AUT-Talk: A Farsi Talking Head

Abstract

This paper discusses the design and implementation procedure for AUT-Talk, a Farsi talking head. This procedure consists of developing a Farsi TTS system and using facial animation to generate visible speech. Having the reuse concept in mind, we employed CSLU Toolkit as the framework. The text-to-speech engine uses diphone concatenative approach to synthesize Farsi speech based on Festival TTS system. Our evaluations show that the resulting system works very well in producing visible speech, i.e. the adaptation of mouth and tongue movement to speech. The voice is well synchronized with the lip motion. The resulting Farsi voice is robot-like, but completely intelligible