BigVGAN: A Universal Neural Vocoder with Large-Scale Training

Benchmark Model Rank Results
speech-synthesis-on-librittsBigVGAN-v2#2PESQ: 4.362M-STFT: 0.7026MCD: 0.2903Periodicity: 0.0593
speech-synthesis-on-librittsBigVGAN#9PESQ: 4.027M-STFT: 0.7997MCD: 0.3745Periodicity: 0.1018
speech-synthesis-on-librittsBigVGAN-base#11PESQ: 3.519M-STFT: 0.8788MCD: 0.4564Periodicity: 0.1287