List your company

ResearchFormerly WizWand

Speech Reconstruction

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

LibriTTS clean (test)

4.644PESQ

67

Jul 7, 2026

LibriSpeech (test-clean)

4.23UT MOS

64

Jun 1, 2026

LibriSpeech clean (test)

4.31UTMOS Score

60

Jun 30, 2026

LibriTTS (test-other)

CleanCodec@12.5

4.17UTMOS

57

Jun 4, 2026

AISHELL-2 Chinese

MOSS-Audio-Tokenizer

0.93SIM

54

Mar 20, 2026

LibriSpeech English (test-clean)

MOSS-Audio-Tokenizer

0.97SIM

54

Mar 20, 2026

LibriTTS + ESC-50 clean-100 (test)

2.97PESQ

48

Jul 7, 2026

4.21PESQ

29

May 28, 2026

SeedTTS en (test)

0.021WER

21

Jun 11, 2026

2.99UTMOS

19

Feb 26, 2026

3.92UTMOS

19

Feb 26, 2026

Salmon Sentiment Consistency emotional 2025b (OOD)

2.9WER

18

Feb 26, 2026

Seed-TTS English

DashengTokenizer

4.125PESQ

17

May 28, 2026

H-Codec-2.0 (Large)

2.77PESQ

12

Feb 26, 2026

Open Track 2 (test)

1.15ScoreQ-ref

12

Feb 26, 2026

Open Track 1 (test)

1.36ScoreQ-ref

12

Feb 26, 2026

MLS non-English OOD (700 utterances (7 languages))

3.16UTMOS

10

Jun 30, 2026

ReSSInt laryngeal subjects (test)

Masked Multimodal Speech Synthesis Framework

40.5WER

9

Jun 9, 2026

LibriSpeech other (test)

6.79WER

9

Jun 1, 2026

MLS (Multilingual LibriSpeech) Non-English (test)

7.3WER

9

Mar 9, 2026

English Read by Japanese accented speech 2007 (OOD)

14.9WER

9

Feb 26, 2026

Japanese Versatile Speech unseen language speech 2019 (OOD)

4.6WER

9

Feb 26, 2026

Gigaspeech noisy speech 2021 (OOD)

9.7WER

9

Feb 26, 2026

LibriSpeech short splits (4s-10s) (test-clean)

5.12WER

8

Jun 23, 2026

Librispeech (test)

0.9STOI

8

Feb 26, 2026