Here We have Transformed the original wav2lip model from pytorch to tensorflow

We also modified the make_video part for making videos from generated images

You can get the tensorflow weights and pytorch weights here in this google drive link

Feel free to Contact with us for details here, [email protected]

Instructions:

1. install ffmpeg (video editor) "sudo apt-get install ffmpeg" (for linux machine)
   
2. install requirements.txt (probably you dont need all of them)
   
3. try to use videos which has faces in all the frames
   
4. try to use videos and audios of 20 secs (max 30 sec)
   
5. try to use videos longer than audios (must)
   

You need all the files in this folder and also files from the drive link to run the make_video.py program

The face detection part was from this repo which we found in the wav2lip repo