Waveform Generation on LJSpeech
4.3894UTMOSPeriodWave-Turbo
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PeriodWave-TurboTraining Steps=0.1M, Params (M)=29.73, inference steps=4, ODE solver=Euler2024.08 | 4.3894 | 0.8352 | 4.418 | 0.0572 | 0.9786 | 13.473 | |
| Ground Truth2024.08 | 4.3804 | — | — | — | — | — | |
| PeriodWave-TurboTraining Steps=10K, Params (M)=29.73, inference steps=4, ODE solver=Euler2024.08 | 4.3779 | 0.888 | 4.387 | 0.0624 | 0.9769 | 14.186 | |
| PeriodWave-TurboTraining Steps=5K, Params (M)=29.73, inference steps=4, ODE solver=Euler2024.08 | 4.3716 | 0.8969 | 4.344 | 0.0588 | 0.9783 | 13.662 | |
| PeriodWave+FreeUTraining Steps=1.0M, Params (M)=29.73, inference steps=16, ODE solver=Midpoint2024.08 | 4.3578 | 1.1132 | 4.293 | 0.0749 | 0.9701 | 15.753 | |
| PeriodWave-TurboTraining Steps=1K, Params (M)=29.73, inference steps=4, ODE solver=Euler2024.08 | 4.3563 | 0.9253 | 4.349 | 0.064 | 0.9766 | 14.943 | |
| PeriodWaveTraining Steps=1.0M, Params (M)=29.73, inference steps=16, ODE solver=Midpoint2024.08 | 4.3243 | 1.1464 | 4.288 | 0.0744 | 0.9704 | 15.042 | |
| BigVGAN-v2Training Steps=3.0M, Params (M)=112.42024.08 | 4.311 | 0.8826 | 4.262 | 0.0663 | 0.976 | 17.325 | |
| HiFi-GAN (V1)Training Steps=2.5M, Params (M)=14.012024.08 | 4.2691 | 1.0341 | 3.646 | 0.1064 | 0.9584 | 26.839 | |
| BigVGANTraining Steps=5.0M, Params (M)=112.42024.08 | 4.2172 | 0.9369 | 4.21 | 0.0782 | 0.9713 | 19.019 | |
| BigVGAN-baseTraining Steps=5.0M, Params (M)=14.012024.08 | 4.1986 | 1.0046 | 3.868 | 0.1054 | 0.9597 | 25.142 | |
| PriorGradTraining Steps=3.0M, Params (M)=2.61, steps=502024.08 | 3.6282 | 1.2784 | 3.918 | 0.0879 | 0.9661 | 17.728 |