WordVoice Adds Decoupled Word-Level Control to LLM-Based TTS

arXiv·medium signal

WordVoice introduces explicit, decoupled, multi-dimensional word-level control for LLM-based text-to-speech, moving beyond utterance-level prosody to per-word steering of speech attributes. Practical for builders wanting fine-grained TTS control (emphasis, style, emotion) that current LLM-TTS systems apply globally. The decoupling of control dimensions is the notable design choice.