Tag
This paper investigates weakly-supervised ByT5 fine-tuning for stress-aware sentence-level Filipino grapheme-to-phoneme conversion, achieving significant improvements in phoneme and character error rates on a manually-corrected test set.
This paper presents FastThaiG2P, a sub-millisecond Thai grapheme-to-phoneme conversion tool for TTS pipelines, achieving 0.15 ms average latency on a 27k-utterance benchmark. The authors demonstrate it by training a StyleTTS 2 Thai TTS model on a phonemized 20-hour open dataset.