List your company

ResearchFormerly WizWand

Lip-to-Speech Synthesis

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

LRS2-BBC (test)

4.166UTMOS

7

Feb 26, 2026

LRS3-TED (test)

4.21UTMOS

7

Feb 26, 2026

Lip2Wav 1.0 (test)

Actual Human Speech

4.82Intelligibility

5

Feb 26, 2026

Lip-to-speech synthesis Subjective (test)

4.02Naturalness

3

Feb 26, 2026

LRS3 unseen speaker protocol (test)

LipVoicerofficial

21.9WER

3

Feb 26, 2026

25.15GF0

2

Feb 26, 2026

0.342WER

2

Feb 26, 2026