VoiceGuider: Enhancing Out-of-Domain Performance in Parameter-Efficient Speaker-Adaptive Text-to-Speech via Autoguidance

TL;DR

VoiceGuider is a novel parameter-efficient TTS system that uses autoguidance to significantly improve out-of-domain speaker adaptation, narrowing the performance gap with full-finetuned models.

Contribution

It introduces autoguidance into parameter-efficient speaker adaptation for TTS, achieving robust out-of-domain performance improvements.

Findings

Enhanced out-of-domain speaker adaptation performance

Autoguidance strategy outperforms other methods

Robust adaptation on extreme out-of-domain data

Abstract

When applying parameter-efficient finetuning via LoRA onto speaker adaptive text-to-speech models, adaptation performance may decline compared to full-finetuned counterparts, especially for out-of-domain speakers. Here, we propose VoiceGuider, a parameter-efficient speaker adaptive text-to-speech system reinforced with autoguidance to enhance the speaker adaptation performance, reducing the gap against full-finetuned models. We carefully explore various ways of strengthening autoguidance, ultimately finding the optimal strategy. VoiceGuider as a result shows robust adaptation performance especially on extreme out-of-domain speech data. We provide audible samples in our demo page.