Title: SSL-TTS
DNr: Berzelius-2023-268
Project Type: LiU Berzelius
Principal Investigator: Éva Székely <>
Affiliation: Kungliga Tekniska högskolan
Duration: 2023-10-10 – 2024-05-01
Classification: 10208


This project explores new methods for neural speech synthesis. We hypothesize that a set of new neural speech synthesis methods would provide several benefits to language technologies that incorporate speech synthesis. The language technologies that would potentially benefit from postive results of this project include speech interface, conversational systems, social robots. Neural models for speech synthesis require considerable GPU resources. We hope to utilize resources at Berzelius to conduct our planned experiments.