High Fidelity Text-to-Speech Via Discrete Tokens Using Token Transducer and Group Masked Language Model

Joun Yeop Lee, Myeonghun Jeong, Minchan Kim +3

We propose a novel two-stage text-to-speech (TTS) framework with two types of discrete tokens, i.e., semantic and acoustic tokens, for high-fidelity speech synthesis. It features t…