In This Article

  1. What Is Frigate NVR?
  2. Requirements
  3. Step 1: Install Frigate Add-on
  4. Step 2: Configure Your Cameras
  5. Step 3: Add Coral TPU Acceleration
  6. Step 4: Integrate with Home Assistant
  7. Step 5: Set Up Automations
  8. Tips & Troubleshooting

Frigate is the gold standard for local AI-powered camera object detection in the Home Assistant ecosystem. Unlike cloud-based camera services, Frigate runs entirely on your own hardware, processes video locally, and sends rich detection events — person detected, car in driveway, dog in yard — directly to Home Assistant with zero cloud dependency.

This guide walks you through a complete Frigate setup from scratch, including the optional (but highly recommended) Google Coral TPU accelerator that transforms detection performance.

What Is Frigate NVR?

Frigate is an open-source Network Video Recorder (NVR) built specifically for Home Assistant integration. It ingests RTSP streams from your IP cameras, runs neural network object detection on every frame (or a configurable subset), records clips when objects are detected, and publishes events to Home Assistant via MQTT.

Key capabilities include:

The key technical detail: Frigate uses a low-resolution detection stream (typically 720p or lower) for object detection, and a separate high-resolution stream for recording. This keeps CPU usage manageable even on a NUC or Raspberry Pi.

Requirements

Hardware

Software

Step 1: Install the Frigate Add-on

The easiest way to run Frigate alongside Home Assistant is via the add-on store. You'll need to add the Frigate community repository first.

  1. In Home Assistant, navigate to Settings → Add-ons → Add-on Store
  2. Click the three-dot menu (⋮) in the top right and select Repositories
  3. Add: https://github.com/blakeblackshear/frigate-hass-addons
  4. Refresh the page. Find Frigate NVR and click Install
  5. Install the Mosquitto broker add-on from the official store if you haven't already

After installation, don't start Frigate yet — you need to configure it first.

Step 2: Configure Your Cameras

Frigate is configured via a YAML file. Navigate to the Frigate add-on and open the configuration editor, or create /config/frigate.yml in your Home Assistant config directory.

Here's a minimal working configuration for a single camera:

mqtt:
  host: 127.0.0.1
  port: 1883

cameras:
  front_door:
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/h264Preview_01_main
          roles:
            - record
        - path: rtsp://admin:[email protected]:554/h264Preview_01_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5
    objects:
      track:
        - person
        - car
        - dog

record:
  enabled: true
  retain:
    days: 7
    mode: motion

snapshots:
  enabled: true
  retain:
    default: 14

Key points: use your camera's main (high-res) stream for recording and a substream (lower-res) for detection. Most PoE cameras offer both. Amcrest and Reolink cameras both work well here — check your camera's manual for the correct RTSP URL format.

Step 3: Add Google Coral TPU Acceleration (Recommended)

Running AI object detection on CPU alone works, but it's slow and CPU-intensive. The Google Coral USB Accelerator is a USB dongle that contains a dedicated Edge TPU chip, purpose-built for running neural network inference. With the Coral, Frigate can process detections in under 10ms per frame vs. 50-100ms on a typical CPU.

To add Coral support, plug in the USB accelerator and add the following to your frigate.yml:

detectors:
  coral:
    type: edgetpu
    device: usb

Then add usb: "0" under the add-on's device configuration to pass through the USB device. Restart Frigate and check the logs — you should see "EdgeTPU: initialized" confirming it's working.

Google Coral USB Accelerator

The single best upgrade for Frigate NVR performance. Reduces detection latency from 50-100ms to under 10ms. A must-have for multi-camera setups.

Buy on Amazon →

Step 4: Integrate with Home Assistant

Once Frigate is running, install the Frigate Home Assistant integration:

  1. Go to Settings → Devices & Services → Add Integration
  2. Search for "Frigate" and select it
  3. Enter your Frigate URL (typically http://localhost:5000 for the add-on)
  4. The integration will auto-discover your cameras and create entities

You'll get entities for each camera including:

Step 5: Set Up Automations

With Frigate integrated, you can build powerful automations. Here's a basic person detection notification example in Home Assistant's automation YAML:

alias: "Front Door Person Alert"
trigger:
  - platform: state
    entity_id: binary_sensor.front_door_person_occupancy
    to: "on"
action:
  - service: notify.mobile_app
    data:
      title: "Person detected"
      message: "Someone is at the front door"
      data:
        image: /api/frigate/notifications/front_door/person/snapshot.jpg

You can also use Frigate's MQTT events directly for more granular control, filtering by score, zone, or object type.

Tips & Troubleshooting

Performance Tips

Common Issues

Summary

Frigate NVR is the most powerful local AI camera system available for Home Assistant. With a Coral TPU accelerator, it runs efficiently even with 6-8 cameras on a modest home server. The setup takes a few hours the first time, but the result is a fully local, privacy-respecting security system that rivals commercial offerings.

Related Articles

SmartWired participates in the Amazon Associates Programme. We may earn a commission from qualifying purchases at no extra cost to you.