Controllable Generation

UDDETTS: Unifying Discrete and Dimensional Emotions for Controllable Emotional Text-to-Speech
A universal LLM-based TTS framework unifying discrete labels and the continuous Arousal–Dominance–Valence space, with semi-supervised training and an ADV predictor for fine-grained, interpretable emotion control.
Jiaxuan Liu
PDF Code Project
DiffStyleTTS: Diffusion-based Hierarchical Prosody Modeling for Text-to-Speech with Diverse and Controllable Styles
A diffusion-based acoustic model with hierarchical prosody modeling and improved classifier-free guidance, jointly enabling diverse and controllable prosody for TTS.
Jiaxuan Liu, Zhaoci Liu, Yajun Hu, Yingying Gao, Shilei Zhang, Zhenhua Ling
PDF Project