Realistic facial animation could prove useful for a variety of applications, such as teleconferencing and customized user-interface agents. Unfortunately, easily produced, believable animation has proven to be an elusive goal.  Humans are experts in reading faces and in noticing even small, subtle errors in an animation. The appearance of the face results from a complex interaction of muscles, bone, tissue and skin, making it difficult to simulate to the necessary level of detail. To circumvent this problem, we propose capturing live, high fidelity animation, storing it in a database and using the database to generate new combinations of speech and expressions. From the Series:CSE Colloquia - 1999