Whispered-to-Normal speech conversion on wTIMIT
3.9WER (Normal Target)Normal Speech
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Normal Speech2026.03 | 3.9 | 9.5 | — | 3.21 | |
| Real Whisper2026.03 | 8.3 | 26.5 | — | 1.21 | |
| FlowW2NTraining Data=Synthetic only2026.03 | 14 | 27.9 | 70 | 2.91 | |
| QuickVCTraining Data=Real pairs2026.03 | 26 | 39.1 | 72 | 3.04 | |
| DistillW2N-TTraining Data=Real pairs + Distill2026.03 | 37.5 | 61.8 | 64 | 3.02 | |
| DistillW2NTraining Data=Real pairs + Distill2026.03 | 110 | 205 | 61.8 | 2.82 |