The IRISA Text-To-Speech System for the Blizzard Challenge 2018
Résumé
This paper describes the implementation of the IRISA unit selection-based TTS system for our participation to the Blizzard Challenge 2018. We describe the process followed to build the voice from given data and the architecture of our system. It uses a selection cost which integrates notably a bottleneck DNN-based embedding prediction. Unit selection is based on a Viterbi-based algorithm with preselection filters used to reduce the search space. The system achieves average results compared to others.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...