BigVSAN: Enhancing GAN-based Neural Vocoders with Slicing Adversarial Network

Benchmark Model Rank Results
speech-synthesis-on-librittsBigVSAN (w/ snakebeta)#6PESQ: 4.120M-STFT: 0.7992MCD: 0.4129Periodicity: 0.0924
speech-synthesis-on-librittsBigVSAN#7PESQ: 4.116M-STFT: 0.7881MCD: 0.3381Periodicity: 0.0935