Skip to content
arXiv · Audio and Speech· Shiao Zhu, Lianbo Liu, Sizhen Lyu, Yuzhe Wang, Sheng Li, Takahiro Shinozaki·· 1d agoAI score24

Beyond Speech Captions:面向对话式 TTS 的语音奖励风格规划

Beyond Speech Captions: Speech-Rewarded Style Planning for Conversational Text-to-Speech

AI brief

论文提出 Speech-Rewarded Style Planning(SRSP),通过冻结的下游 TTS 模型训练文本风格规划器,用目标语音 token 的 teacher-forced likelihood 作为奖励,并以 GRPO 优化候选指令。

Source: arXiv · Audio and Speech · arxiv.org