Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Audio Speech Recognition on LRS3
Loading...
0.7
WER
Whisper-Flamingo
0.536
1.643
2.75
3.857
Nov 4, 2024
Jan 27, 2025
Apr 21, 2025
Jul 14, 2025
Oct 6, 2025
Dec 29, 2025
Mar 23, 2026
WER
Updated 25d ago
Evaluation Results
Method
Method
Links
WER
Whisper-Flamingo
Labelled hours=680,000...
2026.02
0.7
Llama-AVSR
Preprocessing=Lip crop...
2026.03
0.79
Llama-AVSR
Labelled hours=680,000...
2026.02
0.8
USR 2.0
Labelled hours=656, Un...
2026.02
0.9
Auto-AVSR
Labelled hours=1,902,...
2024.11
1
Auto-AVSR
Labelled hours=3,448,...
2024.11
1
Auto-AVSR
Labelled hours=1,902,...
2026.02
1
Auto-AVSR
Labelled hours=3,448,...
2026.02
1
BRAVEn w/ ST
Labelled hours=433, Un...
2026.02
1.1
USR
Labelled hours=433, Un...
2024.11
1.2
USR
Labelled hours=433, Un...
2026.02
1.2
RAVEn w/ ST
Labelled hours=433, Un...
2024.11
1.4
RAVEn w/ ST
Labelled hours=433, Un...
2026.02
1.4
Qwen2.5-Omni
Preprocessing=Raw video
2026.03
3.63
HumanOmni-Speaker
Preprocessing=Raw video
2026.03
3.63
OLA
Preprocessing=Raw video
2026.03
4.7
RNN-T
Labelled hours=31,000,...
2024.11
4.8
RNN-T
Labelled hours=31,000,...
2026.02
4.8
Feedback
Search any
task
Search any
task