End-to-End Zero-Shot Voice Conversion with Location-Variable Convolutions

TL;DR

This paper introduces LVC-VC, an end-to-end zero-shot voice conversion model using location-variable convolutions to directly synthesize speech, achieving balanced voice transfer and intelligibility.

Contribution

It presents a novel end-to-end zero-shot voice conversion approach with location-variable convolutions that jointly model conversion and synthesis without separate vocoders.

Findings

Balanced voice style transfer and speech intelligibility

Effective joint modeling of conversion and synthesis

Outperforms several baseline models

Abstract

Zero-shot voice conversion is becoming an increasingly popular research topic, as it promises the ability to transform speech to sound like any speaker. However, relatively little work has been done on end-to-end methods for this task, which are appealing because they remove the need for a separate vocoder to generate audio from intermediate features. In this work, we propose LVC-VC, an end-to-end zero-shot voice conversion model that uses location-variable convolutions (LVCs) to jointly model the conversion and speech synthesis processes. LVC-VC utilizes carefully designed input features that have disentangled content and speaker information, and it uses a neural vocoder-like architecture that utilizes LVCs to efficiently combine them and perform voice conversion while directly synthesizing time domain audio. Experiments show that our model achieves especially well balanced performance between voice style transfer and speech intelligibility compared to several baselines.