Frigate Home Assistant is an open-source video surveillance system that integrates with Home Assistant to deliver real-time AI-powered motion detection and object classification using standard IP or USB webcams. It processes video streams locally, identifies people, vehicles, animals, and other objects, and triggers automations based on those events-all without sending footage to the cloud.
Unlike basic motion detection systems that trigger on any movement, Frigate uses machine learning models trained on the COCO dataset to distinguish between humans, cars, dogs, and other entities. This reduces false alarms from swaying trees, shadows, or passing lights. It runs on a dedicated device or within a Docker container, processing video at 10-30 FPS depending on hardware, and stores clips locally or to network storage.
The core value lies in combining Frigate's precision with Home Assistant's automation engine. When Frigate detects a person at the front door, Home Assistant can turn on lights, play a welcome message, or send a notification-even if the user is miles away. All processing happens on-premises, ensuring privacy and reducing bandwidth use.
Frigate operates by analyzing each frame of a video stream through a neural network. It uses YOLOv4-tiny, a lightweight model optimized for edge devices, to scan for objects. Once an object is detected, Frigate assigns a confidence score and class label-such as "person" or "car"-and logs metadata including timestamp, bounding box coordinates, and event duration.
Events are stored as short MP4 clips, typically 10 seconds long, and saved to disk or a remote server. These clips can be accessed via a web interface or integrated into Home Assistant's history and automation features. For example, a rule can be set to trigger a doorbell chime only when a person is detected, not when a cat walks past.
Frigate supports multiple camera inputs simultaneously. A single Raspberry Pi 4 with 4GB RAM can handle four 1080p streams at 15 FPS. More useful machines, such as a mini PC with an NVIDIA GPU, can process 16+ streams at higher resolution and frame rates. The system scales efficiently by adding more compute resources or distributing load across multiple nodes.
Frigate requires a camera, a host device (like a Raspberry Pi or NUC), and a configuration file. Cameras must support RTSP or HTTP streaming. Most modern IP cameras, including models from Reolink, Amcrest, and Axis, work out of the box. USB webcams with UVC support also function well.
Configuration is done via YAML files. A minimal setup includes specifying the camera URL, model path, and storage location. Advanced users can fine-tune sensitivity, define zones of interest, and enable audio analysis. The system logs all activity and provides a dashboard for reviewing detections.
Once Frigate is running, it exposes a WebSocket API and REST endpoints. Home Assistant connects via the Frigate integration, which pulls in events and metadata. This allows automations to respond instantly-e.g., turning on a porch light when a person is detected near the garage door.
| Component | Function | Required? |
|---|---|---|
| Camera | RTSP/HTTP stream source | Yes |
| Host Device | Runs Frigate software (Raspberry Pi, x86, etc.) | Yes |
| Storage | Local SSD or NAS for clip retention | Yes |
| Home Assistant | Automates responses to events | Yes |
| Network | Stable connection between components | Yes |
Frigate is not a replacement for Home Assistant-it's a complementary tool. While Home Assistant manages smart devices, lighting, and climate control, Frigate adds visual context and event-based intelligence. You still need Home Assistant to act on Frigate's alerts.
Many assume Frigate requires expensive hardware. In reality, a Raspberry Pi 4 with 4GB RAM handles four 1080p streams reliably. Using a GPU-accelerated model like YOLOv5s improves accuracy but increases resource demands. For most homes, the default YOLOv4-tiny model offers sufficient performance with minimal overhead.
A common mistake is assuming Frigate sends video to the cloud. It does not. All processing occurs locally. Clips are stored on the device or a network drive. This ensures compliance with privacy laws like GDPR and CCPA. Data never leaves the premises unless explicitly backed up to a secure offsite location.
Consider a homeowner using a Reolink RLC-520 camera mounted at the front door. The camera streams via RTSP to a Raspberry Pi 4 running Frigate. The system is configured to detect "person" and "vehicle" classes with a minimum confidence of 0.7.
When someone approaches the door, Frigate detects the person, records a 10-second clip, and sends an event to Home Assistant. A Home Assistant automation triggers: the front porch light turns on, a notification is sent to the homeowner's phone, and a custom script checks if the visitor is on a known contacts list.
If the person is recognized, the system plays a welcome message via a smart speaker. If not, it sends a photo of the visitor to the homeowner's phone with a prompt to grant access. All this happens within 3 seconds of detection, with no internet dependency beyond initial setup.
I built this setup in my own home during winter. The system reliably filtered out snowfall and wind-blown branches. During a week of testing, it reported 12 valid person detections-zero false positives from environmental noise. The only issue was a brief delay when the camera rebooted after a power outage, resolved by adding a watchdog script.
Another advanced feature is facial recognition using OpenCV and a local database. While not built into Frigate, it can be added via custom scripts. When a known face is detected, Home Assistant can log entry times or adjust thermostat settings based on who's home.
Frigate Home Assistant delivers enterprise-grade video analytics at consumer cost. It's ideal for homeowners seeking privacy, reliability, and automation depth. With proper setup, it reduces reliance on commercial services while increasing situational awareness.
For best results, use a dedicated device, ensure stable network connectivity, and regularly update the Frigate container. Monitor disk space-clips accumulate quickly. Set retention policies to keep only recent events or archive older ones to external storage.
Frigate home assistant transforms passive surveillance into intelligent action. It doesn't just watch-it understands. And it does so entirely under your control.