dot

dot, short for Deepfake Offensive Toolkit, is a Python tool that generates real-time face-swap deepfakes from a webcam feed and routes the result into a virtual camera. The intended audience is security researchers and red team professionals who test whether identity verification systems, video call platforms, or biometric onboarding services can be fooled by a digitally altered face. The README notes that users are responsible for complying with local laws, and that the authors take no responsibility for misuse.

The core idea is that you give the tool a still photo of a target face, point it at your webcam, and it replaces your face with the target face in real time. The output is then available as a virtual camera device, so any app that accepts a webcam as input will see the swapped face instead of your real one. No additional training is required on your end: the models ship pre-trained and work immediately once downloaded.

Several face-swap techniques are bundled. The highest quality option uses a method called SimSwap at two resolution settings, with an optional enhancement step for sharpening the result. A lighter-weight option uses the OpenCV library for lower-fidelity swaps. A third approach, based on a model called First Order Motion Model, animates a still photo to mimic head and facial movements rather than doing a direct pixel swap.

You can use dot through a graphical desktop application available as a downloadable executable for Windows and Mac, or through a command-line interface that requires more setup including Python, Conda, and pre-trained model files. GPU support is available and recommended for performance, though CPU-only mode works at reduced speed.

The project is built by Sensity AI and is licensed under the BSD 3-Clause license. The full README is longer than what was shown.