Skip to Main content Skip to Navigation
Master Thesis

Validation d'un processus de création de voix contextuelle et intégration de nouvelles langues à une application de synthèse vocale grand public

Abstract : With speech technology playing an increasing role in our everyday lives, the quest for quality and naturalness in speech synthesis has become a major objective. This report highlights two techniques developed by Voxygen to ensure high-quality, natural and expressive TTS: the use of a particular unit, the vocalic sandwich; and the creation of Slot'N'Fill voices. It also details how these techniques can be applied to various languages. This paper sums up a six-months internship at Voxygen, during which the Slot'N'Fill voice creation process was refined in two languages, and new languages incorporated into mainstream voice-synthesis application. Both of these tasks aimed at producing high-quality synthesis, each in their own way. Therefore, this report evaluates the relevance of the sandwich as a sound-database unit, as well as the impact of corpus choice in creating natural and expressive voices. Since the internship allowed the author to participate in all stages of the voice-creation process in several languages, notions will be expounded from a multilingual and trans-disciplinary perspective.
Document type :
Master Thesis
Complete list of metadatas

Cited literature [20 references]  Display  Hide  Download

https://dumas.ccsd.cnrs.fr/dumas-01215488
Contributor : Uga - Bulles <>
Submitted on : Wednesday, October 14, 2015 - 12:28:47 PM
Last modification on : Wednesday, July 15, 2020 - 1:06:02 PM
Long-term archiving on: : Friday, January 15, 2016 - 12:20:42 PM

Identifiers

  • HAL Id : dumas-01215488, version 1

Citation

Camille Lecorgne. Validation d'un processus de création de voix contextuelle et intégration de nouvelles langues à une application de synthèse vocale grand public. Sciences de l'Homme et Société. 2015. ⟨dumas-01215488⟩

Share

Metrics

Record views

216

Files downloads

1156