Audio-driven Speech Animation for Virtual Characters

Abstract

We propose a procedural audio-driven speech animation method in which animation is generated from raw audio. Given any audio with its respective speech transcript, we generate speech animation that could drive any 3D animated character. We introduce a new co-articulation method based on the state-of-the-art but with unique dynamic elements. The generated animation is configurable by the user by modifying the relative parameters that are required to drive it. The realism of the final animation is enhanced by studying the emotional aspect of speech and applying it to the animation itself. We evaluate our approach against two popular speech animation packages and we show that our method could surpass them in terms of user preference.