Scholarlink

Pretrained phoneme-level models such as Phoneme-Level BERT and XPhoneBERT have shown promising results in enhancing prosody and expressiveness in English TTS systems. However, their effectiveness in less-studied languages with different prosodic characteristics—such as Czech—remains underexplored. This paper investigates their applicability in Czech text-to-speech synthesis by evaluating PL-BERT within the StyleTTS 2 framework and XPhoneBERT within the VITS architecture. We conduct experiments under both high- and low-resource conditions using professionally read Czech news-style speech to determine the benefits of these pretrained phoneme-level models in Czech speech synthesis and to compare them to each other.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluating Phoneme-Level Pretraining in Czech Text-to-Speech Synthesis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Lukáš Vladař,
  • Jindřich Matoušek,
  • Jan Lehečka,
  • Markéta Řezáčková

摘要

Pretrained phoneme-level models such as Phoneme-Level BERT and XPhoneBERT have shown promising results in enhancing prosody and expressiveness in English TTS systems. However, their effectiveness in less-studied languages with different prosodic characteristics—such as Czech—remains underexplored. This paper investigates their applicability in Czech text-to-speech synthesis by evaluating PL-BERT within the StyleTTS 2 framework and XPhoneBERT within the VITS architecture. We conduct experiments under both high- and low-resource conditions using professionally read Czech news-style speech to determine the benefits of these pretrained phoneme-level models in Czech speech synthesis and to compare them to each other.