---
title: "Ruby-OPD: Enhancing Pronunciation Controllability and Accuracy for Flow-Matching Text-to-Speech via On-Policy Distillation"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.06403v1"
permalink: "https://feed.astralor.com/items/itm_7577b4bb4a8434275b34699b"
published: "2026-11-04T12:05:34.000Z"
score: 58
channel: "tech"
tags: ["Ruby-OPD","Enhancing Pronunciation Controllability","Accuracy","Flow-Matching Text-to-Speech","On-Policy Distillation Achieving","Japanese","TTS"]
---

# Ruby-OPD: Enhancing Pronunciation Controllability and Accuracy for Flow-Matching Text-to-Speech via On-Policy Distillation

## 速览

Achieving both high pronunciation accuracy and fine-grained controllability remains a central challenge in Japanese text-to-speech (TTS), as Japanese is an unsegmented language mixing logographic Kanji and phonographic …

## 关键点

- Achieving both high pronunciation accuracy and fine-grained controllability remains a central challenge in Japanese text-to-speech (TTS), as Japanese is an unsegmented language mixing logographic Kanji and phonographic …
- Traditional cascaded systems convert text into phoneme sequences via a front-end, ensuring pronunciation control but discarding lexical semantics and prosodic cues in surface orthography.
- Conversely, recent zero-shot TTS models directly consume raw orthographic text to improve naturalness, yet suffer from polyphone errors and lack explicit pronunciation control.

## 全文

Achieving both high pronunciation accuracy and fine-grained controllability remains a central challenge in Japanese text-to-speech (TTS), as Japanese is an unsegmented language mixing logographic Kanji and phonographic Kana with extensive polyphony. Traditional cascaded systems convert text into phoneme sequences via a front-end, ensuring pronunciation control but discarding lexical semantics and prosodic cues in surface orthography. Conversely, recent zero-shot TTS models directly consume raw orthographic text to improve naturalness, yet suffer from polyphone errors and lack explicit pronunciation control. To bridge this gap, we propose a two-stage framework for flow-matching TTS. First, we investigate Ruby text representations with orthography perturbation to jointly preserve surface forms and target pronunciations for reliable control. Second, we introduce On-Policy Distillation (OPD) on large-scale text-only corpora to enhance pronunciation accuracy on raw orthographic text without paired speech. On the Joyo-Kanji-Yomi benchmark, our model achieves 97.13% Accuracy, 2.02% Target PER, 0.70% Sentence PER, and 4.21% Sentence CER under raw text while retaining strong controllability.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_7577b4bb4a8434275b34699b