Frigate

Frigate is a standout in the open-source NVR space because it genuinely solves the false-positive problem that plagues motion-based cameras. Instead of recording every wind-blown leaf, it uses AI object detection to trigger on what matters — people, cars, animals. The key is that all processing is local, so your footage never touches the cloud. This is a huge privacy win and eliminates recurring cloud fees. The zone-based alerting is a killer feature. You can define specific regions (like your porch or driveway) and get notified only when a person or vehicle enters those zones. This is far more precise than generic motion alerts and turns a security camera into a truly smart sensor. For Home Assistant users, Frigate is basically a must-have. It integrates directly into the media browser, provides low-latency camera entities, and exposes real-time sensors and switches. You can build automations like turning on lights when a person is detected at night, or sending a notification with a snapshot when a car pulls into the driveway. The MQTT support also makes it work with OpenHab, NodeRed, and any other platform that speaks MQTT. The main drawback is the learning curve. You need to be comfortable with Docker, Linux, network camera configuration, and ideally a Coral TPU or a decent GPU. Without an accelerator, detection speed drops significantly, which leads to missed events. The free community models are decent, but the Frigate+ subscription offers custom models that noticeably improve accuracy, so there is a cost if you want the best detection. That said, once set up, Frigate is rock-solid. The community is active, and the recent release of a camera-compatibility viewer (see news) shows the ecosystem is maturing. For self-hosters who prioritize privacy and control, Frigate is arguably the best option out there. It's not for everyone, but if you're willing to put in the work, the payoff is huge.