Reproduce “EmpDPO: Enabling Emphasis Control in TTS via Direct Preference Optimization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haonan Chen, Yunyun Wang, Megan Wei, Zeyu Jin
The authors introduce EmpDPO, a text-to-speech system that emphasizes a selected word without changing the text. A multilingual detector labels examples and scores generated speech, enabling direct preference optimization with corpus replay. Across six languages, EmpDPO gains 0.136 ± 0.005 in detector score versus 0.069 for SFT. The quality-preserving result is limited to the tested setting and corpus replay.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.