Neural Vocoding on LibriTTS (test)
4.269PESQBigVGAN
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| BigVGANTraining steps=5M steps, Source=Open-sourced model checkpoint, #Param. (M)=112.39, #MACs (Giga/5s)=454.082026.03 | 4.269 | 4.963 | 0.736 | 1.871 | 0.079 | 96.7 | 24.281 | 3.744 | — | |
| RNDVoC-shared#Param. (M)=3.14, #MACs (Giga/5s)=37.202026.03 | 4.226 | 4.915 | 0.746 | 1.914 | 0.074 | 97 | 23.966 | 3.657 | — | |
| RNDVoC-nonshared#Param. (M)=9.48, #MACs (Giga/5s)=28.292026.03 | 4.085 | 4.873 | 0.798 | 2.251 | 0.09 | 96.1 | 27.041 | 3.607 | — | |
| BigVGANTraining steps=1M steps, Source=Open-sourced model checkpoint, #Param. (M)=112.39, #MACs (Giga/5s)=454.082026.03 | 4.027 | — | 0.8 | — | 0.102 | 96 | — | — | — | |
| ComVo (large)Params. (M)=114.562026.03 | 3.8831 | — | 0.8443 | — | 0.0871 | 96.29 | — | 3.7337 | — | |
| BigVGAN-baseTraining steps=5M steps, Source=Open-sourced model checkpoint, #Param. (M)=14.01, #MACs (Giga/5s)=166.412026.03 | 3.841 | 4.907 | 0.829 | 2.431 | 0.107 | 95.4 | 32.541 | 3.617 | — | |
| Vocos (large)Params. (M)=114.512026.03 | 3.8362 | — | 0.8625 | — | 0.0933 | 95.96 | — | 3.6923 | — | |
| RNDVoC-Lite#Parm. (M)=0.71, Inference Speed (CPU)=0.0261(38.33×), Inference Speed (GPU)=0.0020(502×)2026.03 | 3.834 | 4.874 | — | — | — | — | — | — | — | |
| BigVGAN (large)Params. (M)=112.412026.03 | 3.8197 | — | 0.8644 | — | 0.0888 | 96.07 | — | 3.5489 | — | |
| VocosSource=Open-sourced model checkpoint, #Param. (M)=13.53, #MACs (Giga/5s)=6.352026.03 | 3.615 | 4.879 | 0.854 | 3.105 | 0.115 | 94.8 | 35.584 | 3.566 | — | |
| BigVGAN-baseTraining steps=1M steps, Source=Open-sourced model checkpoint, #Param. (M)=14.01, #MACs (Giga/5s)=166.412026.03 | 3.519 | — | 0.879 | — | 0.129 | 94.6 | — | — | — | |
| RNDVoC-UltraLite#Parm. (M)=0.08, Inference Speed (CPU)=0.0253(39.46×), Inference Speed (GPU)=0.0018(543×)2026.03 | 3.499 | 4.804 | — | — | — | — | — | — | — | |
| Avocodo#Param. (M)=14.01, #MACs (Giga/5s)=166.472026.03 | 3.217 | 4.762 | 1.113 | 4.314 | 0.161 | 91.3 | 51.6 | 3.358 | — | |
| HiFiGAN-V1#Param. (M)=14.01, #MACs (Giga/5s)=166.412026.03 | 3.056 | 4.721 | 1.104 | 4.284 | 0.167 | 92.1 | 52.529 | 3.511 | — | |
| APNet#Param. (M)=73.33, #MACs (Giga/5s)=33.922026.03 | 2.897 | 4.666 | 1.273 | 4.15 | 0.159 | 92.7 | 39.663 | 2.416 | — | |
| iSTFTNet-V1#Param. (M)=13.33, #MACs (Giga/5s)=117.302026.03 | 2.88 | 4.655 | 1.157 | 4.48 | 0.167 | 91.8 | 53.072 | 3.464 | — | |
| APNet2#Param. (M)=31.52, #MACs (Giga/5s)=14.792026.03 | 2.834 | 4.582 | 1.143 | 4.153 | 0.153 | 92.3 | 46.373 | 2.956 | — | |
| HiFiGAN-V2#Parm. (M)=0.92, Inference Speed (CPU)=0.0288(34.72×), Inference Speed (GPU)=0.0014(704×)2026.03 | 2.195 | 4.37 | — | — | — | — | — | — | — | |
| GTParams. (M)=-2026.03 | — | — | — | — | — | — | — | 3.8712 | — | |
| GTBackbone=BigVGAN-base2026.03 | — | — | — | — | — | — | — | — | 4.804 | |
| LSGANBackbone=BigVGAN-base2026.03 | — | — | — | — | — | — | — | — | 4.526 | |
| RAFBackbone=BigVGAN-base2026.03 | — | — | — | — | — | — | — | — | 4.592 |