An open-source application that enables anyone to experiment with deep learning by extracting, training, and swapping faces in images and videos without requiring advanced AI expertise. It supports local execution via CLI or a GUI and relies on user-provided training data.
Project overview
It was the first deepfakes code that anyone could download, run, and learn from by experimentation, lowering the barrier to exploring deep learning face recognition and swapping.
Project type
Image & Vision · Video · Learning Resources
Use cases
Image Processing · Video Creation · Learning & Education
Deployment
Refer to project documentation
License
GPL-3.0
Best for
Creators, developers, and general users who want to learn about AI by experimenting with face recognition and swapping without needing advanced expertise.
Key capabilities
Extracts faces from raw photos into a designated folder.
Trains a deep learning model on faces extracted from photos or videos.
Applies trained models to swap and convert faces in source photos.
Provides a GUI as an alternative to command-line execution.
Offers a dedicated video conversion tool accessible via command line.
Limitations and risks
The tool requires sufficient training data to work effectively.
The technology can be used to create inappropriate or unethical content without consent.
Getting started
Setup difficulty is medium, requiring an understanding of Python setup or the use of pre-built installers, along with collecting and preparing training data. The path to first success involves getting the code, running 'python faceswap.py extract', running 'python faceswap.py train', and running 'python faceswap.py convert'.
Evidence and sources
GitHub project description: Deepfakes Software For All
README: it was the first AI code that anyone could download, run and learn by experimentation without having a Ph.D. in math, computer theory, psychology, and more.
Release: Faceswap Windows, Linux and macOS Installers v3.0.0
README: Your best bet is to join the FaceSwap Discord server where there are plenty of users willing to help.