On October 15, the Giant Network AI Lab, in collaboration with the SATLab from the Department of Electronic Engineering at Tsinghua University, unveiled DiaMoE-TTS—a large-scale model framework designed for multi-dialect speech synthesis. This initiative stands out for its complete open-sourcing of data, code, and methodologies, reflecting a commitment to fostering equitable and inclusive advancements in the realm of dialect speech synthesis. At the core of this framework lies a unified IPA (International Phonetic Alphabet) expression system, meticulously crafted with the expert guidance of linguists. The framework's scalability and robustness have been rigorously tested and validated across a spectrum of multilingual environments, encompassing languages such as English, French, and German. Additionally, it boasts specialized versions tailored for Chinese dialects, including Cantonese and Sichuanese, further underscoring its versatility and global applicability.
