Magic Leap One Developer Review – An Ambitious Headset with Untapped Potential

It’s real and it’s here — but does it live up to the hype?

21

There’s a lot that can be said about the Magic Leap One. It’s trying to do a ton—eye-tracking, hand-tracking, 6DoF controllers, real-time meshing and a number of other features that haven’t been seen in a mobile MR device before. And although its OS and apps don’t fully utilize the tech available to them, Magic Leap One is an ambitious, well-made, but imperfect MR devkit that doesn’t quite live up to the hype, but is still the most complete and affordable mixed-reality (MR) computer out there.

Lucas Rizzotto

Guest Article by Lucas Rizzotto

Lucas Rizzotto is an award-winning Immersive Experience Designer, Artist and Creator. He’s part of the team behind Where Thoughts Go, and has developed several other MR projects.

You can follow him on Instagram, Twitter, Facebook,or contact him through his website. You can also sign up to his mailing list and support his work on Patreon.

Part 1: The Hardware

If you’re not familiar, a mixed reality computer is a device that interacts with your senses and understands your physical space. It uses that information to allow you you to place digital objects in your world that feel real.

The Magic Leap One consists of three elements: a light-pack, a headset and a controller — all of these components are beautifully designed, with colorful lights coming from unexpected places.

Meet the trio | Image courtesy Magic Leap

The headset is lightweight, easy to put on and comfortable, and that allowed me to use it for hours on end without tiring up my neck. It usually sits well on your head if you don’t move it too quickly, but can fall out of place if you perform extreme actions such as running. I also felt some heat on my forehead after a few hours of use due to the placement of one of the processing units.

The controller is intuitive, comfortable to hold and works well in most cases, although I’ve noticed some considerable delay in the 6DoF tracking. Magic Leap says this is unusual, but I’m still waiting for a fix.

Photo by Road to VR

The light-pack carries the bulk of the processing power for the ML1, so it is the heaviest object of the bunch. It’s actually quite heavy when you hold it on your hand, but the weight disappears when you clip it in your pocket — this has some trade-offs, however: the pocket-centered design ignores the possibility that users may be wearing things like a dress or more elastic pants that wouldn’t provide a strong grip.

Adding a bit of complexity to the UX are wires that can tangle in the user’s clothes and limbs if they’re not careful. This makes trying out the Magic Leap One a little bit awkward as the user is set up for a potential accident. For example, while I was stress-testing the device the light-pack fell from my pocket at one point due to the type of pants I was wearing and the thick wires briefly choked me as they got stuck around my neck and almost brought the headset down.

Thankfully, Magic Leap sells an optional $40 strap that allows you to wear the light-pack over your shoulder. This option was universally preferred by me and my colleagues due to its friendly and context-agnostic nature, but it should have been a part of the original packet.

Update: While the strap is listed as a $40 item on the website, I was just told by Magic Leap that they’ve been making them free for all those who purchase it with their device. This was unclear at the time of my purchase and I ended up not getting one.

Magic Leap One’s display is great, but a bit less crisp and legible than 2016’s HoloLens. ML1 supports a higher field-of-view than the HoloLens, but it seems to have done it by sacrificing image resolution, which can make text cloudy. Its lenses are also considerably darker to compensate for lack of image brightness, which can sometimes remove you too much from the environment.

The display can showcase some occasional visual artifacts at times, the strangest one being an “oily” texture that appears when you’re looking at brightly-colored virtual objects. Artifacts are nothing new in the realm of Mixed Reality, but they will start showing up if you start looking.

Magic Leap has for a long time talked about the “multi-focal light-field” properties of its display, and chances are you won’t notice it in action. Apparently it’s one of those features that is working if you’re not noticing it, so it’s hard to make an assessment of how much of a difference it actually makes.

This video is perhaps the best demonstration of how reading through the display feels like, minus some absent visual artifacts.

Issues aside, most components of the product come together well enough to make the ML1 a very capable platform for developers to build experiences for. The Magic Leap has catching up to do in some areas of the headset, but it was a joy to play with at first pass.

Part 2: Spatial Meshing & Tracking

Magic Leap One’s real-time meshing is solid, responsive and relatively fast, being comparable to the higher-end tracking solutions we see today.

While the meshing process isn’t perfect and can miss reflective surfaces, as well as black objects, this stands as a common issue in AR in general. Developers must pick the rooms they choose to exhibit this in carefully to guarantee an ideal experience.

The Magic Leap One also had a hard time recognizing rooms that were already previously scanned and constantly asked me to re-scan my environment every time I booted up the device. This would be easy to overlook, but it messes with one of the coolest things in MR: is digital object persistence (virtual objects you place in the real world staying where you placed them in between sessions). I hope this is something they improve in an update.

For the most part, the tracking of virtual objects feels solid with only the occasional jittering, which was never too noticeable in my experience with the device. All and all, tracking and real-time meshing of the Magic Leap One works well and lags only a bit behind the HoloLens in terms of accuracy.

But what about the diverse forms of input tracking?

Eye-tracking has worked pretty well in the small tests that I’ve done and is one of the things I’m most excited for. Eye-tracking offers so many exciting new possibilities for telling stories and creating interfaces (a topic I explored in a separate article) and I’m excited to see how people turn this into a key aspect of this device.

