Varjo has announced its latest high-end enterprise headset, the XR-4. The company is going all out on the headset’s mixed reality capabilities, saying that the view of the outside world as seen through the headset is “indistinguishable” from how the world appears with your own eyes.

That’s a seriously lofty claim, but Varjo hasn’t made a habit of hyperbole. We’ll wait until we can actually look through the headset ourselves, but clearly the company is confident in what it’s built.

But not every XR-4 headset will have what the company says is a passthrough view with a whopping 51 pixels per-degree resolution. Only the more expensive variant—the ‘Focal Edition’, priced at $10,000—will reach that peak visual quality thanks to an eye-tracked auto-focus system which adjusts the cameras to keep the world in sharp focus wherever you’re looking around the scene.

A look at the mixed reality view through Varjo XR-4 (captured through the headset’s cameras, but not lenses)

The considerably less expensive XR-4 standard edition, priced at $4,000 nixes the auto-focus system and delivers only 33 pixels per-degree (though this is still very high resolution passthrough compared to the majority of headsets you can buy today).

Achieving such a high resolution mixed reality view has required that the headset also includes some pixel-packed displays. With XR-4 the company is also moving fully to inside-out tracking as the default, along with built-in audio, and the company’s own controllers.

Let’s look at the spec breakdown here:

Varjo XR-4 Specs, Price, Editions, and Release Date

XR-4 Focal Edition

Visuals
Display 2x mini-LED (200 nits with local dimming),
96% DCI-P3 colors
Resolution 3,840×3,744 (14.4MP)
Pixels Per-degree (claimed) 51
Refresh Rate 90Hz
Optics Full-dome aspheric
Field-of-view (claimed) 120° × 105°
Pass-through view Yes (51 PPD)
Optical Adjustments IPD (automatic)
IPD Adjustment Range 56–72mm
Input & Output
Connectors 1x DisplayPort,
1x USB-C
Input XR-4 controllers
Audio In-headstrap speakers,
3.5mm aux port
Microphone Dual-microphone
Weight 665g (headset) + 356g (headstrap)
Sensing
Headset-tracking Inside-out (no external beacons),
SteamVR Tracking (external beacons) [optional]
Controller-tracking Headset-tracked (headset line-of-sight needed)
Eye-tracking Yes (200Hz)
Expression-tracking No
On-board cameras 6x tracking,
2x RGB (20MP) eye-tracked auto-focus
Depth-sensor LiDAR (300 Kpix)
Price
MSRP $10,000

This is the XR-4 ‘Focal Edition’ which includes eye-tracked auto-focus passthrough cameras to achieve the claimed 51 PPD passthrough resolution. The ‘standard edition’ XR has nearly identical specs, except without the auto-focus camera, the company says the headset’s passthrough resolution drops to 33 PPD.

As for pricing, while the Focal Edition is seriously pricey, the standard edition is actually cheaper than its predecessor; the XR-3 was priced at $5,500 for the headset alone, plus a required $1,500 annual support charge. XR-4 standard edition meanwhile is priced at $4,000 and does not require an annual support charge.

Varjo is also making two ‘Secure Edition’ variants of the XR-4 (which mirror the specs of the Focal Edition and standard edition, but these are TAA compliant and can be ordered without any wireless radios (this is for particularly niche applications where data security is critical, for instance in military applications). These are priced even higher, at $8,000 and $14,000 respectively.

The headset-tracked controllers are made in partnership with Razer, which has previously dabbled with various VR accessories.

Varjo says the XR-4 will begin shipping by the end of 2023.

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.


Ben is the world's most senior professional analyst solely dedicated to the XR industry, having founded Road to VR in 2011—a year before the Oculus Kickstarter sparked a resurgence that led to the modern XR landscape. He has authored more than 3,000 articles chronicling the evolution of the XR industry over more than a decade. With that unique perspective, Ben has been consistently recognized as one of the most influential voices in XR, giving keynotes and joining panel and podcast discussions at key industry events. He is a self-described "journalist and analyst, not evangelist."
  • fcpw

    Meh so what- without the ecosystem backing of a Meta or Apple these products are islands unto themselves. Neat tech with zero future.

