Frigate

Frigate is a free, open source NVR (Network Video Recorder) built specifically for homelab and home server users who want local, private AI-powered security cameras without sending footage to the cloud. It connects to your existing IP cameras via RTSP streams, runs real-time object detection using TensorFlow locally on your own hardware, and integrates natively with Home Assistant to trigger automations, notifications, and dashboards directly from camera events. You deploy it on your home server using Docker Compose, and everything, including the video, the AI processing, and the event history, stays entirely on your own hardware.

What sets Frigate apart in the security and AI category is how efficiently it handles detection. Rather than processing every single frame, it uses a lightweight motion detection pass to determine where and when to run object detection, dramatically reducing CPU load. It supports hardware-accelerated AI inference via Google Coral TPU, Hailo-8, NVIDIA GPU (TensorRT/CUDA), Intel GPU and NPU (OpenVINO), and Rockchip NPU (RKNN), meaning you can run multi-camera detection on modest hardware without the setup requiring a dedicated server. For homelabbers who are already running something like a Raspberry Pi 5 or a small Intel N100 mini PC, Frigate is designed to run comfortably on exactly that kind of hardware.

Because Frigate communicates over MQTT and integrates directly into Home Assistant's media browser, it becomes the brain of a fully local smart home security system. You can expose detection events as Home Assistant entities, use person detection to trigger lights or lock doors, and search your entire recording history by object type rather than scrubbing through hours of footage. It is a self-hosted replacement for cloud-dependent camera systems like Wyze Cam, Arlo, or Ring that gives you the same smart detection features with full data ownership and no subscription fees.

Key Features

• Real-time object detection (person, car, animal, face)
• Hardware AI acceleration: Google Coral TPU, Hailo-8, NVIDIA TensorRT, Intel OpenVINO
• Native Home Assistant integration with MQTT camera entities and automations
• RTSP stream support with go2rtc re-streaming
• Motion zones, detection masks, and per-camera detection regions
• Event-based and 24/7 continuous recording with retention policies
• WebRTC and MSE low-latency live view
• Docker Compose and Home Assistant add-on deployment

Use Cases

Frigate is ideal for homelabbers who want to replace cloud-dependent cameras like Ring or Wyze with a fully private, self-hosted NVR that runs AI object detection locally on a Raspberry Pi 5 or mini PC with a Coral TPU or Hailo accelerator. It works especially well paired with Home Assistant, where Frigate person detection events can trigger automations like turning on lights, sending a notification with a snapshot, or unlocking a door when a specific person is detected at the front door. For anyone already running Docker at home who wants a self-hosted ZoneMinder alternative with modern AI detection, tight Home Assistant integration, and zero cloud dependency, Frigate is the most capable open source option in the NVR and security camera category.