Blow: a single-scale hyperconditioned flow for non-parallel raw-audio voice conversion

TL;DR

Blow is a novel single-scale hyperconditioned flow model that enables effective many-to-many non-parallel raw-audio voice conversion, outperforming existing methods in quality and flexibility.

Contribution

The paper introduces Blow, a new end-to-end flow-based model with hypernetwork conditioning for non-parallel voice conversion using raw audio.

Findings

Blow achieves equal or better performance than existing models.

The model requires less training data for effective conversion.

Ablation studies highlight the importance of its key components.

Abstract

End-to-end models for raw audio generation are a challenge, specially if they have to work with non-parallel data, which is a desirable setup in many situations. Voice conversion, in which a model has to impersonate a speaker in a recording, is one of those situations. In this paper, we propose Blow, a single-scale normalizing flow using hypernetwork conditioning to perform many-to-many voice conversion between raw audio. Blow is trained end-to-end, with non-parallel data, on a frame-by-frame basis using a single speaker identifier. We show that Blow compares favorably to existing flow-based architectures and other competitive baselines, obtaining equal or better performance in both objective and subjective evaluations. We further assess the impact of its main components with an ablation study, and quantify a number of properties such as the necessary amount of training data or the preference for source or target speakers.