Modeling Multi-speaker Latent Space to Improve Neural TTS: Quick Enrolling New Speaker and Enhancing Premium Voice

TL;DR

This paper introduces a multi-speaker neural TTS model that effectively adapts to new speakers with minimal data and enhances premium voices by leveraging multi-speaker data for improved naturalness and similarity.

Contribution

The study proposes a multi-speaker latent space approach in neural TTS that enables quick speaker adaptation and premium voice enhancement using limited data and multi-speaker information.

Findings

Achieves MOS of 4.16 for new speakers with less than 5 minutes of data.

Attains MOS of 4.5 for premium voices on out-of-domain texts.

Outperforms single speaker models in naturalness and similarity metrics.

Abstract

Neural TTS has shown it can generate high quality synthesized speech. In this paper, we investigate the multi-speaker latent space to improve neural TTS for adapting the system to new speakers with only several minutes of speech or enhancing a premium voice by utilizing the data from other speakers for richer contextual coverage and better generalization. A multi-speaker neural TTS model is built with the embedded speaker information in both spectral and speaker latent space. The experimental results show that, with less than 5 minutes of training data from a new speaker, the new model can achieve an MOS score of 4.16 in naturalness and 4.64 in speaker similarity close to human recordings (4.74). For a well-trained premium voice, we can achieve an MOS score of 4.5 for out-of-domain texts, which is comparable to an MOS of 4.58 for professional recordings, and significantly outperforms single speaker result of 4.28.