AUTOVC: Zero-Shot Voice Style Transfer with Only Autoencoder Loss

TL;DR

AUTOVC introduces a novel autoencoder-based approach for zero-shot voice style transfer that achieves state-of-the-art results without requiring adversarial training, simplifying the process and improving perceptual quality.

Contribution

The paper presents a new autoencoder scheme with a designed bottleneck that enables distribution-matching style transfer and zero-shot voice conversion, outperforming existing methods.

Findings

Achieves state-of-the-art many-to-many voice conversion results.

First method to perform zero-shot voice conversion.

Simplifies training by avoiding GANs, using only self-reconstruction loss.

Abstract

Non-parallel many-to-many voice conversion, as well as zero-shot voice conversion, remain under-explored areas. Deep style transfer algorithms, such as generative adversarial networks (GAN) and conditional variational autoencoder (CVAE), are being applied as new solutions in this field. However, GAN training is sophisticated and difficult, and there is no strong evidence that its generated speech is of good perceptual quality. On the other hand, CVAE training is simple but does not come with the distribution-matching property of a GAN. In this paper, we propose a new style transfer scheme that involves only an autoencoder with a carefully designed bottleneck. We formally show that this scheme can achieve distribution-matching style transfer by training only on a self-reconstruction loss. Based on this scheme, we proposed AUTOVC, which achieves state-of-the-art results in many-to-many voice conversion with non-parallel data, and which is the first to perform zero-shot voice conversion.