Nos_Brais-GL: A FAIR Galician TTS Corpus for Neural Speech Synthesis

Abstract

This paper introduces Nos_Brais-GL, a new open-access high-quality Galician speech corpus designed for the development of neural Text-to-Speech (TTS) systems. Nos_Brais-GL contains approximately 18 hours of professionally recorded male speech and a carefully curated set of utterances selected to ensure linguistic variation and phonetic and prosodic richness. Beyond its immediate application in synthetic speech generation, Nos_Brais-GL exemplifies good practices in TTS corpus design for lesser-resourced languages, emphasizing methodological transparency, open licensing, and interoperability.

Resources

Details