Magic Leap - sad trash
The title of this article [in the original Magic Leap is a Tragic Heap - approx. lane.] carefully chosen, and not for red word. I wish the best for VR and all other technologies in the continuum between reality and virtuality , including Magic Leap. Unfortunately, the current proposal is a tragedy in the classical sense, especially considering that their abundant funding and thoughtful advertising sucked all the air out of the AR room. This is not so much a functional developer kit as a hype gadget that is almost impossible to use in a meaningful way, and many of their design decisions seem to be due to this reality. He fulfills almost none of the promises that allowed them to monopolize financing in the AR investment community.
At the moment there are many general reviews of ML1, so I will focus on a few specific points that are not so widely covered. If you want to read the full review, this video from Tested is great for a start. If you want to see the insides and how they work, I helped iFixit disassemble my ML1 .
Tracking bad. Otherwise you will not tell. The controller responds slowly, floats everywhere and practically does not work near large steel objects - good for a wooden house made of sticks, but bad for any industrial environment. Magnetic positioning works with difficulty, but this is probably the worst implementation I've seen in public products. VR enthusiasts who are familiar with the Polhemus (magnetic tracker) systems , the Razer Hydra or the ever-elusive SIXENSE STEM , know where the bottom bar is. From the Magic Leap Developer Guide: “6DoF tracking works stably with slow and moderate movement. It also quickly recovers and moves in case of sudden or unexpected movements (for example, boxing or fishing-like movements). ”
I understand that Magic Leap wanted to make a controller that works without a headset in direct line of sight or a clumsy protrusion to track the position, but this is a terrible compromise, especially for developers who need just a working controller - there are good reasons why no other company went along this path. Controller-type gimmicks behind the back are funny, but ML1 could and should have used a different tracking system. Several other companies have managed to implement optical tracking without funding billions of dollars like Magic Leap, and even if they can't bring tracking to mind, they will definitely be able to use the system with an external guideline. In the end, many of the limitations of the Magic Leap software and user interface seem to be caused by a poor controller.
Another strange difference from competing devices is the non-clickable trackpad. The Steam controller, the HTC Vive sticks, the Oculus Go, the Lenovo Mirage Solo, even the Playstation 4 controller - all have a clickable trackpad, and game developers rely heavily on this feature. Its absence in practice means that selecting using the touch panel requires either lifting the finger and pressing (terrible accuracy), or pressing the trigger while holding (also terrible accuracy). It also means that the trackpad cannot be used to emulate buttons or another selection scheme. Everyone in the industry uses components from ALPS (great company, by the way), it was enough just to call them and say that we need a custom trackpad with fun RGB LEDs.
A final note about controllers: unlike most magnetic tracking systems, here the transmitter is located in the controller. This means that a giant metal core with a copper winding hangs directly above the trigger. In order to balance the weight, Magic Leap needed to install metal weights at the bottom of the controller. Now the controller initially feels like a “premium”, but really loses in long-term ergonomics.
They call it Lightpack. This is basically the filling of a tablet computer in a fat hockey puck that you wear on your belt. Of course, this is still the best part of the Magic Leap kit, a score of five plus! I expected the company to make a fashionable gadget and, for the sake of beauty, transfer the graphics subsystem and battery to the head-mounted display, but some group of sensible people seemed to admit that placing the heaviest components on the most sensitive part of the body is a bad idea if people need to wear device for a long time. This is already a topic for a separate article, but the data shows that it should be VERY tough to reduce the weight of the head-mounted display. The transfer of iron to the belt also allows the use of much more powerful chips than if necessary to squeeze them into a head-mounted wearable device.
The cables are strong, and the pressure from the back of the head actually helps to balance the balance a bit. Developers should have provided a replaceable battery, although no one is going to use the ML1 long enough for it to matter to anyone other than collectors in order to preserve the AR and VR history.
They call them Lightwear. It was this part that for many years caused a stir, with endless talk about “photonic light field chips”, “laser fiber optic scanning displays”, “projecting a digital light field into the user's eye”, and with the Holy Grail - solving the vergence-adaptability conflict , which for decades haunted head-mounted displays (vergence accommodation conflict: focus problem caused by asynchronous eye movement - approx. lane). In other words, the focus of the eyes is always accompanied by the convergence of the image that Magic Leap presented as a critical element for eliminating the "permanent neurological deficit"and brain damage. This is even more important for AR than VR, since here you need to mix digital elements with always the correct real elements.
TL; DR: The supposed “photonic light field chips” are simply waveguides paired with reflective LCoS color displays and LED backlighting, the same technology that everyone else has been using for many years, including Microsoft in their latest generation HoloLens. ML1 is not a “light field projector” or a display by any generally accepted definition, but as a bifocal display solves a vergence-adaptability conflict only in contrived demonstrations where all elements of the UI and the environment are located in one of the two focusing planes. The discrepancy is observed in all other cases. Similarly, a broken clock shows the correct time twice a day.
