Kinship in Speech: Leveraging Linguistic Relatedness for Zero-Shot TTS in Indian Languages

TL;DR

This paper introduces a zero-shot TTS approach for Indian languages by leveraging linguistic relatedness and shared phonetic representations, enabling rapid adaptation and synthesis for resource-scarce languages.

Contribution

It proposes a novel method of augmenting shared phone representations and adjusting text parsing to facilitate zero-shot TTS for diverse Indian languages.

Findings

Natural speech generated for multiple Indian languages

Effective cross-lingual transfer demonstrated through evaluations

Reduced adaptation overhead for under-resourced languages

Abstract

Text-to-speech (TTS) systems typically require high-quality studio data and accurate transcriptions for training. India has 1369 languages, with 22 official using 13 scripts. Training a TTS system for all these languages, most of which have no digital resources, seems a Herculean task. Our work focuses on zero-shot synthesis, particularly for languages whose scripts and phonotactics come from different families. The novelty of our work is in the augmentation of a shared phone representation and modifying the text parsing rules to match the phonotactics of the target language, thus reducing the synthesiser overhead and enabling rapid adaptation. Intelligible and natural speech was generated for Sanskrit, Maharashtrian and Canara Konkani, Maithili and Kurukh by leveraging linguistic connections across languages with suitable synthesisers. Evaluations confirm the effectiveness of this approach, highlighting its potential to expand speech technology access for under-represented languages.