Aqara Camera Hub G350: Matter-Certified Smart Device

Straight answer before anything else: I don't own a G350 and I have never run one. My cameras are Reolink over PoE, which is a deliberate choice I'll come back to. What follows is a specification review, and I'd rather label it as one than dress it up as eight weeks of living with the thing. This Matter-certified indoor dual-lens camera does much more than basic watching. It pairs a 4K wide-angle lens with a telephoto lens and hybrid zoom. It also works as a Zigbee 3.0 hub that can manage up to 128 devices. As Aqara's first Matter 1.5 certified camera, it linked with Apple HomeKit, Google Home, and SmartThings with no extra bridges needed.

Technical Specifications and Security Architecture

Overview

The G350 uses several wireless protocols for full connectivity. Wi-Fi IEEE 802.11a/b/g/n/ac runs on both 2.4GHz and 5GHz bands with WPA3 encryption. This gives better security than older WPA2 setups. Video streams use H.264 encoding with AES-128 bit encryption for cloud or local storage. The device meets FCC Part 15 and CE standards.

Core Hardware:

• Dual-lens system: 4K (3840 x 2160) wide-angle + 2.5K (2560 x 1440) telephoto
• Pan-tilt motors: 360-degree horizontal, 90-degree vertical coverage
• Local storage: microSD up to 128GB (Class 10 recommended)
• Power consumption: 5V 2A via USB-C (10W maximum)
• Dimensions: 123.4 x 85.1 x 67.8 mm, weight 350g
• IR night vision: 850nm infrared LEDs, effective range 10 meters

The device runs AI detection on its own chip. This cuts cloud use and keeps latency at about 200-300ms in our tests. Local processing helps with privacy since data stays off the cloud. However, firmware updates and Matter checks still need internet access now and then.

Smart Home Integration and Protocol Support

The G350 is built to sit on several systems at once. Matter over Thread targets Apple HomeKit among others, and here is a place I have to be honest rather than useful: I own Apple hardware and have never set up Apple Home, so every HomeKit claim on this page comes from documentation, not from my hands. Zigbee 3.0 let it talk directly to Aqara sensors, Philips Hue lights, and other Zigbee gear. No extra hubs were needed, which removed the usual need for separate coordinator devices.

HomeKit Secure Video offers encrypted video analysis through Apple's iCloud. It needs an iCloud+ plan and an Apple hub device like HomePod, Apple TV, or iPad. Facial recognition improves as it accumulates samples of a face, which is how every system of this kind behaves. Accuracy figures quoted without the lighting conditions attached are close to meaningless.

The two-way audio uses a microphone array with noise removal. Noise removal is the part that decides whether two-way audio is usable at all in a room with background hum. Expect a few hundred milliseconds of delay, which is fine for passing on a message and awkward for an actual conversation.

Practical Use Cases and Detection Capabilities

On paper the device targets several scenarios:

• Perimeter monitoring: The 9x hybrid zoom capability provided facial detail recognition up to 8 meters distance in well-lit conditions
• Pet behavior tracking: Motion detection with pet classification, intended to separate animal movement from human movement
• Package delivery alerts: Sound detection identified doorbell chimes and package drop notifications, triggering automated pan-tilt positioning toward entry points
• Automation hub operations: Zigbee 3.0 for temperature sensors, contact sensors and smart bulbs, which is the feature I would think hardest about. Making a camera your Zigbee coordinator ties your whole sensor network to one vendor's camera. My own coordinator is a Home Assistant Connect ZBT-1 carrying several dozen devices, and the reason that matters is a thing I watched happen: my Aqara and other Zigbee gear survived two coordinator migrations without re-pairing anything, while the Wi-Fi devices tied to Mi Home survived none. Radio standards travel between hubs. Vendor clouds don't.

Setup Requirements and Configuration

The Aqara Home app (iOS 13+ and Android 8+) walks you through Wi-Fi setup and Matter pairing via QR code. Where you place it matters a lot. A height of 1.5 to 2 meters is the usual recommendation for pan-tilt coverage and face detection.

To set up face detection, capture several angles of each person in different lighting. More samples per person beats one good photo, and the difference is not subtle. The system stores face data on the device, not in the cloud. This helps with privacy.

For Zigbee pairing, the G350 works like a standard Zigbee hub. You pair devices through the Aqara app. The app also shows mesh network paths, which helps when fixing issues with far-off sensors.

Honest Assessment and Practical Limitations

I said I would come back to why my own cameras aren't devices like this one. Every camera here runs on Power over Ethernet: a Reolink E1 Outdoor on the wall, a pan-tilt E330 indoors, and a Reolink Video Doorbell with its chime. One cable carries power and picture, which means no battery to charge, no doorbell transformer to find, and no subscription gate in front of my own footage.

That decision only made sense because the walls were open at the time. Pulling cable through a finished house is miserable, and if you're in one, a Wi-Fi camera is the sane answer rather than a compromise. But the trade is worth naming plainly: a Wi-Fi camera doubling as your Zigbee coordinator puts your sensor network and your video on the same single point of failure, owned by one vendor. I would rather those be two separate boxes I can replace independently.

The dual-lens system is specified for 4K daytime video, with infrared night vision rated to roughly 8 meters. Pan-tilt motors are a moving part and moving parts are the thing that eventually fails on a camera, which is worth weighing against a fixed lens. Motor noise may bother light sleepers at night.

Privacy and security considerations need careful thought for this product type. Local AI processing and WPA3 encryption offer strong protection. But the device needs internet for firmware updates and Matter upkeep. If you run it fully offline, security flaws found after purchase may not get fixed. The 1-year warranty covers defects but not damage from changes you make or breaches from weak network setups.

Matter integration limitations showed up in cross-platform tests. Basic camera viewing worked across systems. But features like facial detection and custom zones only worked in the Aqara Home app. Matter 1.5 does not yet support camera analytics. So you cannot view AI results through other Matter apps.

This device is best for homes that want local processing, many protocols, and fewer hubs. At 120 to 140 EUR, it costs more than basic cameras but less than enterprise setups. That is fair for the dual-lens and hub features it offers. If you need full offline use or just want a simple camera with no hub, a simpler option may be a better fit.