Frigate
Frigate is an open-source NVR (Network Video Recorder) built around real-time local AI object detection for IP cameras. Instead of relying on cloud services, it runs detection on your own hardware — with optional support for Coral TPU accelerators for fast, low-power inference. It is one of the most popular projects in the self-hosted community, with 34,000+ GitHub stars.
Key Features
- Real-time object detection — Identifies people, cars, pets and other objects as they appear in camera feeds, using local AI models via OpenCV, TensorFlow Lite or Coral TPU.
- Fully local processing — All detection and recording happens on your hardware; no footage or metadata is uploaded to the cloud.
- Event-based recording — Only records when motion or a detected object appears, saving disk space compared to 24/7 recording.
- Multi-camera support — Manage many IP cameras from a single web interface, with live view, snapshots and clip review.
- Home Assistant integration — First-class integration with Home Assistant, with rich MQTT events for automation and alerts.
- WebRTC live streaming — Low-latency live viewing in the browser without plugins.
Why Use Frigate?
Frigate turns ordinary IP cameras into a smart security system with no subscription fees and no cloud dependency. The AI detection is fast enough to distinguish a person from a passing car, which dramatically cuts false-positive alerts. If you already run Home Assistant, Frigate slots in as the eyes of your home automation setup.
Use Cases
- Home security with person/car/pet detection and instant notifications
- Monitoring entry points, driveways and gardens without cloud cameras
- Building a privacy-respecting NVR that keeps footage on your own NAS or server
Platform
Linux (Docker container) · Runs on x86_64 and ARM (including Raspberry Pi 4/5)
Licence
MIT