    • ViRGiN

      Yeah this poses as some revolutionary product for the ‘enterprise’ but it’s all just pumping the baloon.
      The software is still not catching up to what hardware can provide.

    • Cless

      … huh? Nobody that buys this wants Meta or Apple integration though, so why bother…?

    • psuedonymous

      For buyers of these headsets, that’s a nonissue. These are not purchased by consumers to play with CotS software, these are used to develop bespoke applications for this particular hardware setup.

      • ViRGiN

        That’s what every company unable to tap into consumer market always say – take Lynx for example. While I have infinietly more high confidence in Varjo actually partnering in serious ways with enterprises, we still don’t have a true idea what these headsets are really used for in truly productive ways.

    • Mike

      Where did you get that it doesn’t have compatibility with an ecosystem? The article says it can use SteamVR base stations, which implies it can run on SteamVR, which means it can run any PC VR software.

    • Dave

      If they have a reasonable customer base or shareholders backing there products, then what does it matter, they’ll continue to inovate and produce highend headsets. I do believe with controllers and inside out tracking, it opens up PCVR, so I don’t really know what you are talking about. Please explain to me why it matters to not have an eco system when customers are developing there own applications and the headset can tap into the biggest eco system of them all, Steam Desktop and SteamVR.

  • Bob

    The specifications are most definitely impressive. However, the form factor and general bulkiness of the design isn’t a step in the right direction. Obviously, the headset isn’t intended for the mass consumer but Varjo isn’t a huge company with unlimited pockets (much like Apple, or Meta) so they cannot afford to advance both function and form; it’s one or the other. And at least in terms of form, it appears to be a regression even though the technical specifications have improved.

    Until these companies have access to holographic lenses with curved displays or some other exotic lens and display technologies that allow wide FOV and absolute clarity edge-to-edge, there will continue to be relatively big and bulky headsets. Even the Apple Vision Pro, despite its ski-goggle like design, is actually quite a chunky device which is made more evident by users with smaller heads.

    • Christian Schildwaechter

      The Varjo XR-3 made up for bulk with excellent weight balance and adjustability, allowing it to be worn comfortably for a long time. But the high mass increases inertia during fast head turns, causing a special type of physical latency. This (and the price) made it a bad choice for shooters that permanently require quickly turns, or gaming in general.

      For a typical XR-3/XR-4 use case like virtual prototype testing that’s not an issue. Testing often involves something hand operated in front of the user, moving mostly the eyes, with moderate, slow head movements, needing a presentation as close as possible to the final, physical product.

      The very small and light form factor of the Bigscreen Beyond is a good direction for high end VR games, even at the cost of inside-out/eye tracking, stand-alone, integrated audio etc. The seemingly opposite XR-4 (Focus) is a good direction for high end XR simulation, as the added features/weight allow to simulate work environments with a combination of now almost indistinguishable physical and virtual interfaces. And the heavy and unbalanced Quest 3 is a good direction for affordable, low resolution/power consumer HMDs. There are different use cases for VR, with quite different requirements, still justifying seemingly contradictory form factors.

  • g-man

    Sounds interesting. Changing camera focus based on eye tracking is a good idea but not overly complex to implement and certainly can’t turn a 32ppd display into 51. Curious to see what’s actually happening there.

    • Ben Lang

      They claim the display is 51 PPD and the focal edition gets 51 PPD passthrough resolution (which would indicate the display is the limiting factor).

      The standard edition also has a 51 PPD display but 33 PPD passthrough.

      Specs say they both use 20MP cameras…

      Ostensibly that leaves only the focusing system as the reason that one can achieve 51 PPD passthrough but not the other. But you’re right it’s curious how that could make such a difference. We’re in touch with Varjo to understand more.

