Lip-to-speech synthesis on LRS3 unseen speaker protocol (test)
21.9WERLipVoicerofficial
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LipVoicerofficialvariant=official, vocoder=HiFi-GAN [16]2026.02 | 21.9 | 93 | 3.19 | 8.166 | 6.308 | |
| LipVoicerreconvariant=reimplementation, training_settings=default2026.02 | 22.1 | 93 | 3.17 | 8.516 | 6.4 | |
| LipSodyprosody_modeling=explicit2026.02 | 22.5 | 92 | 3.18 | 8.435 | 6.404 |