Hand-tracking is also here, but it can feel pretty limited if you’ve played with other superior forms of tracking like the Leap Motion. It’s more about tracking entire hands than it is about tracking individual fingers, although it does track some movement of the thumb and the index finger.

The 6DoF controller works through magnetic tracking and it’s generally good, but as mentioned before, it has presented me with some positional tracking issues that I’m hoping will be resolved.

All things aside, the fact that the device has eye-tracking, hand-tracking and a 6DoF controller is impressive and it’s a solid enough toolbox for people to create natural interfaces with, especially when you add voice recognition on top of it all. Now we just have to see how developers will choose to use it and how these features will improve throughout the year.

Continued on Page 2: The Operating System

1
2
3

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. See here for more information.


  • dk
    • Lucas Rizzotto

      Hey! I fixed it in the article, so this one is corrected. Thanks for the heads up ( :

  • NooYawker

    Can you do a full review of the Hololens? Does Magic Leaps device do anything better than the 2 year old Hololens?

    • Mac

      It sounds like it has more input options, eye tracking, a wider field of view and a more powerful cpu/gpu in the pocket computer. I was surprised about the eye tracking, I had no idea they planned to include this. It’s clearly an early dev kit of what’s to come for AR.

      • JJ

        yeah but every developer says that ML doesnt use the eye tracking for anything, sooo its possible but nobody knows if its any good yet. The FOV is slightly larger, by a small fraction while the details are much worse especially in the alignment of the RGB. Not really more input options tbh, maybe plug n play for non programmers, but with the hololense being in the windows environment meant that I had various ways to hook up all sorts of devices.

    • Anonymous247

      You should compare the 3D video. Can Hololens open 8 3D movies at once ?

  • beestee

    It is nice to finally have this thing grounded in reality, and by the looks of these impressions it seems like there is a lot of potential lurking within it. Show me a few killer applications and a UX update that thoroughly embrace all the tech available within the hardware and I would consider a purchase even at it’s current price.

    • jj

      any potential you see in this, was also in the hololense, 2 years ago.

      • beestee

        Eye tracking, multi-focal display, and a 6 degrees of freedom controller?

        • jj

          those arent game changers or enough innovation to set this apart from a product that came out two years ago… and like others and i have said nobodies used the eye tracking. literally the only demo they have is blink where they can tell if you’re blinking or not.

          • beestee

            But that is hardware that is in the ML1 that is not in the HLv1 is it not? You said the HLv1 had all of the same potential two years ago…

            As I said above, I would only consider purchasing if the hardware was properly leveraged by the software, which currently it is not.

          • JJ

            yeah but none of those hardware features make a difference. there are only 2 multi-focal displays and ML themselves said 6 is the ideal minimum…. plus the 6dof controller is via magnets and subpar compared to other 6dof controllers so to me its not the same. and the eye tracking helps with the PID but thats it because the multi-focal feature barely works.

            plus my hololense remembers what room im in upon tarcking while the ml appears to never do so.

          • beestee

            I think it is too early to say they would make a difference or not before they have been developed for.

            Given some time, I bet the spatial mapping will improve.

            There has to be a reason it is $700 cheaper, right? It can’t all be unicorns and rainbows.

          • MosBen

            I would understand taking the position that if you already won a Hololens it probably doesn’t make sense to also get a Magic Leap One, but that can’t apply to most people. If someone is interested in buying an AR headset today and is willing to spend a significant amount of money, it seems like the Magic Leap One is the best option available today. Granted, I’d tell such a person to wait until CES 2019 to see if Hololens 2 is announced, just as a comparison, but whether eye tracking, 6DOF, a slightly wider FOV, etc. are game changers or not, they’re clearly upgrades over what Hololens offers.

  • Lucidfeuer

    I’d like to try one, but I already know I’ll get the same impression as the HoloLens because well…that’s exactly what the Magic Leap (if it ever existed) would be no more than that.

  • Foreign Devil

    thanks for the review! Nice to see something substantial about this.

  • Gato Satanista

    The thing is: Magic leap has not fulfilled their promise. This is clearly not a consumer-ready product. Buy it if you are a serious developer who wants to develop some custom AR sollution/service for large companies. If you just want to play games or use some professional apps, stay away for now. Unless you have a pocket full of cash and doesn’t care to buy a new dust collector (for now) device.

    • Get Schwifty!

      Exactly….

  • thewebdood

    Great write-up Lucas! Have you played with the actual APIs yet?

  • thewebdood

    I’m with you on the FOV bit, BTW. Over-rated metric that people focus on (sic) for no reason. It’s a solved problem that will eventually be a non-topic.

    • benz145

      I disagree. While there is a point where increasing it further mostly serves to enhance immersion, there’s a minimum threshold which is key for real usability, where users can mostly interact with the augmented world as they would the real world, without having to re-train their ingrained proprioceptive model to account for objects disappearing outside of the active engagement area.. 90 degrees is thought by many to roughly meet that threshold.

      Not only that, but assuming it will be solved is an oversimplification in my opinion. It’s actually a very complex problem that right now has no obvious answers without lots more R&D.