BodySpeech: A configurable facial and gesture animation system for speaking avatars

Speaking avatars are present in many Human Computer Interaction (HCI) applications. Their importance lies in communicative goals which entail interaction within other avatars in virtual worlds or in marketing where they have become useful in customer push strategies. Generating automatic and plausible animations from speech cues have become a challenge. We present BodySpeech, an automatic system to generate gesture and facial animations driven by speech. Body gestures are aligned with pitch accents and selected based on the strength relation between speech and body gestures. Concurrently, facial animation is generated for lip sync, adding emphatic hints according to intonation strength. Furthermore, we have implemented a tool for animators. This tool enables us to modify the detection of pitch accents and the intonation strength influence on output animations, allowing animators to define the activation of gestural performances.