console.