AccentBox: Towards High-Fidelity Zero-Shot Accent Generation
This addresses the issue of generating high-fidelity accents in speech synthesis for applications requiring diverse or unseen accents, representing an incremental improvement over existing zero-shot TTS methods.
The paper tackles the problem of low accent fidelity and control in zero-shot text-to-speech models by proposing a two-stage pipeline for zero-shot accent generation, achieving a state-of-the-art 0.56 f1 score on accent identification for unseen speakers and higher accent fidelity in generation tasks.
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.