Speech Super-Resolution on VCTK Clean
0.81LSD (8->16kHz)TF-Restormer (off)
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| TF-Restormer (off)streaming=false, trained specifically for super-resolution=true2025.09 | 0.81 | 4.42 | 0.82 | 4.58 | 0.82 | 4.4 | 0.81 | 4.57 | |
| NVSRfixed output sampling rates=true2025.09 | 0.83 | 4.15 | 0.89 | 4.24 | 0.94 | 4.16 | — | — | |
| TF-Restormer (off)streaming=false2025.09 | 0.89 | 4.53 | 0.95 | 4.61 | 1.01 | 4.54 | 0.97 | 4.62 | |
| AP-BWEfixed output sampling rates=true2025.09 | 0.9 | 4.2 | 0.86 | 4.34 | 0.88 | 4.26 | 0.85 | 4.33 | |
| VoiceFixerfixed output sampling rates=true2025.09 | 1.05 | 4.2 | 1.05 | 4.27 | 1.06 | 4.21 | — | — | |
| Frepainterfixed output sampling rates=true2025.09 | 1.33 | 3.94 | 1.4 | 3.79 | 1.37 | 3.71 | 1.31 | 4.01 | |
| Input2025.09 | 2.53 | 3.78 | 2.91 | 3.78 | 3.44 | 3.78 | 3.17 | 4.4 |