Whispered-to-Normal speech conversion on CHAINS
5.9WER (N)Normal Speech
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Normal Speech2026.03 | 5.9 | 11.9 | — | 3.06 | |
| Real Whisper2026.03 | 11.1 | 31.6 | — | 1.5 | |
| FlowW2NTraining Data=Synthetic only2026.03 | 17.9 | 28.8 | 63.1 | 2.93 | |
| QuickVCTraining Data=Real pairs2026.03 | 24.1 | 36.8 | 66.8 | 2.99 | |
| DistillW2N-TTraining Data=Real pairs + Distill2026.03 | 37.3 | 54.6 | 60.4 | 2.81 | |
| DistillW2NTraining Data=Real pairs + Distill2026.03 | 105 | 201 | 57.6 | 2.2 |