      • XRC

        Please ask more about the controllers

      • g-man

        But 38040px over 120 degrees is 32 not 51, it doesn’t seem to be a 51ppd display

        • Christian Schildwaechter

          All Varjo lenses are aspheric glass lenses, meaning they have a custom profile to reduce lens distortions. The lenses on the XR-4 have a specific profile that widens magnification at the center, while reducing it towards the edges. So when looking straight forward, the same eye FoV actually covers more pixels, leading to the increased PPD there.

          Consequently the PPD changes depending on where you look, leading to a similar effect as on the XR-3, where the 70PPD was only achieved for a rather small FoV at the center with a dedicated micro-display, and much lower towards the edge. Which is fine, as our eyes also achieve their highest resolution when looking straight forward, causing light to fall onto the very small fovea. Beyond about 18° FoV, our color perception resolution drops by 75%. The XR-4 “sliding PPD” of course requires fast eye tracking and rendering the image accordingly, but matches our vision pretty well, without the sudden change in resolution as on the XR-3.

    • Christian Schildwaechter

      The auto-focusing cameras based on eye tracking may seem simple to implement, but were what actually surprised me, as I assumed that we have nothing that is fast enough to be usable. The issue here is the very fast eye movement and on-the-fly depth adjustment based on cues like blurriness that allow us to change gaze direction instantly, while a HMD would have to first wait for the eye movement to stop, then analyze the focus based on the eyes vergence, calculate the required focus and adjust the lens focus accordingly.

      This will inevitably come with some latency, causing the image to be blurry for a few milliseconds. I am very impressed that they seemingly managed to keep this low enough to be usable, though I don’t even know about research regarding people’s sensitivity to this type of image distortion. So similar to ETFR, technically implementing it is trivial, because tracking the pupils with lores cameras is trivial. But implementing it fast and precise enough to not lead to mayor distortions and visible artifacts is far from trivial, and necessary to make it actually usable. Which may be why the XR-4 Focal Edition costs USD 10K compared to 4K to the regular version.

      • g-man

        Autofocus systems are very fast now; I wouldn’t expect it to take long to change focus the depth determined by your gaze and the existing depth map. Small cameras also have small apertures and so fairly wide depth of field, leaving some room for error/delay.

        Anyway definitely interested to try it out someday, it is a cool idea.

        • Christian Schildwaechter

          I really hope so, as the fixed focus on passthrough cameras causes very serious issues for our depth perception, making wearing an HMD with passthrough pretty much a no-go for activities that require using the hands or quickly determining depth like driving a car.

          SadlyItsBradley’s video from yesterday is about his experience with the XR-4 Focus Edition. Varjo had invited him to the presentation, and he had an XR-3 for review for a few months before, so he could compare them. The XR-4 got a pretty raving review, esp. for the auto focus ability, which improves passthrough usability significantly. Unfortunately the speed was indeed a problem:

          Now my biggest complain about it was, the speed of the auto-focus was … very delayed, I would say.

          He asked the Varjo engineers about it, and they responded that they could have reduced that delay significantly, but for that day filled with demos chose to go for worse, but more stable parameters, to ensure that the focus worked reliably throughout the day. So the judgement is still out how fast/noticeable it can actually be when tuned for the best performance.

  • Naruto Uzumaki

    Quest 5 or 6 will have same specs for around 800$ and with more games1

    • Christian Schildwaechter

      Pixel count per eye increased by 53% from Quest 1 to 2, 30% from Quest 2 to 3, or almost doubled from Q1 to 3, for an average of 41% per generation. Varjo XR4 is currently at 315% of Quest 3, and with the same average increase, would be at 160% of Quest 5 and 113% of Quest 6, finally beaten by Quest 7.

