Metraipolis › archive › dc_7fb97c831a73
Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models — arXiv:2610.40505
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.40505
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T06:17:27.588Z
- SHA-256 of the archived text
d85faf6baa6897d4076f5a4a47ab14328ca29ee7105e8fac48f76fcf8405e084- Stable URL of this page
https://metraipolis.com/documento/dc_7fb97c831a73
How to cite
Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models — arXiv:2610.40505 — arXiv (en), https://arxiv.org/abs/2610.40505 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7fb97c831a73 (Metraipolis archive dc_7fb97c831a73, sha256 d85faf6baa6897d4076f5a4a47ab14328ca29ee7105e8fac48f76fcf8405e084)
BibTeX
@misc{dc7fb97c831a73,
title = {Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models — arXiv:2610.40505},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.40505},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7fb97c831a73; sha256 d85faf6baa6897d4076f5a4a47ab14328ca29ee7105e8fac48f76fcf8405e084},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.40505. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Voice-cloning TTS models synthesize an unseen speaker's voice from brief speech prompts, but it remains unclear which acoustic properties they represent and use. We introduce a feature-based interpretability framework for autoregressive voice-cloning TTS models, analyzing features from sparse autoencoders (SAEs), PCA, and random directions via sparse probing, steering, and quantitative auto-interpretability. We focus on Qwen3-TTS and extend probing and steering to $5$ additional models to test generalization across architectures. These models encode concepts like pitch, energy, noise, and accent, but sparse prediction and steering are effective only for some concepts. Strong probing does not always translate to steering or vice-versa. SAEs are useful for sparse probing of human-labelled concepts and novel-concept discovery, while PCA is competitive for dense probing and steering. Auto-interpretability finds $684$ automatically validated SAE features in Qwen3-TTS covering speaker clusters, speaking style, and recording conditions beyond known concepts. Overall, we perform a comprehensive interpretability study of autoregressive voice-cloning TTS models. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Audio and Speech Processing (eess.AS) ; Computation and Language (cs.CL); Machine Learning (cs.LG); Sound (cs.SD). Pagina del lavoro: https://arxiv.org/abs/2610.40505 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-31T06:07:23.423Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).