Papers › SyncVSR: Data-Efficient Visual Speech Recognition with End-to-End Crossmodal Audio...

SyncVSR: Data-Efficient Visual Speech Recognition with End-to-End Crossmodal Audio Token Synchronization

18 Jun 2024arXiv:2406.12233archive 2025-07-28

Young Jin Ahn, Jungwoo Park, Sangha Park, Jonghyun Choi, Kee-Eung Kim

Visual Speech Recognition (VSR) stands at the intersection of computer vision and speech recognition, aiming to interpret spoken content from visual cues. A prominent challenge in VSR is the presence of homophenes-visually similar lip gestures that represent different phonemes. Prior approaches have sought to distinguish fine-grained visemes by aligning visual and auditory semantics, but often fell short of full synchronization. To address this, we present SyncVSR, an end-to-end learning framework that leverages quantized audio for frame-level crossmodal supervision. By integrating a projection layer that synchronizes visual representation with acoustic data, our encoder learns to generate discrete audio tokens from a video sequence in a non-autoregressive manner. SyncVSR shows versatility across tasks, languages, and modalities at the cost of a forward pass. Our empirical evaluations show that it not only achieves state-of-the-art results but also reduces data usage by up to ninefold.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

KAIST-AILab/SyncVSR officialmentioned on GitHubjax report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Landmark-based LipreadingLipreadingSpeech RecognitionVisual Speech Recognitionspeech-recognition

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Landmark-based Lipreading LRS2 SyncVSR Word Error Rate (WER) 74.6 #1 of 1 Archive leaderboard report
Landmark-based Lipreading LRW SyncVSR (Word Boundary) Top 1 Accuracy 80.3 #1 of 5 Archive leaderboard report
Landmark-based Lipreading LRW SyncVSR Top 1 Accuracy 75.1 #2 of 5 Archive leaderboard report
Lipreading CAS-VSR-W1k (LRW-1000) SyncVSR (Word Boundary) Top-1 Accuracy 58.2 #1 of 9 Archive leaderboard report
Lipreading LRS2 SyncVSR Word Error Rate (WER) 16.5 #3 of 25 Archive leaderboard report
Lipreading LRS2 SyncVSR Word Error Rate (WER) 28.9 #11 of 25 Archive leaderboard report
Lipreading LRS3-TED SyncVSR Word Error Rate (WER) 21.5 #3 of 23 Archive leaderboard report
Lipreading LRS3-TED SyncVSR Word Error Rate (WER) 31.2 #12 of 23 Archive leaderboard report
Lipreading Lip Reading in the Wild SyncVSR (Word Boundary) Top-1 Accuracy 95.0 #1 of 22 Archive leaderboard report
Lipreading Lip Reading in the Wild SyncVSR Top-1 Accuracy 93.2 #3 of 22 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections