In the project, we build a comprehensive framework that allows to equip arbitrary robots with highly customized voices. Therefore, we make use of novel human-in-the-loop optimization techniques to modify existing state-of-the-art TTS models.
Currently, most speech frameworks that are available are mostly limiting their customization capabilities to very basic traits like gender or age. For social robots, which need to be thoroughly well-designed to foster acceptance and trust, such basic customization possibilities are not sufficient. In a variety of scientific publications, we have already proven that our specific approach of enabling a fine-grained personalization process through human-in-the-loop optimization techniques is feasible for a huge variety of robots. In this project, we now aim for a complete and comprehensive software system that makes our solutions accessible to the whole robot community.