List your company

ResearchFormerly WizWand

Speech Reconstruction

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

LibriTTS 16 kHz (test)

3.56ViSQOL

7

May 26, 2026

LibriTTS (test)

4.16PESQ

7

Feb 26, 2026

2.36PESQ (WB)

7

Feb 26, 2026

GRID (speaker-dependent)

Proposed Method

0.738STOI

7

Feb 26, 2026

VCTK 48 kHz (test)

3.62ViSQOL

6

May 26, 2026

Full-band SR 48kHz

Re. 48kHz HuBERT Base

85.93STOI

6

Mar 25, 2026

Full-band SR 24kHz

Re. 24kHz HuBERT Base

89.34STOI

6

Mar 25, 2026

Full-band SR 22.05kHz

94.38STOI

6

Mar 25, 2026

Full-band SR 16kHz

90.26STOI

6

Mar 25, 2026

Combined HVAC and VCTK 500 Hz to 8 kHz upsampling 8.33-hour (test)

1.24LSD

6

Mar 10, 2026

Combined HVAC and VCTK 1 kHz to 8 kHz upsampling 8.33-hour (test)

1.15LSD

6

Mar 10, 2026

Combined HVAC and VCTK 2 kHz to 8 kHz upsampling 8.33-hour (test)

1LSD

6

Mar 10, 2026

LRAC Challenge Track 1 (Multi-talkers) 2025 (test)

1.26DMOS

6

Feb 26, 2026

LRAC Challenge Track 1 (Noisy) 2025 (test)

4.44DMOS

6

Feb 26, 2026

LRAC Challenge Track 1 (Clean) 2025 (test)

81.75MUSHRA Score

6

Feb 26, 2026

LibriSpeech and LJSpeech

98.08MUSHRA Score

6

Feb 26, 2026

4.42MCD

6

Feb 26, 2026

0WER

5

Jun 4, 2026

AISHELL-3 (OOD)

0CER

5

Jun 4, 2026

0WER

5

Jun 4, 2026

1.9WER

5

Jun 1, 2026

English dataset 2 kHz sampling

1.01LSD

5

Mar 9, 2026

English dataset 1 kHz sampling

1.19LSD

5

Mar 9, 2026

English dataset 500 Hz sampling

1.29LSD

5

Mar 9, 2026

LibriSpeech 30% packet loss (test)

2.63PESQ

5

Feb 26, 2026