Generalized Multilingual Text-to-Speech Generation with Language-Aware Style Adaptation

TL;DR

LanStyleTTS tackles multilingual TTS by addressing phoneme mismatches and language-specific prosody with a non-autoregressive, language-aware framework that standardizes phonemes using IPA and employs a style adaptation module for phoneme-level prosody control. It supports both FastSpeech-like backbones and VITS end-to-end waveform generation through two variants, and demonstrates improvements in WER and MOS across English and Chinese. The study also shows that IPA-based tokenization and latent acoustic features can significantly reduce model size and inference time while preserving quality, offering practical benefits for large-scale multilingual deployment. Overall, LanStyleTTS provides a scalable, flexible approach to high-quality multilingual speech synthesis with tangible efficiency gains and actionable guidance for tokenization and feature representation.

Abstract

Text-to-Speech (TTS) models can generate natural, human-like speech across multiple languages by transforming phonemes into waveforms. However, multilingual TTS remains challenging due to discrepancies in phoneme vocabularies and variations in prosody and speaking style across languages. Existing approaches either train separate models for each language, which achieve high performance at the cost of increased computational resources, or use a unified model for multiple languages that struggles to capture fine-grained, language-specific style variations. In this work, we propose LanStyleTTS, a non-autoregressive, language-aware style adaptive TTS framework that standardizes phoneme representations and enables fine-grained, phoneme-level style control across languages. This design supports a unified multilingual TTS model capable of producing accurate and high-quality speech without the need to train language-specific models. We evaluate LanStyleTTS by integrating it with several state-of-the-art non-autoregressive TTS architectures. Results show consistent performance improvements across different model backbones. Furthermore, we investigate a range of acoustic feature representations, including mel-spectrograms and autoencoder-derived latent features. Our experiments demonstrate that latent encodings can significantly reduce model size and computational cost while preserving high-quality speech generation.

Figures (2)

• Figure 1: LanStyleTTS Architecture Overview
• Figure 2: Phoneme Embedding Visualization. (a) Alphabet embeddings. (b) Phoneme embeddings without language-specific style. (c) Phoneme embeddings with language-specific style..