bannedbook/faceswap

FaceSwap is a free, open-source tool that lets you replace one person's face with another in photos and videos. You might use it to swap a friend's face into a movie scene, create a parody video, or experiment with AI techniques as a learning exercise. The project bills itself as "deepfakes software for all," meaning it aims to make this technology accessible to anyone with a decent computer, not just researchers with advanced math degrees. The workflow has three main steps. First, you collect a set of photos or video frames containing the faces you want to use. The tool extracts and isolates those faces from the surrounding images. Second, you train a model, essentially, the software studies the two faces and learns how to map one onto the other through repetition and adjustment. Third, you run the conversion, which applies the trained model to your target images or video, producing the final swapped result. There's also a graphical interface if you prefer buttons to command-line instructions. The project runs on Windows, Linux, and macOS, but you'll want a modern graphics card with CUDA support (typically an NVIDIA GPU) for reasonable performance. AMD cards get partial support. The README notes that reusing an existing model trains much faster than starting from scratch, and if you don't have enough training data, you can start with someone who looks similar and then switch. Who would use this? Hobbyists learning about machine learning, VFX artists exploring new techniques, filmmakers working on projects that require face replacement, or anyone curious about how deepfakes actually work under the hood. The developers stress that the tool is meant for ethical uses, commentary, education, movies, experimentation, and explicitly prohibit using it for nonconsensual content or deception. One notable aspect is the project's framing. The README spends considerable space on a "manifesto" defending the technology, arguing that deepfakes opened up AI learning to people without PhDs and that the media focus on misuse overshadows legitimate applications. The developers take a zero-tolerance stance on unethical use, though the software itself is freely available.