Magic Leap 2 Hands-On: AR Glasses That Can Dim the Real World

The headset I'm wearing lets me look at virtual objects that don't exist. It's also the first time I've been in person at a product demo event in more than two years. 

As I circle around a very real table in front of me, readouts of mountain ranges and small flitting helicopters on a map pop up as virtual information cards. On the wall behind the table, there are more readouts of data, like giant monitors. My hand-held controller is used as a pointer to click these things. With another click, the room around me goes dark. I've dimmed the real world to see the augmented one with these Magic Leap 2 glasses.

Nearly four years ago, I flew to Florida to see dancing, glowing musical ghosts and fish swimming through the air, all part of a wide-eyed and artistic-focused vision of augmented reality created by Magic Leap, one company among many that's trying to create everyday AR glasses. 

Magic Leap made big promises in 2018, but its hardware wasn't much more than a development kit for AR ideas that still weren't ready for the mainstream. 

Magic Leap's vision has changed since then. New CEO Peggy Johnson, formerly vice president of business development at Microsoft, and before that an executive at Qualcomm, has pivoted the company into a business-focused one, mirroring the goals of Microsoft's HoloLens. Alongside this pivot, there's finally a Magic Leap 2, arriving by the end of this year. This is my first time trying it out.

Promises of Magic Leap's next-gen hardware have been around since, well, my first visit. Then-CEO Rony Abovitz showed me a table with a cloth over it back in 2018, claiming the Magic Leap 2 was under it. I never saw what was really there. But this Magic Leap 2 headset I wore is real, if relatively similar in concept to the original. There's a pair of goggle-eyed glasses, which unfortunately don't work over my own prescription glasses. They still tether to a processor box, which is large and hums loudly as it clips onto my belt. It uses an AMD-based processor this time, which Johnson says is more powerful than the previous Nvidia-based chipset. And there's still a handheld, wireless, pointer-style controller. It's not phone-connected like Qualcomm's upcoming AR glasses: It's stand-alone, but with that hip pack as part of its package.

All of these parts feel upgraded. Plus, the Magic Leap 2's field of view is notably larger, showing more holographic objects in front of me in a taller, wider area that isn't as cut off on the sides. But the most notable thing is a feature I've never seen before on any AR headset or glasses: The lenses can dim the outside world, becoming a pair of blackout sunglasses that AR can display on top of.

Dimming reality

My first demo showed a familiar holographic situation-room-type experience, a common style of demo for AR. (I gathered at a similar round table for a HoloLens 2 demo back in 2019 at Microsoft's Redmond headquarters, too.) The demo, which was made using satellite data and meant to represent a potential use for how AR can represent large-scale map data and multiple virtual monitors at once, turned the round table into a 3D landscape, while the monitors on the walls provided extra simultaneous streams of data. One thing I immediately noticed was that the wider viewing area allowed me to get closer to the table while still seeing everything on it. The taller viewing area also meant that stuff on the walls would still be visible when I was looking at the table.

Field of view has been a big limiter in AR glasses: Existing hardware has a restricted zone where AR effects can appear, and outside of that box, they disappear from view. VR headsets avoid this problem because they have larger fields of view and a goggle-like blindered design, and they don't have to worry about blending the real world with the virtual. Snap's first pair of AR glasses has a very limited viewing area: The first Magic Leap was better, and the Microsoft HoloLens 2 better still. The Magic Leap 2 is the best of all, but still not perfect.

The Magic Leap 2's selective dimming technology can make a whole room, or only parts of it, suddenly darken to offer better contrast or focus on AR effects. Sliders can adjust the dimming intensity, from near blackout to a slight touch.

A second demo showed me another spin on the dimming effect: A giant 3D luxury watch floated in front of me in the office demo room, but with dimming activated it created a halo around the watch and blocked out the real world from behind it. The watch face didn't seem ghostly anymore, and the whole object seemed more solid. My videographer walked behind the floating watch and I didn't see him.

AR already has ways of making virtual objects hide behind real ones, a technique called occlusion (that happens by mapping the real world with depth-sensing cameras). But this dimming effect allows real objects to slip behind virtual ones, too.

Dimming had an uncanny effect at times, almost theatrical, like I was somehow entering a museum exhibit that focused my attention to an object like a spotlight. But Magic Leap's goal for dimming is to make it easier to see essential projections in bright light or outdoors, or to focus and eliminate distractions when using it for important work. The headset is positioned as an enterprise and training tool, not for entertainment this time, but Johnson says that future possibilities for installations and live performance could be next, similar to the way Magic Leap originally birthed a number of fascinating art and theater projects.

