VCTK
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
VCTK noisy (test)
3.89NISQA-MOS
12
VCTK 24 kHz (test)
0.74LSD
11
VCTK 16 kHz target sampling rate 0.92 (test)
0.78LSD
11
VCTK
100Accdet
10
VCTK Google Pixel7
0.84LSD
10
VCTK Desktop
0.82LSD
10
VCTK 8 kHz to 44.1 kHz (test)
4.73VISQOL
10
VCTK 100 audio clips (unseen)
0.093MAE
10
VCTK (test)
99.3Accuracy
10
VCTK
4.63MOS
9
VCTK English Corpus with Unseen Speakers (out-of-domain)
4.117UTMOS
9
VCTK (test)
4.26nMOS
9
VCTK (OD)
4.5PESQ
9
VCTK
97.92Preference Score
8
VCTK (test)
4.4MOS
8
VCTK (unseen speakers)
4.37MOS
8
VCTK Clean
0.81LSD (8->16kHz)
7
VCTK (test)
0.78LSD
7
VCTK (test)
0.86LSD
7
VCTK-BWE BW=2K (test)
4.306WVMOS
7
VCTK 2 Speech
14.52SI-SDR
7
VCTK 4 kHz input sampling rate (test)
1WER
7
VCTK 2 kHz input sampling rate (test)
1WER
7
VCTK subset
2.36PESQ (WB)
7
VCTK++
90F1 Score
7