Parallel WaveGAN: Fast and High-Quality GPU Text-to-Speech -English version-

Ryuichi Yamamoto (r9y9) / LINE

■ Session Overview
Text-to-speech is a technology that synthesizeshuman voice from a text by computer. For services that require real-time interaction with users, such as LINE CLOVA smart speakers, text-to-speech systems are required to have high synthesis quality and generate speech at high speed. In this session, we will present the results of our research on GPU-
based speech synthesis, which we developed jointly with NAVER and LINE. Typical conventional methods suffer from the slow generation or long training time. This session covers how we addressed these issues, based on a paper accepted for ICASSP 2020, a top conference on speech signal processing, with recent developments in related fields.

■ Official Site #linedevday
https://linedevday.linecorp.com/2020/en/sessions/8721

■ Category
AI&Data - AI

Видео Parallel WaveGAN: Fast and High-Quality GPU Text-to-Speech -English version- канала LINE Developers