OverFlow: Putting flows on top of neural transducers for better TTS

Benchmark Model Rank Results
text-to-speech-synthesis-on-ljspeechOverFlow#11Audio Quality MOS: 3.37Word Error Rate (WER): 2.30