Introduction
Running Frigate with Home Assistant is one of the most powerful ways to build a local AI-based security system.
However, many people try to run it on a Raspberry Pi and run into performance issues, misconfiguration, or unstable detection.
This article is based on a realistic setup approach, not theory.
What Frigate does
Frigate is a local NVR (Network Video Recorder) with AI object detection.
It can:
detect people, cars, animals
trigger automations in Home Assistant
record events only when motion is relevant
run fully local (no cloud required)
Hardware setup (realistic baseline)
Option A: Raspberry Pi 5
Good for light workloads
Limited for heavy AI inference
Recommended minimum setup:
Raspberry Pi 5 (8GB preferred)
NVMe SSD (strongly recommended)
Coral TPU (USB or PCIe) for AI acceleration
USB 3.0 powered hub (if multiple cameras)
Basic architecture
Cameras → Frigate → MQTT → Home Assistant → Automations
Key installation concept
Frigate runs best in Docker.
Typical setup:
Frigate container
MQTT broker (Mosquitto)
Home Assistant integration
Common setup mistake #1: no hardware acceleration
If you try to run detection only on CPU:
Raspberry Pi 5 becomes overloaded
lag appears
detections drop or freeze
👉 Solution:
Use Coral TPU or reduce detection load drastically.
Common setup mistake #2: too many cameras
People add 4–8 cameras immediately.
On Raspberry Pi:
this will saturate CPU + memory
system becomes unstable
👉 Realistic limit:
1–2 cameras without Coral TPU
3–4 with TPU (depending on resolution)
Common setup mistake #3: wrong resolution
High resolution = heavy processing.
Recommended:
720p or 1080p max for Pi setups
avoid 4K unless dedicated hardware exists
Basic Frigate configuration idea
Key concepts:
detect objects: person, car
record only events
reduce continuous recording
This keeps storage and CPU usage under control.
Integration with Home Assistant
Once connected:
You get entities like:
binary_sensor.front_door_person
camera.front_door
These can trigger automations:
Example:
person detected → turn on lights
person detected at night → notify phone
Performance reality
On Raspberry Pi 5:
Frigate works, but is resource sensitive
tuning is required
expectations must be realistic
It is not an enterprise NVR system.
Recommended stable approach
If you want stability:
Start with 1 camera
Use low resolution
Enable object detection only for “person”
Add Coral TPU before scaling
Future expansion path
Once stable:
add more cameras gradually
add zones (driveway, door, backyard)
integrate alerts
connect to a solar backup system
Related: Basic DIY Solar Backup System for Home →
Conclusion
Frigate + Home Assistant is powerful, but only if you respect hardware limits.
The Raspberry Pi 5 can handle it, but success depends more on configuration and restraint than raw installation.
