TTS in codemix domain

Abstract

We aim to develop a Text-to-Speech (TTS) system for code-mixed corpora, addressing the unique challenges posed by multilingual and mixed-language text. To establish a robust foundation, we first reviewed TTS literature comprehensively, focusing on state-of-the-art architectures, datasets, and techniques for multilingual and code-mixed scenarios. Key insights from this review guided our approach to develop a TTS model tailored for code-mixed data. We present an analysis of relevant studies and their reported results, highlighting the strengths and limitations of existing methods, which serve as a benchmark for our system.