Spatial audio extends your senses, sort of

I got one more demo of the Magic Leap 2, this one showing off spatial audio. 3D positional audio has already emerged as a key feature in many VR headsets like the Meta (formerly Oculus) Quest 2, and has popped in headphones like Apple's AirPods. Magic Leap's demo showed virtual mannequin heads surrounding me, talking with and without spatial audio turned on. The 3D audio helps position sound to keep track of multiple conversations better. Magic Leap also showed how the head, turned away from me and behind a virtual wall, could also project realistically balanced voice levels in a room where real people were simultaneously talking to me. 

A little sound alert then pinged at me, and I turned to face it. It was a virtual arrow pointing at a 3D object just out of reach -- what looked like a machine or engine. The point was that the sound cues can direct to things out of the AR headset's field of view. This is exactly what Meta and others have already discussed as a key part of AR audio, but it was a solid reminder that audio, for now, is going to be a part of Magic Leap's vision to make compelling AR that works.

Smaller goggles, AMD hip processor, brand new controller

The Magic Leap 2 is notably smaller than the first version I tried nearly four years ago, and more compact than the HoloLens 2. But it still needs a big processor unit that awkwardly clips onto my pants pocket. A new AMD processor is, according to Johnson, more like a "PC for your eyes" than Qualcomm's phone-based Snapdragon XR2, which is in most other VR/AR stand-alone devices at the moment. But it's tough to judge any performance from a few demos.

The goggles stretch a bit to fit onto my head, the whole headset forming a continuous band. They still have circular lenses, and I still feel like a crazy steampunk astronaut wearing them together with my mask. 

The glasses have improved eye tracking and hand tracking, but I didn't get to try those features out. Instead, I used the Magic Leap 2's new handheld pointer controller. Just like the first Magic Leap, it's a one-handed device with a touchpad and trigger and vibrating feedback, and works more like a simple mouse than an advanced hand-tracking VR controller. The tracking's been rebooted, shifting away from the first's magnetic-based tech and adding its own cameras to make more reliable positioning.

The whole experience, between the hip pack, the glasses and the pointer, feels similar to the original Magic Leap, but better. I still wonder whether hand tracking and eye tracking will be as reliable. Johnson admits it's not as accurate as using the pointer, so I have questions.

I haven't even discussed the price of this headset, because Magic Leap hasn't confirmed it yet. But I was told it's more expensive than the first Magic Leap, which was already $2,300. The HoloLens 2 is a $3,000-range headset as well, and this hardware isn't even trying for the nonbusiness sector right now.

It doesn't work over my glasses, though

The Magic Leap 2, like the Magic Leap 1 (and many other AR glasses), won't work on top of my own glasses like the HoloLens 2 can. This means you need a prescription insert that slides into the goggles. At my demo, I was promised there would be one, but unfortunately their inserts ended at -5 and my vision is over -8. Luckily, I brought contact lenses, which I wore to demo the headset.

That's not ideal, but according to Johnson, the closer distance to the eyes is necessary to get that larger field of view. Apparently, Magic Leap is making prescription inserts to fit my vision… but I haven't seen them yet.

Are these the leap, or just a step?

I got the feeling that the Magic Leap 2 is a small but key step for a company that, like many others, is aiming to solve the riddle of AR glasses. No one's cracked it yet. Meta, Snap, Microsoft, Qualcomm, Niantic, Apple, Google, Samsung and many more could be ready to dive into it, but while VR headsets are an actual everyday product now, AR glasses aren't. 

Johnson sees the Magic Leap 2's much narrower targeting of industries like manufacturing, medicine and defense as a way of testing the waters while the headset can eventually be made smaller and lighter, without needing that big processor hip pack: "I do believe at some point down the road, we'll circle back to consumer, but it's going to be when the devices are more fully integrated and lightweight, and we can make the field of view even bigger."

As other companies try to solve AR glasses slowly, and Microsoft seems ready to pivot away from its HoloLens 2 to possibly phone-connected devices, is Magic Leap's tech going to seem advanced enough, or trapped on its own island? 

Magic Leap isn't making its own apps this time, though, nor a proprietary OS. The headset's software is Android Open Source, and it looks like the company is hoping developers find it easier to create for that than the Magic Leap 1. "This is about launching not just a piece of hardware, but an entire ecosystem around it," says Johnson. "And frankly, with Magic Leap 1, we could have done a better job with that." Johnson promises that the hardware will work with phones, at least to share 3D models with other collaborators (Unity compatibility promises that 3D models will be easy to import).

We'll know more when the hardware arrives later this year, but ideas like the headset's reality-dimming mode seem like the beginning of a shift in AR glasses design. The hardware is clearly better this time, and may be the best I've ever seen. Will the software and enterprise success follow?