Demo Of VoiceFixer
Voicefixer aims at the restoration of degraded human speech. It can handle noise, reveberation, low resolution (2kHz~44.1kHz) and clipping (0.1-1.0 threshold) effect within one model.
Built by Haohe Liu on Sep.14.2021 in Shanghai. With wild typhoon Chanthu outside.
1. About
2. Testset Data Demos
2.1 General Speech Restoration (ALL-GSR)
| Speaker info |
|
|
|
|
|
|
|
| 33_simulated |
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
| 127_simulated |
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
2.2 Speech Super Resolution (SR)
2.2.1 From 4kHz to 44.1kHz
| Speaker info |
|
|
|
|
|
|
|
|
| p361_056 |
|
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
|
| p374_098 |
|
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
|
2.2.2 From 8kHz to 44.1kHz
| Speaker info |
|
|
|
|
|
|
|
|
| p360_002 |
|
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
|
| p361_002 |
|
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
|
2.2.3 From 24kHz to 44.1kHz
| Speaker info |
|
|
|
|
|
|
|
|
| p360_002 |
|
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
|
| p361_002 |
|
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
|
2.3 Speech Enhancement (DENOISE)
| Speaker info |
|
|
|
|
|
|
| p232_005 |
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
| p257_008 |
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
2.4 Speech Dereverberation (DEREV)
| Speaker info |
|
|
|
|
|
|
| p361_001 |
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
| p363_004 |
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
2.5 Speech Declipping (DECLI)
Clipping threshold 0.1
| Speaker info |
|
|
|
|
|
|
|
| p360_001 |
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
2.5.1 Clipping threshold 0.25
| Speaker info |
|
|
|
|
|
|
|
| p360_001 |
|
|
|
|
|
|
|
| Spectrogram |
|
|
|
|
|
|
|
3.1 Comparison between enhancement only SSR, Regression based GSR and VoiceFixer GSR
| Speaker info |
|
|
|
|
| A recording of my voice. |
|
|
|
|
| Spectrogram |
|
|
|
|
| TV news interviews. |
|
|
|
|
| Spectrogram |
|
|
|
|
| Chinese Youtuber. |
|
|
|
|
| Spectrogram |
|
|
|
|
3.2 More demos