In more detail: ML1 uses six waveguides stacked on top of each other, three for each RGB color channel in two different focal planes. You can think of it as a bifocal display — a display system that shifts the focus of the display between two different values based on eye tracking without any variability between them, unlike constantly changing multifocal (vari-focal) displays, such as Oculus Half-Dome or displays the true Nvidia light field . I do not have accurate measurements yet, but it seems that the near plane is focused at about 0.75 meters, and the far one - at about 5 meters. If they adhere to this technology (and I have not seen any signs that they can do anything else,especially loudly advertised fiber optic display ), then each additional focusing plane will require additional sets of waveguides and an unacceptably high frame rate (each plane consumes at least 60 Hz from the time budget). I do not think that this can be done within reasonable weight, image quality and cost).
More than one plane is good, don't get me wrong! This allows developers to avoid extreme inconsistencies with very close or very distant objects. Nevertheless, unwinding the hype and monopolizing investments with promises that cannot be fulfilled is bad for the entire XR industry, and not just for Magic Leap. Equipment manufacturers are required to clearly inform the developers about the capabilities of their equipment, even if these capabilities do not match their preferences.
As for the other parts: tracking is good compared to most other products in the AR / VR industry, but worse than most leading products, including Hololens. Expect jitter (jitter) in perfect surroundings. If you want to compare, imagine a cross between PSVR and Rift. Meshing-system is good, but not as fast as Hololens. It is quite similar to the fact that they produce companies that have financing by several orders of magnitude less, such as Stereolabs .
In addition to bifocal functionality, image quality is acceptable. Have you seen Hololens? Here, about the same, but with a little more FOV. The rainbow artifacts are slightly worse because of the large number of stacked waveguides, and the black level is slightly better, but the Magic Leap is quite comparable with other players. Despite the high power consumption, the head-on display remained beautiful and compact (seriously, it is terrible to touch the magnesium shell in a warm room), the display is too dim for use outdoors. This is a shame, because its transparency is about the same as that of sunglasses - the material is not quite for indoor use. How does eye tracking work? It is difficult to say, because it is not used anywhere. This is not a good sign.
A true achievement would be a significant and useful increase in FOV, which Magic Leap could have done if it had put user convenience in the first place, and not the size of the device. For an example, check out Dreamworld at 90 degrees - the tracking is absolutely incomparable, but the experience is very exciting.
Magic Leap says that it “built a completely new operating system” called LuminOS to use its “spatial computing system” . In fact, this is just Android with custom modifications — most people use this approach when they want to declare that they have built an entire operating system.
I will be brief here. I hope that in the future, Magic Leap will do cool things, but the current UI is essentially the Android Wear watch menus floating in front of you. Menus are made of flat panels with which you can interact only through the previously discussed non-clickable trackpad. Eye tracking and controller rotation / position are ignored, as are head tilts. You can scatter windows of Windows 8 style applications everywhere, floating in space or even attached to walls! This is elegant, mostly useless, and this is exactly what Microsoft began to demonstrate about three years ago. Here, the worst elements of the phone UI are crammed into the most bizarre elements of the VR user interface, and I very much hope that developers will correct the situation in the near future.
Sales and developer adoption rates
The first few days after launch, the Magic Leap ordering system was very simple. I collected several order numbers from friends, compared order times and numbers — and quite confidently predicted sales in the first week. Unfortunately, when I tweeted about it, they changed the ordering system. Based on the available data, they sold about 2,000 units in the first week, with a very strong bias towards the first 48 hours. If I had to guess, I would put total sales at less than 3,000 units at the moment. This is regrettable for obvious reasons - I know more than a hundred people with ML1, and almost none of them is an AR developer. Most of them are technical leaders, “opinion leaders” or enthusiasts who work in the industry but do not plan to actually build AR applications. This was a big problem at the dawn of VR industry development, even with many tens of thousands of developers among hundreds of thousands of development kits sold! In the case of ML, the problem is several orders of magnitude worse.
Magic Leap was supposed to be a truly revolutionary product to justify the last few years of hype and excessive funding. They had a fairly solid gadget, but not even close as promised, and with several flaws that prevent it from becoming a widely accepted useful tool for developing AR applications. This is bad for the XR industry. The product is slightly better in some respects than Hololens, in other respects worse. By and large, this is a small step forward compared to the most advanced product three years ago - this is more Hololens 1.1 than ready for the mass market AR 1.0. The latter cannot be done without technological progress, and it seems that other companies will make this progress. Of course, there is a chance that Magic Leap will surprise us: what if a truly revolutionary product will be shown later!
Above is a characteristic photo from the Magic Leap article published a couple of years ago in Wired magazine when they fanned HYIP about fiber-optic scanning displays. Do you see fashionable, high-tech luminous threads? They do nothing. These are just electroluminescent wires. It looks great for casual observers, but does not stand up to scrutiny from competent people. If you want to decorate with such wires a suit, a gaming computer or a multi-billion dollar machine for inflating the HYIP, then a good assortment of light lines for $ 20 can be found here .
UPD. Approximately 45 minutes after the publication of the article, the executive director of Magic Leap decided to share his love for the television series Avatar: The Legend of Aange. He also mentioned that the stereoscope was invented in 1838, very interesting!