研究人员开发了GLASS，一个用于零样本文本到语音（TTS）系统声学风格控制的新框架。与以往将说话人身份与韵律交织在一起的方法不同，GLASS将语速和音高之类的属性视为独立的、由奖励定义的控制方向。通过使用GRPO训练轻量级LoRA适配器，该系统允许通过线性算术进行可组合的风格调整，从而在不重新训练核心TTS模型的情况下实现语音特征的定向转变。 AI
影响 能够对合成语音特征进行更精细、更灵活的控制，有可能提高TTS的自然度和用户体验。
排序理由 该集群包含一篇详细介绍文本到语音合成新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2606.05889v1 Announce Type: cross Abstract: We propose GLASS, a framework for composable acoustic style control in zero-shot autoregressive text-to-speech (TTS) that learns controls from post-generation rewards rather than style labels. In zero-shot TTS, a speaker prompt of…
We propose GLASS, a framework for composable acoustic style control in zero-shot autoregressive text-to-speech (TTS) that learns controls from post-generation rewards rather than style labels. In zero-shot TTS, a speaker prompt often entangles speaker identity with prosodic attri…
We propose GLASS, a framework for composable acoustic style control in zero-shot autoregressive text-to-speech (TTS) that learns controls from post-generation rewards rather than style labels. In zero-shot TTS, a speaker prompt often entangles speaker identity with prosodic attri…