WaveGlow: A Flow-Based Generative Network for Speech Synthesis

Use a generative flow network to generate audio samples conditioned on an MFCC. Gaussian noise goes in, audio samples come out.