Visual speech synthesis using shape and appearance models