Audio Super-resolution on VCTK 16 kHz input (test)
0.813LSDStructure-aware regularization
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Structure-aware regularizationInput SR=16 kHz, Target SR=48 kHz2026.05 | 0.813 | 0.217 | 1.047 | 19.18 | |
| Self-adaptive timestep samplingInput SR=16 kHz, Target SR=48 kHz2026.05 | 0.838 | 0.206 | 1.049 | 19.09 | |
| Decayed SSL guidanceInput SR=16 kHz, Target SR=48 kHz2026.05 | 0.843 | 0.219 | 1.053 | 19.54 | |
| NVSRInput SR=16 kHz, Target SR=48 kHz2026.05 | 0.863 | 0.232 | 1.042 | 18.53 | |
| Uniform baselineInput SR=16 kHz, Target SR=48 kHz2026.05 | 0.878 | 0.231 | 1.079 | 19.17 | |
| AudioSRInput SR=16 kHz, Target SR=48 kHz2026.05 | 1.108 | 0.473 | 1.307 | 18.71 |