Activation Steering for Accent-Neutralized Zero-Shot Text-To-Speech
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914374798540800 |
|---|---|
| author | Yang, Mu Hansen, John H. L. |
| author_facet | Yang, Mu Hansen, John H. L. |
| contents | Zero-shot Text-to-Speech (TTS) models can generate speech that captures both the voice timbre and accent of a reference speaker. However, disentangling these attributes remains challenging, as the output often inherits both the accent and timbre from the reference. In this study, we introduce a novel, post-hoc, and training-free approach to neutralize accent while preserving the speaker's original timbre, utilizing inference-time activation steering. We first extract layer-specific "steering vectors" offline, which are derived from the internal activation differences within the TTS model between accented and native speech. During inference, the steering vectors are applied to guide the model to produce accent-neutralized, timbre-preserving speech. Empirical results demonstrate that the proposed steering vectors effectively mitigate the output accent and exhibit strong generalizability to unseen accented speakers, offering a practical solution for accent-free voice cloning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_05977 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Activation Steering for Accent-Neutralized Zero-Shot Text-To-Speech Yang, Mu Hansen, John H. L. Audio and Speech Processing Zero-shot Text-to-Speech (TTS) models can generate speech that captures both the voice timbre and accent of a reference speaker. However, disentangling these attributes remains challenging, as the output often inherits both the accent and timbre from the reference. In this study, we introduce a novel, post-hoc, and training-free approach to neutralize accent while preserving the speaker's original timbre, utilizing inference-time activation steering. We first extract layer-specific "steering vectors" offline, which are derived from the internal activation differences within the TTS model between accented and native speech. During inference, the steering vectors are applied to guide the model to produce accent-neutralized, timbre-preserving speech. Empirical results demonstrate that the proposed steering vectors effectively mitigate the output accent and exhibit strong generalizability to unseen accented speakers, offering a practical solution for accent-free voice cloning. |
| title | Activation Steering for Accent-Neutralized Zero-Shot Text-To-Speech |
| topic | Audio and Speech Processing |
| url | https://arxiv.org/abs/2603.05977 |