The landmark open-source lip-sync foundation model accurately synchronizing spoken speech to real video faces.
Pretrained models and expert lip-sync discriminator fully open-sourced on GitHub, aligning any speech audio to any video target.
Wav2Lip is universally recognized as the foundational gold standard in AI video lip synchronization (published at ACM Multimedia 2020). In multilingual film dubbing, educational localized voiceovers, and avatar broadcasting, matching an existing actors mouth shapes seamlessly to an entirely new audio recording has historically been an elusive hurdle. Wav2Lips genius breakthrough stems from its pre-trained Expert Lip-Sync Discriminator: a specialized neural network trained on vast audiovisual corpora that penalizes even microscopic timing deviations between audio phonemes and visual mouth shapes. Regardless of dynamic head tilting, varied lighting, or shifting facial expressions in the target video, Wav2Lip seamlessly synthesizes accurate, synchronized lip movements onto the target face while keeping background scene physics and upper-face emotions pristine. The vast majority of modern AI video translation and dubbing services are architectural descendants of Wav2Lip.