New Japanese text-to-speech model reads kanji correctly

Original: Ruby-OPD: Enhancing Pronunciation Controllability and Accuracy for Flow-Matching Text-to-Speech via On-Policy Distillation

Source: arxiv.org ↗