Neural Vocoding / Waveform Synthesis on MUSDB18 HQ (Out-of-Distribution samples)
4.34Vocal SMOSBigVGAN-v2
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| BigVGAN-v2training_data=large-scale dataset including environmental sounds and instruments2024.08 | 4.34 | 4.34 | 4.34 | 4.3 | 4.31 | 4.33 | |
| PeriodWave-Turbo-BSteps=4, Sampling Method=Euler, Model Variant=B2024.08 | 4.34 | 4.3 | 4.2 | 4.27 | 4.29 | 4.28 | |
| BigVSAN2024.08 | 4.31 | 4.31 | 4.12 | 4.21 | 4.22 | 4.23 | |
| PeriodWave-L + FreeUSteps=16, Enhancement=FreeU, Model Variant=L2024.08 | 4.31 | 4.29 | 4.17 | 4.25 | 4.2 | 4.24 | |
| BigVGAN2024.08 | 4.3 | 4.34 | 4.06 | 4.21 | 4.24 | 4.23 | |
| PeriodWave-B + FreeUSteps=16, Enhancement=FreeU, Model Variant=B2024.08 | 4.3 | 4.28 | 4.09 | 4.16 | 4.21 | 4.21 | |
| PeriodWave-Turbo-LSteps=4, Sampling Method=Euler, Model Variant=L2024.08 | 4.3 | 4.35 | 4.22 | 4.29 | 4.32 | 4.3 | |
| PeriodWave-Turbo-SSteps=4, Sampling Method=Euler, Model Variant=S2024.08 | 4.28 | 4.34 | 4.18 | 4.23 | 4.2 | 4.25 | |
| Vocos2024.08 | 4.15 | 4.24 | 3.77 | 3.81 | 3.95 | 3.98 | |
| UnivNet-c322024.08 | 4.1 | 4.04 | 3.73 | 3.64 | 3.74 | 3.85 |