Local NVR vs Cloud Cameras: Frigate, Blue Iris, UniFi Protect, and Why It Matters

For years, home security cameras meant picking a brand, paying a monthly subscription, and trusting a company in California (or wherever) to store video of the inside of your home on their servers. Ring, Arlo, Nest — they all work the same way. Your footage goes to the cloud. You pay to access it. And if the company gets hacked, goes bankrupt, or gets a law enforcement subpoena, your footage goes with it.

There’s a growing alternative: local NVR (Network Video Recorder) systems that keep every second of footage on hardware in your home. No subscriptions. No cloud. No third parties. And with modern AI detection software like Frigate, the performance often beats what cloud services offer — at a fraction of the long-term cost.

This guide breaks down exactly how local NVR compares to cloud-based cameras, which software options make sense for different users, and how to decide whether the setup effort is worth it for your home.

Local NVR vs Cloud Cameras: The Core Difference

The fundamental difference is simple: where does your footage live?

With cloud cameras (Ring, Arlo, Nest/Google), your cameras send footage to the company’s servers. You access it through their app. Motion alerts trigger from their AI running on their hardware. If you want to save more than a few days of footage, or access smart detection features, you pay a monthly fee.

With a local NVR system, your cameras send footage to a device in your home — a mini-PC, a NAS, a Raspberry Pi, or a dedicated NVR appliance. Your footage stays there. Detection runs locally. You access it through your local network or via a VPN. Nobody else sees it.

Cloud Cameras: Ring, Arlo, and Nest — Honest Pros and Cons

Cloud camera systems are popular for good reasons. They’re easy to set up, the apps are polished, and they work well for most people who just want basic coverage without technical overhead.

What Cloud Cameras Do Well

• Setup speed: Most cloud cameras are running in 10–15 minutes per device. No server to configure, no YAML files to edit.
• Remote access: The apps handle remote viewing without any VPN or port forwarding setup.
• Automatic updates: Firmware updates roll out automatically, including security patches.
• Off-site backup: If someone steals your camera, the cloud still has the footage.

Where Cloud Cameras Fall Short

• Ongoing cost: Ring Protect Plus runs about $100/year. Arlo Secure is similar. Over 3 years with 4 cameras, subscriptions alone add $300–$468 on top of the camera hardware cost.
• Privacy exposure: Ring has a documented history of sharing footage with law enforcement without warrants — a practice that drew an FTC settlement in 2023. Nest footage lives on Google’s servers. What happens to your data when these companies are acquired or change policies is genuinely uncertain.
• False alarms: Cloud AI isn’t as accurate as local AI when tuned properly. Independent testing in 2026 found Ring generating false positives at a 22.7% rate in daylight, climbing to 39.4% at dawn and dusk. Arlo ran 15–20% false positives in mixed lighting conditions.
• No offline operation: If your internet goes down, many cloud cameras stop recording or alerting entirely.
• Platform changes: Nest dropped free continuous recording without warning. Arlo changed subscription tiers. Ring has altered its privacy policy multiple times. You’re dependent on the company’s decisions.

Local NVR Options: Frigate, Blue Iris, UniFi Protect, and Scrypted

The local NVR space has matured significantly in the past few years. Here are the four most capable options and who each one is right for:

Frigate: Best for Home Assistant Users and Privacy-First Setups

Frigate is an open-source NVR built specifically for Home Assistant integration. It runs in Docker, uses YAML for configuration, and offloads AI detection to a Google Coral TPU or Hailo accelerator — keeping all processing local and fast.

What makes Frigate stand out:

• Under 3% false positive rate in daylight after tuning (vs. 22.7% for Ring)
• Person, vehicle, animal, and package detection — all local
• Facial recognition via Double Take or CompreFace integration
• License plate recognition via CodeProject.AI
• MQTT event stream for real-time Home Assistant automations
• No accounts, no cloud, fully offline-capable

Hardware needed: An Intel N100-class mini-PC (~$150–250), a Google Coral USB TPU (~$60–80), PoE cameras (~$50–100 each), a PoE switch (~$40–80), and 2TB of storage (~$60–80). Total upfront cost for 4 cameras runs $510–890, with no monthly fees. It typically breaks even against Ring Protect Plus in about 12 months.

The catch: Setup takes 2–6 hours. You need to be comfortable with Docker and YAML. Configuration requires ongoing tweaking to tune detection zones and thresholds. It’s not plug-and-play.

UniFi Protect: Best for Ubiquiti Households

If you’re already running Ubiquiti networking gear (UniFi routers, switches, access points), UniFi Protect is the cleanest local NVR option. It runs on Ubiquiti’s consoles (UNVR, UDM Pro, UDM SE) and pairs natively with their G5 and G6 camera lines, which include on-camera AI detection.