      In reality this is more complex, because the increase in pixels mostly depends on GPU performance, which increases about 30% per year. We already had 2.5K and even 4K per eye HMDs based on Qualcomm’s XR2, they are just not capable to render anything moderately complex at such high resolutions. We already have ~4K micro-OLED displays like those used in the Apple Vision, but it takes desktop grade GPU performance plus ETFR to utilize these.

      So technically a Quest 5 or 6 may feature similar specs as the non-focus version of the Varjo XR-4, but with high end mobile GPUs lagging about a decade in performance compared to desktop, you probably still won’t get the same performance from a Quest 6 as you could get today from an XR-4 plus high end PC. There are of course no games that could utilize this today, but gamers really aren’t the target group here. And for many industrial users being given almost a decade of advance for their custom application will make this a no-brainer purchase. The “almost a decade” is mostly due to improvements in upscaling and image-reconstructions that will significantly reduce the GPU performance required to drive hires VR displays.

  • ViRGiN

    Mrtv already shoveling affiliate links

  • knuckles625

    Eh, I’m going to go against the grain here and say that there really is a valid use-case for this. I’ve seen what companies pay for prototype and visual models on a regular basis, very often with the sole purpose of gaining approval from a higher-up (President, CEO).

    If a company can pipeline a believably high fidelity render model from their design program to the headset and put it on CEO’s head & virtual hands (for approval/design changes), even the most expensive model could pay for itself in a month

    • Christian Schildwaechter

      The main value in hires is the display of real world user interfaces. The Varjo XR-3 with even higher 70PPD at the center was used for example to test car dashboards for readability while actually driving, where the driver has to be able to recognize small symbols at a glance. They sell a lot of headsets for flight simulators, with real cockpits being a Christmas tree like collection of small glowing buttons, and the advanced passthrough of the XR-3/XR-4 allowing to seamlessly blend in physical cockpit hardware in simulators. There are indeed a large number of use-cases for a HMD like the XR-4, limited mostly by the price. And not by the price of the HMD, but the much higher development costs of the usually required custom software needed to even take advantage of the HMDs features.

      • silvaring

        You’ve highlighted something that id never thought of, the glancing to check symbols on a dashboard… which is very different to moving your head to look at the symbols. For example in PSVR2 the dashboard looks photo realistic, even looking at symbols (Gran Turismo), but when quickly glancing does it have the same clarity?

  • timelessicons

    $10,000.00 msrp. lol why not.

  • Dave

    Varjo are basically milking the cow while it’s sunny – I don’t see these prices for 4K VR being that relevent in the next 12 months, maybe for passthrough, but we’ll see. So I think Varjo are taking a big risk here selling for such a high price. I wonder for how long it will be seen as just an overpriced peice of metal… I suppose that’s why they target Enterprise who don’t care as long as it fills there usecase.

    • Christian Schildwaechter

      The price is high for a consumer headset, but for their product design/flight simulator customers, USD 4000 or even USD 10000 may be the best offer they ever got. For a long time you could chose between a professional VR HMD for pilot training, or buying a house for about the same money.

      Just do a quick calculation how much more it cost to pay a software engineer USD 100/h to create or integrate a custom application for use with the headset. With tailor made projects, the level of support you can buy is often way more important than the actual hardware or price. If delaying your project by a day costs USD 1000 due to several people having to be payed while being stuck, you absolutely want that USD 1500/year support contract that gives you a direct phone line to Varjo’s engineers, so they can sort out your problems.

      HTC has been selling VR HMDs with less bang for the buck than Meta’s for years, yet nobody in their right mind would pick a Meta HMD for a mission critical project, where you need full access to the device, a stable operating system and someone you can call. Meta killed all their business programs, doesn’t support the Quest Pro, offers no version control or release management tools that would prevent your system from updating right before an important demo, crashing everything. Price is the least of the problems, and businesses will continue to buy from Varjo and HTC, long after Meta releases a 4K standalone for less than USD 1000, or Apple’s AVP turns out to be the best thing since sliced bread.