USED CAMERAS AND AN OLD PC BECOME AI HOME SECURITY WITH ZERO MONTHLY FEES
Turn old IP cameras and a spare PC into a smart security system that runs entirely at home, with AI object detection and no cloud subscription.
HomeAI
- difficulty
- ●●●○○
- time
- a weekend
- license
- MIT
- repo
- repo ACTIVE34,719 stars
●●●○○ · a weekend · MIT · 34,719 stars · repo ACTIVE
WHAT YOU’LL NEED
- Dev board / microcontroller — runs the project firmware
Partner
COMPAREE VERDICT
Frigate is the go-to open-source NVR for people who want AI object detection without a subscription, and it has earned its 34,000 stars by actually working. You point it at any IP camera's RTSP stream, it records clips when it sees a person or a car (or a raccoon), and nothing leaves your network. The Home Assistant integration is first-class if you want automations, but it works fine standalone via its own web UI. The catch: this is not plug-and-play consumer software. You will spend your first evening debugging Docker networking, working out which cameras expose a usable RTSP feed, and learning what a Coral accelerator is and whether you need one. If you have never run a Docker container or edited a YAML file, budget extra time for that learning curve. But once it is running, it is reliable, fast, and the lack of a cloud dependency means it keeps working even when your internet drops. The most common mistake is buying cameras that do not expose RTSP or using an underpowered host without acceleration — read the hardware recommendations before you buy anything.
IN THE REPO
NOT IN THE REPO
- —Docker containers and a complete configuration reference are provided. The repository contains all the software you need to run the system.
- —There is no prescriptive hardware BOM — you choose your own cameras and host machine based on the documentation's compatibility tables and performance recommendations.
- —The docs are extensive but assume you are comfortable with Docker, YAML configuration files, and basic networking concepts like RTSP streams.
- —MIT licensed — use it commercially, modify it, no restrictions.
- —Works standalone or integrates deeply with Home Assistant if you already run it.
- —The biggest hidden cost is often the Coral USB accelerator (around $60-80 when available) — Frigate will run without it, but AI detection becomes much slower and CPU-heavy.
Can I build this?
Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.
Partner · KickstarterBluHub reads and burns Blu-ray/DVD/CD, adds USB-C, USB-A, SD/TF and 10 Gbps data — plug and play, grand-piano finish.
Gallery
Start here
Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.
- 1.Read the hardware requirements page in the docs and decide on your host (Pi, mini PC, or NAS) and whether you need a Coral accelerator.(The performance comparison table will tell you if your hardware will bottleneck without acceleration.)
- 2.Confirm your cameras support RTSP. Check the manufacturer's docs or test the stream URL with VLC before you buy.(Most modern IP cameras do, but cheap no-name models and some consumer Wi-Fi cameras do not expose it.)
- 3.Install Docker on your host, then follow the Docker Compose setup in the official docs.(The Docker route is the most supported. The docs also cover Home Assistant OS add-on and Proxmox LXC installs.)
- 4.Edit the config.yml to add your first camera's RTSP URL and define detection zones.(Start with one camera and get that working before adding more. The config file has extensive inline comments.)
- 5.Open the web UI, check the camera feed, and tune detection thresholds and masks.(You will spend time masking out trees that trigger motion and adjusting the confidence threshold to stop false positives.)
Resources
Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.
Partner · KickstarterCORE POWER BANK S: 0.2" thin, 2.8 oz, 5000 mAh with 22.5W fast charging. Semi-solid battery, carbon fiber, IPX7 waterproof.
KNOWN ISSUES
- Buying cameras that do not expose RTSP or only work through a proprietary app. Check this before you buy — it is the most common dead end.
- Running on underpowered hardware without a Coral accelerator. A Raspberry Pi 4 will work, but detection latency will be high and CPU will spike. Budget for a Coral if you want real-time AI.
- Not reading the config reference and copying someone else's YAML verbatim. Every camera and network is different — you will need to adjust RTSP paths, resolution, and masks.
- Skipping detection masks and zones. Without them, you will get alerts every time a tree sways or headlights sweep past a window.
- Expecting instant setup. Budget a weekend to learn the system, not an hour.
- Using Wi-Fi cameras in a setup with many streams. Wired PoE is more reliable for always-on recording.
Do I need Home Assistant to run this?
No. Frigate has its own web UI and works standalone. The Home Assistant integration adds automations and dashboard cards, but it is optional.
Can I use old webcams or USB cameras?
Technically yes via go2rtc bridges, but IP cameras with native RTSP are much more reliable. USB cameras are fiddly and not the typical use case.
What is a Coral accelerator and do I need one?
A Google Coral is a USB or PCIe stick with a chip purpose-built for running AI inference fast. Without it, Frigate uses your CPU for object detection, which is slower and creates lag. Most people buy one after their first week.
Will this work with my existing Reolink / Wyze / Amcrest camera?
If the camera exposes an RTSP stream, yes. Check the docs — there is a community hardware page listing tested models and their quirks.
How much storage do I need?
Depends on how many cameras, resolution, and retention length. The docs include a storage calculator. A typical setup with four 1080p cameras and 7-day retention needs around 500GB.
Community builds
No community builds yet — be the first, we feature the best ones.
Discussion1
FROM THE COMPAREE TEAM
34,000 stars and it runs entirely at home — no cloud, no subscription. Did you go with a Coral accelerator from the start, or run CPU-only first and upgrade later?
Blake Blackshear
Blake Blackshear built Frigate to solve his own frustration with cloud-dependent security cameras that recorded everything and sent it offsite. He wanted AI detection that ran at home and only saved what mattered. The project became one of the most-starred self-hosted surveillance systems on GitHub and is now a cornerstone of the Home Assistant ecosystem.
Keep building — related projects
Same category, picked by how close they are to this build.
NO CAD FILESTurn any plushie into a talking AI friend with a 10 dollars chip inside
AIHome●●●●○advanced
FULL CADA used 20 dollars Google speaker becomes your private AI assistant with one swapped board
AIHome●●●○○moderate
NO CAD FILESThis AI camera feeder recognizes your pet and refuses to feed the wrong one
HomeAI●●●○○moderate
SOFTWARE ONLYAn AI inbox triage that drafts replies but physically cannot hit send
AIAutomation●●○○○starter
DISCLAIMER
- Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
- Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
- Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
CompareeTEAM23d agoedited
Practical notes from our verification: the repository is active with recent commits, the documentation is extensive but assumes Docker and networking comfort, and the single biggest decision is the Coral accelerator — Frigate will run without it, but real-time AI detection on multiple streams becomes CPU-bound quickly. The Home Assistant integration is mature and well-supported, but the standalone web UI is perfectly usable if you do not run HA. Most first-time builders underestimate the time spent on camera RTSP configuration and detection zone tuning — that is where the weekend goes, not the Docker setup.