Communication Dans Un Congrès Année : 2025

The GIPSA-Lab Text-To-Speech System for the Blizzard Challenge 2025

Résumé

This paper describes the GIPSA-Lab submission to the Blizzard Challenge 2025. The Text-To-Speech system trained for this challenge is an autoregressive encoder-decoder architecture based on Tacotron2. Updates of the Tacotron2 framework were provided to specifically train the model on orthographic inputs, which is our main focus for this edition of the challenge. This model was trained with both orthographic and phonetic transcriptions but with no phonetic alignments with audio. An additional phonetic prediction layer was added to the model. This additional layer enables to train the text encoder on phonetic prediction alone, without the need for audio recordings.

Fichier principal
Vignette du fichier
gb_Blizzard2025.pdf (353.91 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05368789 , version 1 (17-11-2025)
hal-05368789 , version 2 (05-01-2026)

Licence

Identifiants

  • HAL Id : hal-05368789 , version 1

Citer

Gérard Bailly, Olivier Perrotin. The GIPSA-Lab Text-To-Speech System for the Blizzard Challenge 2025. 19th Blizzard Challenge Workshop, Aug 2025, Leeuwarden, Netherlands. ⟨hal-05368789v1⟩
453 Consultations
198 Téléchargements

Partager

  • More