Audio Generation on LibriTTS (dev)
1.3647M-STFTHiFi-GAN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HiFi-GANversion=V12024.01 | 1.3647 | 0.16 | 0.9309 | 2.911 | |
| WaveGlow-256Params (M)=99.432024.08 | 1.3099 | 0.1485 | 0.9378 | 3.138 | |
| PeriodWaveParams (M)=29.8, Inference steps=16 steps, Training steps=1M2024.08 | 1.2129 | 0.0762 | 0.9652 | 4.224 | |
| WaveFlow-128*Params (M)=22.582024.08 | 1.112 | 0.1416 | 0.941 | 3.027 | |
| HiFi-GANsize=BASE2024.01 | 1.0269 | 0.123 | 0.9523 | 3.5485 | |
| PeriodWave + FreeUParams (M)=29.8, Inference steps=16 steps, Training steps=1M2024.08 | 1.0269 | 0.0765 | 0.9651 | 4.248 | |
| HiFi-GAN (V1)Params (M)=14.012024.08 | 1.0017 | 0.1565 | 0.93 | 2.947 | |
| PeriodWave-MBParams (M)=37.08x4, Inference steps=16 steps, Training steps=1M2024.08 | 0.9729 | 0.0704 | 0.9678 | 4.262 | |
| EVA-GANsize=BASE2024.01 | 0.9485 | 0.0942 | 0.9658 | 4.033 | |
| UnivNet-c32configuration=c322024.01 | 0.8959 | 0.1333 | 0.9444 | 3.2566 | |
| UnivNet-c32Params (M)=14.872024.08 | 0.8947 | 0.1305 | 0.9347 | 3.284 | |
| BigVGANsize=BASE2024.01 | 0.8788 | 0.1287 | 0.9459 | 3.519 | |
| BigVGAN-baseParams (M)=14.01, Training steps=5M2024.08 | 0.8788 | 0.1287 | 0.9459 | 3.519 | |
| Vocos2024.01 | 0.858 | 0.1103 | 0.9555 | 3.6328 | |
| VocosParams (M)=13.532024.08 | 0.8544 | 0.1113 | 0.947 | 3.615 | |
| BigVGANsize=BIG2024.01 | 0.7997 | 0.1018 | 0.9598 | 4.027 | |
| BigVGANParams (M)=112.4, Training steps=5M2024.08 | 0.7997 | 0.1018 | 0.9598 | 4.027 | |
| EVA-GANsize=BIG2024.01 | 0.7982 | 0.0751 | 0.9745 | 4.3536 |