TTS4ALL - Text-to-Speech in Low Resource Scenarios - Data Management, Methodology, Models, Evaluation

Explore text-to-speech (TTS) technology for low-resource languages through this comprehensive workshop presentation from JSALT 2025's closing day. Delve into the challenges and solutions for developing TTS systems when working with limited linguistic resources, covering essential aspects of data management, research methodology, model development, and evaluation techniques. Learn from leading researchers Yannick Estève and Fethi Bougares as they present findings from five weeks of intensive collaborative research at Brno University of Technology. Discover innovative approaches to creating speech synthesis systems for underrepresented languages, understand the complexities of working with scarce training data, and examine evaluation frameworks specifically designed for low-resource TTS scenarios. Gain insights into practical methodologies for data collection and management in resource-constrained environments, explore state-of-the-art model architectures adapted for limited data scenarios, and understand the unique challenges faced when developing TTS systems for languages with minimal digital presence. The presentation includes interactive elements with opportunities for questions and discussion, making it valuable for researchers, developers, and practitioners working in speech technology, computational linguistics, and language preservation efforts.