Wav2lip is a lip-sync expert, that is all you need for speech to lip generation in the wild

We need a video and an audio, both files have to be exact same length. The audio we will extract from a Vietnamese video (named vi-news.mp4) and the video we will download an English video (named en-news.mp4). See this guide to download these videos.

# Cutting the video and audio (14 seconds)
ffmpeg -i en-news.mp4 -ss 00:00:00.000 -to 00:00:14.000 -async 1 en-news-cut.mp4
ffmpeg -i vi-news.mp4 -ss 00:01:27.500 -to 00:01:41.500 -async 1 vi-news-cut.mp4

# Extract audio from vi-news-cut.mp4
ffmpeg -i vi-news-cut.mp4 -q:a 0 -map a vi-news-cut.mp3

Upload en-news-cut.mp4 and vi-news-cut.mp3 to the wav2lip web demo and download the result video

References

Wav2Lip