AccentBox: Towards High-Fidelity Zero-Shot Accent Generation

TL;DR

This paper introduces AccentBox, a two-stage zero-shot accent generation system that improves accent fidelity and control in TTS by combining accent identification and speaker-agnostic accent embeddings.

Contribution

It unifies FAC, accented TTS, and ZS-TTS into a novel pipeline with state-of-the-art accent identification and enhanced accent fidelity in zero-shot scenarios.

Findings

Achieves 0.56 F1 score on unseen speakers for accent identification.

Outperforms previous methods in accent fidelity for zero-shot accent generation.

Enables generation of unseen accents with high fidelity.

Abstract

While recent Zero-Shot Text-to-Speech (ZS-TTS) models have achieved high naturalness and speaker similarity, they fall short in accent fidelity and control. To address this issue, we propose zero-shot accent generation that unifies Foreign Accent Conversion (FAC), accented TTS, and ZS-TTS, with a novel two-stage pipeline. In the first stage, we achieve state-of-the-art (SOTA) on Accent Identification (AID) with 0.56 f1 score on unseen speakers. In the second stage, we condition a ZS-TTS system on the pretrained speaker-agnostic accent embeddings extracted by the AID model. The proposed system achieves higher accent fidelity on inherent/cross accent generation, and enables unseen accent generation.