Frigate changes the way a home NVR decides what matters. Instead of triggering recordings on motion alone, it sends frames to a supported AI accelerator and checks whether the movement is actually a person, a car, a pet or another object of interest. That dramatically cuts the false alerts that traditional recorders produce when a shadow moves or the wind blows.
Because processing is local, your camera streams never need to leave your network. Frigate runs in Docker on modest hardware, keeps a rolling buffer of recent footage, and records clips around real detections. Zones let you define precise areas, so you are only notified when someone steps onto your porch rather than when a car passes on the street.
Integration is a major part of the project. Frigate plugs into Home Assistant's media browser, exposes sensors and switches via MQTT, and works with OpenHAB, Node-RED and similar tools. The core application is free and open source; Frigate+ is an optional paid service that provides custom-trained detection models tuned for your specific camera angles and conditions.