FaceFusion

Face swapping and face enhancement for images and video are the core of FaceFusion, the most widely used open face manipulation platform at more than 29,000 GitHub stars. Processing runs on ONNX Runtime with execution providers for NVIDIA CUDA, Apple CoreML, and AMD ROCm alongside plain CPU, and a browser-based GUI sits next to a fully headless CLI whose job system creates, queues, and retries batch work for automation pipelines. Terminal installation targets developers comfortable with Python environments, while Windows and macOS installers cover everyone else, and docs.facefusion.io documents every processor and option. The OpenRAIL-AS license keeps the code open while contractually restricting harmful applications such as non-consensual imagery, and the maintainers build content safeguards into the default pipeline. A GPU is not strictly required, though acceleration matters for video, where every frame passes through detection, swapping, and enhancement models. The project grew out of the abandoned roop lineage and remains under active development in 2026.