Wav2Lip

Wav2Lip: Accurately Lip-syncing Videos In The Wild

Wav2Lip wrapper pypi package code for this package is available at: https://github.com/mehdihosseinimoghadam/Wav2Lip also original code from writers of Wav2Lip is available at: https://github.com/mehdihosseinimoghadam/Wav2Lip

Prerequisites

• Python 3.6
• ffmpeg: sudo apt-get install ffmpeg
• Face detection pre-trained model should be downloaded to face_detection/detection/sfd/s3fd.pth. Alternative link if the above does not work.

Getting the weights

Features

• Easy to use
• Fast
• Accurate

Usage

First of all get the weights:

wget "https://www.adrianbulat.com/downloads/python-fan/s3fd-619a316812.pth" -O "/usr/local/lib/python3.7/dist-packages/Wav2Lip/face_detection/detection/sfd/gfs3fd.pth"

gdown https://drive.google.com/uc?id=1jQOJInh8cDj2mrbUgcQxhCc7rpAgyV1-

After pip installation add these directories:

mkdir /usr/local/lib/python3.7/dist-packages/Wav2Lip/results
mkdir /usr/local/lib/python3.7/dist-packages/Wav2Lip/temp
mkdir /usr/local/lib/python3.7/dist-packages/Wav2Lip/checkpoints

Then add weights:

mv /content/wav2lip_gan.pth /usr/local/lib/python3.7/dist-packages/Wav2Lip/checkpoints/
&&
cd /usr/local/lib/python3.7/dist-packages/Wav2Lip

Import Wav2Lip wrapper function:

from Wav2Lip.wrapper_app import main
main("/path/to/wav/file","path/to/image")

The resulting video would be in

/usr/local/lib/python3.7/dist-packages/Wav2Lip/results/result_voice.mp4

If you run in colab you can use this script:

from IPython.display import HTML
from base64 import b64encode
mp4 = open('/usr/local/lib/python3.7/dist-packages/Wav2Lip/results/result_voice.mp4','rb').read()
data_url = "data:video/mp4;base64," + b64encode(mp4).decode()
HTML("""
<video width=400 controls>
      <source src="%s" type="video/mp4">
</video>
""" % data_url)

Authors

Github

Source is avaliable at https://github.com/mehdihosseinimoghadam/Wav2Lip

License

MIT

Free Software, Hell Yeah!

Acknowledgements

This is only a wrapper package and the main code of Wav2Lip can be found in https://github.com/mehdihosseinimoghadam/Wav2Lip