We demonstrate a method for generating a 3D virtual character performance from the audio signal by inferring the acoustic and semantic properties of the utterance. Through a prosodic analysis of the acoustic signal, we ...