Reolink’s New Floodlight Camera Uses Sensors and AI to Detect Where It Can’t See

Reolink rolled out a new smart home security camera at IFA 2025 that the company says can see beyond its dual camera lenses. It’s called the TrackFlex Floodlight WiFi, and it looks kind of like the Reolink Elite Floodlight WiFi, a camera I recently reviewed, but with a ball-shaped camera housing that rotates to see things that three sensors above them have detected.

This sensor-based approach gives the camera a 270-degree detection range at any given time, according to a press release that Reolink emailed to Gizmodo. The three motion sensors wrap around the front and sides of the unit. A representative I spoke with at Reolink’s IFA booth suggested installing the camera on the corner of your home, surveying a driveway that stretches from the street to a garage farther back, a scene impossible for one fixed-view camera to cover—the TrackFlex could detect that a car is entering the driveway, swivel to face it, then watch it as it drives to the garage.

Like other Reolink cameras, this one stores recordings locally, either with microSD cards (up to 512GB) or Reolink’s NVR and Home Hub devices. Saving videos to a network-attached storage (NAS) device via FTP is also an option. This is Reolink, so expect to be assaulted with options in the company’s app.

As for the TrackFlex’s dual cameras, they’re not recording one broad field of view like the Elite Floodlight WiFi. Instead, one is a standard wide view while the other is a 6x zoom, capable of capturing a lot more zoomed-in detail than the wide lens can. The two floodlights looked exactly like those of the Elite, and can articulate to point up or down, or to bathe a wider area in light. They also offer the same brightness and temperature adjustment as the Elite.

The camera uses the company’s new local AI system called ReoNeura Core, which enables the TrackFlex to do the same sorts of natural language video search that we’re seeing with a lot of other connected-camera AI systems. (See SwitchBot’s new AI Hub.) So, if you want to search your locally stored video for a moment, like when a person with a brown shirt walked into your garage last, you can do that. The Reolink rep I spoke with took me to a pair of displays that showed me what was happening behind the scenes.

On the left, it was a view of what a user might see while using the event notifications screen; you see the camera’s live feed. While on the right, a list of events with descriptions that were mostly pretty accurate—there were people interacting at a convention, and others were walking around in the background—but it got some details wrong. We weren’t at an outdoor event, for example (although it was quite bright in there).

Meanwhile, on the right, the screen showed individual characteristics of the things the model was identifying. Seeing everyone broken down by their attributes had a very police-station-surveillance vibe. ReoNeuro identified one person as a middle-aged male wearing a green, short-sleeve shirt, along with a hat and a bag. In the bottom panel, another middle-aged male—although that person definitely looked younger than me, and I’m still in my prime, I swear—is identified as wearing a long-sleeve shirt with pants, both blue, and as having short hair. All of that detail is effectively keywords for your searches later, and they all seem like things you’d type if you’re looking for specific events that you know the TrackFlex recorded.

But woof, it’s more than a little unnerving to see this in action, and it felt a little off-key to have a Reolink rep so proudly showing it to me. It’s great that this is all happening on device, as I’d rather that than have it happening in a cloud server over which I have no control. It’s convenient and there’s no doubt that I want the convenience of casually searching my footage, but that sure didn’t stop seeing how the sausage is made from giving me the willies.