Rudrabha/Wav2Lip ? reverse-engineered prompt

I want this repo turned into a simple lip sync tool I can actually run locally. The main thing I need is a clear way to give it a face video and a separate speech audio file, then get back a new video where the lips match the audio as closely as possible. Please set up the open source non commercial version that’s already in the repo, including the pretrained model and inference flow, not just the paid API example.

Make it easy for a normal person to use. A simple command line workflow is fine, and if it’s easy, a tiny upload form would be nice too. Please handle the usual setup, model download, face detection, temp files, and output saving so it works end to end. Add a short README with exact steps, expected input formats, and one example command. If training scripts are already here, leave them usable, but focus on getting inference working reliably first. Look up current docs online if you need to.