---
title: "Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2610.40505v1"
permalink: "https://feed.astralor.com/items/itm_3ac7f4c31c9ebe0ab792b64e"
published: "2026-10-29T17:21:36.000Z"
score: 54
channel: "tech"
tags: ["Discovering Interpretable Acoustic","Features","Autoregressive Voice-Cloning TTS","Models Voice-cloning TTS","Voice-cloning TTS","TTS","SAEs","PCA"]
---

# Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models

## 速览

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 f…

## 关键点

- 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 quant…
- We focus on Qwen3-TTS and extend probing and steering to $5$ additional models to test generalization across architectures.

## 全文

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.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_3ac7f4c31c9ebe0ab792b64e