Learning latent representations for style control and transfer in end-to-end speech synthesis

TL;DR

This paper introduces a VAE-based approach for unsupervised learning of speaking style representations in end-to-end speech synthesis, enabling effective style control and transfer, outperforming previous models like GST.

Contribution

The paper presents a novel VAE framework for style learning in speech synthesis, improving style disentanglement, control, and transfer capabilities.

Findings

Effective style control demonstrated

Outperforms GST in style transfer tests

Addresses KL collapse in training

Abstract

In this paper, we introduce the Variational Autoencoder (VAE) to an end-to-end speech synthesis model, to learn the latent representation of speaking styles in an unsupervised manner. The style representation learned through VAE shows good properties such as disentangling, scaling, and combination, which makes it easy for style control. Style transfer can be achieved in this framework by first inferring style representation through the recognition network of VAE, then feeding it into TTS network to guide the style in synthesizing speech. To avoid Kullback-Leibler (KL) divergence collapse in training, several techniques are adopted. Finally, the proposed model shows good performance of style control and outperforms Global Style Token (GST) model in ABX preference tests on style transfer.