What is frigateane?

Frigate ANE Detector is a native Mac app for Apple Silicon computers that makes it much easier to run Frigate, an existing open source home security camera system, using Apple's Neural Engine chip for fast object detection. Frigate can already detect objects like people or cars in camera footage, but getting it to use the Neural Engine on a Mac required a manual, multi step process involving shell scripts, networking rules, and hand edited configuration files. This app wraps that whole process in a single window with a graphical setup wizard. The app is not a replacement for the existing community project that figured out how to run the detection model on the Neural Engine. Instead, it builds a full interface and automation layer on top of that approach. Through the setup wizard, a user can configure their cameras by entering video stream addresses and login details, connect to an MQTT message broker and Home Assistant for smart home integration, choose where camera recordings are stored, and pick which detection model to use. The app then generates the Frigate configuration file automatically and can start and stop the whole system with one click. Once running, a dashboard shows the status of the system, whether the Apple Neural Engine detector is active, and gives quick access to features like resetting the admin password for the underlying Frigate web interface, viewing or backing up the configuration, and setting up the networking rules needed for the app to reach cameras and other devices on the local network. The project targets a Mac with Apple Silicon, at least macOS 13 for the app itself, and macOS 26 with Apple's newer container feature to actually run Frigate. Building the app from source requires Xcode's command line tools, and a build script downloads and prepares a self contained Python environment with the machine learning libraries Frigate's detector needs, so the final app does not require users to install Python themselves. The project is aimed at people who want Frigate running on the Neural Engine without using the terminal.