AI Paper Atlas

Research map / Benchmarks / LRS2

LRS2 leaderboard

19 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on LRS2. Best WER (%): 4.9 (video-SALMONN audio + video, ICML 2024).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

LRS2 · WER (%)

13 papers, each with its best reported number for its own method; lower is better.

#WER (%) ↓MethodPaperTopic
14.9video-SALMONN audio + video
Vicuna-v1.5 13B
video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models
ICML 2024 · Additional Results on Lip Read
Video-language understanding
26.6H2T-LoRA
T5-large
HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models
NeurIPS 2023 · Table 2
Language, safety and interpretability
314.56TTSF (without motion)Faces that Speak: Jointly Synthesising Talking Face and Speech from Text
CVPR 2024 · Table 3
Autoregressive and latent generation
417.8LipVoicerLipVoicer: Generating Speech from Silent Videos Guided by Lip Reading
ICLR 2024 · Performance comparison table,
Video generation
520.8VALLR (Ours)
Llama 3.2-3B
VALLR: Visual ASR Language Model for Lip Reading
ICCV 2025 · Table 2
Language, safety and interpretability
622.6Ours (expanded training data)Sub-Word Level Lip Reading With Visual Attention
CVPR 2022 · Table 1
Language, safety and interpretability
723.43TalkLip (g + c)Seeing What You Said: Talking Face Generation Guided by a Lip Reading Expert
CVPR 2023 · Table 1
Video generation
823.8Proposed Method
AV-HuBERT Large
Lip Reading for Low-resource Languages by Learning and Combining General Speech Knowledge and Language-specific Knowledge
ICCV 2023 · Table 2
Language, safety and interpretability
935.36LinProVSRLinProVSR: Linguistics-Knowledge Guided Progressive Disambiguation Network for Visual Speech Recognition
AAAI 2026 · Table 4
Cross-modal learning
1039.99HiCoDiT (video only)Hierarchical Codec Diffusion for Video-to-Speech Generation
CVPR 2026 · Table 2
Video generation
3 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

LRS2 · LSE-C

10 papers, each with its best reported number for its own method; higher is better.

#LSE-C ↑MethodPaperTopic
18.53TalkLip (g + c)Seeing What You Said: Talking Face Generation Guided by a Lip Reading Expert
CVPR 2023 · Table 1
Video generation
28.41Ours w/o FRAudio-driven Talking Face Generation with Stabilized Synchronization Loss
ECCV 2024 · Table 1
Video generation
37.95HiCoDiT (video only)Hierarchical Codec Diffusion for Video-to-Speech Generation
CVPR 2026 · Table 2
Video generation
47.925SyncTalkFace (audio-visual + visual-visual sync)
Wav2Lip
SyncTalkFace: Talking Face Generation with Precise Lip-Syncing via Audio-Lip Memory
AAAI 2022 · Quantitative results table
Video generation
57.874LipFormerLipFormer: High-Fidelity and Generalizable Talking Face Generation With a Pre-Learned Facial Codebook
CVPR 2023 · Table 1
Editing and style transfer
66.861Neural Dubber (Mel + PWG)Neural Dubber: Dubbing for Videos According to Scripts
NeurIPS 2021 · Multi-speaker AVD results tabl
Video generation
76.6LipVoicerLipVoicer: Generating Speech from Silent Videos Guided by Lip Reading
ICLR 2024 · Performance comparison table,
Video generation
85.785Wav2Lip + ViT + PrefAce
ViT-Base
PrefAce: Face-Centric Pretraining with Self-Structure Aware Distillation
AAAI 2024 · Table 4
Self-supervised and contrastive
95.686TTSFFaces that Speak: Jointly Synthesising Talking Face and Speech from Text
CVPR 2024 · Table 1
Autoregressive and latent generation
105.528Wav2Lip + ViT + MARLIN
ViT-B
MARLIN: Masked Autoencoder for Facial Video Representation LearnINg
CVPR 2023 · Lip Synchronization table
Video segmentation and tracking

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.