UniFi Protect has no monthly subscription for local recording. The app is polished and well-maintained. Remote access works through Ubiquiti’s cloud relay, which you can disable if you want fully local-only operation. The main limitation is that it works best with UniFi cameras — third-party ONVIF cameras work but don’t get smart detection features.

Blue Iris: Best for Windows Users with Mixed Camera Brands

Blue Iris is a Windows-native NVR application with a one-time license fee of $69.95. It’s been around since 2008 and supports a huge range of ONVIF cameras. If you’re running a mixed fleet of cameras from different manufacturers — Reolink, Hikvision, Dahua, Amcrest — Blue Iris is typically the most compatible option.

It requires a Windows PC to run and has a steeper initial learning curve than UniFi Protect, but it’s extremely powerful once configured. AI detection has improved significantly with DeepStack and CodeProject.AI integrations in recent versions.

Scrypted: Best Bridge for Cloud Cameras Moving to HomeKit or Home Assistant

Scrypted isn’t a full NVR — it’s more of a protocol bridge. If you have Ring, Arlo, or other cloud cameras that you want to integrate into Apple HomeKit or Home Assistant without replacing the hardware, Scrypted makes that possible. It’s often used as a companion to Frigate: Scrypted handles the cloud camera bridging, Frigate handles the AI detection.

Cost Comparison: Local NVR vs Cloud Over 3 Years (4 Cameras)

System	Upfront Cost	3-Year Subscription	3-Year Total
Ring (4 cameras + Protect Plus)	$400–800	$180–468	$580–1,268
Arlo (4 cameras + Secure)	$520–1,400	$468	$988–1,868
Frigate local NVR (4 cameras)	$510–890	$0	$510–890
UniFi Protect (4 G5 cameras + UDM)	$900–1,400	$0	$900–1,400

The numbers make a strong case for local NVR if you’re comfortable with setup. Frigate breaks even against Ring Protect Plus in roughly 12 months, and the gap grows every year after that.

Which Cameras Work Best with a Local NVR?

Cloud cameras like Ring and Arlo are designed to work only with their own cloud infrastructure — they don’t support the RTSP or ONVIF protocols that local NVR systems need. If you want to go local, you need cameras that support RTSP or ONVIF. Good options include:

• Reolink: Strong value, reliable RTSP support, good image quality. The Reolink RLC-810A is a popular Frigate-compatible option around $50–80.
• Amcrest: Wide range of PoE and Wi-Fi cameras with solid ONVIF support. Good for Blue Iris setups.
• Hikvision / Dahua: Enterprise-grade options often used in DIY NVR setups. Strong image quality and feature set, though both are Chinese manufacturers — a consideration for some buyers.
• Ubiquiti G5 / G6: Best with UniFi Protect. Premium build quality with on-camera AI processing.
• Annke: Budget-friendly ONVIF cameras with good Frigate compatibility.

For camera selection guidance specific to indoor and outdoor use cases, our guide to smart home security cameras covers what to look for when evaluating specs.

Is Local NVR Right for You? A Practical Decision Guide

Go local NVR if:

• You’re already running Home Assistant or are comfortable with self-hosted tools
• Privacy is a priority and you don’t want footage stored on third-party servers
• You want more than 4 cameras without subscription costs scaling proportionally
• You want advanced AI features (facial recognition, plate reading) without paying cloud AI prices
• You’re comfortable with an initial technical setup and occasional maintenance

Stick with cloud cameras if:

• You want everything working in under an hour with zero ongoing maintenance
• You rent your home and can’t install PoE cabling
• Off-site backup of footage matters more than privacy
• You have 1–2 cameras and the subscription cost is manageable

Getting Started with Local NVR: The Minimum Viable Setup

If you want to try local NVR without committing to a full hardware overhaul, here’s the smallest viable setup:

1. One RTSP-capable camera: A Reolink RLC-810A (~$60) or an Amcrest IP8M-2496EW (~$70). Both are Frigate-compatible out of the box.
2. A host device: An existing mini-PC, NAS, or even a Raspberry Pi 4 with 4GB RAM can run Frigate for 1–2 cameras without a Coral TPU (CPU-only mode). Performance is limited but functional for testing.
3. Docker and Frigate: Frigate’s documentation at docs.frigate.video has a quick-start guide. The initial YAML configuration for one camera takes about 30–60 minutes if you’re new to it.
4. Scale up from there: Once you see how the system works, add a Coral TPU, additional cameras, and PoE infrastructure as budget allows.

The shift from cloud cameras to local NVR isn’t for everyone. But for anyone who’s been paying $10–15/month per camera system and wondering what they’re actually getting for that money — the answer is convenience, not control. Local NVR gives you both the footage and the keys to it. For a complete picture of how video fits into a broader smart home security strategy, our piece on smart home privacy and what your devices collect is a natural companion read.