# Lip Reading on LRS3 1.0 (test) Source: [Lip Reading on LRS3 1.0 (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:08:25.578Z Evaluation of Lip Reading performance on the test split of the LRS3 1.0 dataset, measured using Word Error Rate (WER). A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Lip Reading | | Dataset (including any stated split) | LRS3 1.0 (test) | | Metric | WER | | Unit | PERCENT | | Direction | Lower is better | | Results loaded | 19 of 19 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [WER](): Lower is better; unit: PERCENT. ## Recorded results | Method | WER | Result date | Report paper | | --- | --- | --- | --- | | AV-HuBERT + relaxed cross attention | 25.51 | Not recorded | 68fa6d6e1e209131cea39cdb | | AV-HuBERT + relaxed cross attention | 28.05 | Not recorded | 68fa6d6e1e209131cea39cdb | | Visual Transformer Pooling (VTP) | 30.7 | Not recorded | 68fa6b631e209131cea2df6c | | RNN-T | 33.6 | Not recorded | 68fa6b631e209131cea2df6c | | Makino et al. | 33.6 | Not recorded | 68fa6d6e1e209131cea39cdb | | Visual Transformer Pooling (VTP) | 40.6 | Not recorded | 68fa6b631e209131cea2df6c | | AV-HuBERT + relaxed cross attention | 42.68 | Not recorded | 68fa6d6e1e209131cea39cdb | | Hyb. + Conformer | 43.3 | Not recorded | 68fa6b631e209131cea2df6c | | Ma et al. | 43.3 | Not recorded | 68fa6d6e1e209131cea39cdb | | Baseline (Shi et al.) | 46.1 | Not recorded | 68fa6d6e1e209131cea39cdb | | Ma et al. | 46.9 | Not recorded | 68fa6d6e1e209131cea39cdb | | CTC-V2P | 55.1 | Not recorded | 68fa6b631e209131cea2df6c | | Xu et al. | 57.8 | Not recorded | 68fa6d6e1e209131cea39cdb | | TM-seq2seq | 58.9 | Not recorded | 68fa6b631e209131cea2df6c | | Afouras et al. | 58.9 | Not recorded | 68fa6d6e1e209131cea39cdb | | CTC + KD | 59.8 | Not recorded | 68fa6b631e209131cea2df6c | | Afouras et al. | 59.8 | Not recorded | 68fa6d6e1e209131cea39cdb | | Afouras et al. | 59.9 | Not recorded | 68fa6d6e1e209131cea39cdb | | Conv-seq2seq | 60.1 | Not recorded | 68fa6b631e209131cea2df6c | ### AV-HuBERT + relaxed cross attention WER: 25.51. Result ID: 698fa1083a6c314849faa7b7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Unlabeled data (pre-training)":"1,326h","Labeled data (fine-tuning)":"433h + 1,326h","self-training":true,"Language Model":"Joint Transformer Decoder","matched inference":false}. This record does not assert independent reproduction. ### AV-HuBERT + relaxed cross attention WER: 28.05. Result ID: 698fa1083a6c314849faa7b6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Unlabeled data (pre-training)":"1,326h","Labeled data (fine-tuning)":"433h","Language Model":"Joint Transformer Decoder","matched inference":false}. This record does not assert independent reproduction. ### Visual Transformer Pooling (VTP) WER: 30.7. Result ID: 698e8635ffc3901de09e54d3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Datasets used (Training)":"LRS2, LRS3, MV-LRS†, TEDxext","Total # hours (Training)":"2,676"}. This record does not assert independent reproduction. ### RNN-T WER: 33.6. Result ID: 698e8635ffc3901de09e54d2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Datasets used (Training)":"YT31k†","Total # hours (Training)":"31,000"}. This record does not assert independent reproduction. ### Makino et al. WER: 33.6. Result ID: 698fa1083a6c314849faa7b3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Labeled data (fine-tuning)":"33,000h","Language Model":"None"}. This record does not assert independent reproduction. ### Visual Transformer Pooling (VTP) WER: 40.6. Result ID: 698e8635ffc3901de09e54cf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Datasets used (Training)":"LRS2, LRS3","Total # hours (Training)":"698"}. This record does not assert independent reproduction. ### AV-HuBERT + relaxed cross attention WER: 42.68. Result ID: 698fa1083a6c314849faa7b5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Unlabeled data (pre-training)":"1,326h","Labeled data (fine-tuning)":"30h","Language Model":"Joint Transformer Decoder","matched inference":false}. This record does not assert independent reproduction. ### Hyb. + Conformer WER: 43.3. Result ID: 698e8635ffc3901de09e54ce. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Datasets used (Training)":"LRS3, LRW","Total # hours (Training)":"639"}. This record does not assert independent reproduction. ### Ma et al. WER: 43.3. Result ID: 698fa1083a6c314849faa7b2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Labeled data (fine-tuning)":"433h + 157h","Language Model":"Joint Transformer Decoder"}. This record does not assert independent reproduction. ### Baseline (Shi et al.) WER: 46.1. Result ID: 698fa1083a6c314849faa7b4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Unlabeled data (pre-training)":"1,326h","Labeled data (fine-tuning)":"30h","Language Model":"None"}. This record does not assert independent reproduction. ### Ma et al. WER: 46.9. Result ID: 698fa1083a6c314849faa7b1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Labeled data (fine-tuning)":"433h","Language Model":"Joint Transformer Decoder"}. This record does not assert independent reproduction. ### CTC-V2P WER: 55.1. Result ID: 698e8635ffc3901de09e54d1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Datasets used (Training)":"LSVSR†","Total # hours (Training)":"3,886"}. This record does not assert independent reproduction. ### Xu et al. WER: 57.8. Result ID: 698fa1083a6c314849faa7b0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Labeled data (fine-tuning)":"433h + 157h","Language Model":"None"}. This record does not assert independent reproduction. ### TM-seq2seq WER: 58.9. Result ID: 698e8635ffc3901de09e54d0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Datasets used (Training)":"LRS2, LRS3, LRW, MV-LRS†","Total # hours (Training)":"1,637"}. This record does not assert independent reproduction. ### Afouras et al. WER: 58.9. Result ID: 698fa1083a6c314849faa7af. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Labeled data (fine-tuning)":"1,362h + 157h","Language Model":"Joint Transformer Decoder"}. This record does not assert independent reproduction. ### CTC + KD WER: 59.8. Result ID: 698e8635ffc3901de09e54cd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Datasets used (Training)":"LRS2, LRS3, VoxCeleb2‡","Total # hours (Training)":"1,032"}. This record does not assert independent reproduction. ### Afouras et al. WER: 59.8. Result ID: 698fa1083a6c314849faa7ad. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Unlabeled data (pre-training)":"334h","Labeled data (fine-tuning)":"433h","Language Model":"Joint Transformer Decoder"}. This record does not assert independent reproduction. ### Afouras et al. WER: 59.9. Result ID: 698fa1083a6c314849faa7ae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Labeled data (fine-tuning)":"1,362h + 157h","Language Model":"None"}. This record does not assert independent reproduction. ### Conv-seq2seq WER: 60.1. Result ID: 698e8635ffc3901de09e54cc. [Reporting paper]() Recorded evaluation configuration: {"Datasets used (Training)":"LRS2, LRS3","Total # hours (Training)":"698"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.