Hacker News new | past | comments | ask | show | jobs | submit login
Relativty – An open-source VR headset (relativty.com)
819 points by karanganesan on Sept 10, 2020 | hide | past | favorite | 222 comments



> We started Relativty because after watching Sword Art Online

This little fact alone makes the background story way funnier than most tech origin stories I've heard. And being able to do PCB layout and system design at 15-16? Kudos to Max and Gabriel.


Hopefully the design won't kill users on launch day.


For those downvoting, this is the plot of Sword Art Online.


It's fine - people will be still perfectly willing to continue using it after it does that. ;-)

(Plot of later part of Sword Art Online - really.)


The second generation of gear claims to have limiters to prevent nerve damage. Kinda like cars installing seat belts and air-bags.


These guys were 15 years old when they started, and are now 20. That's amazing, great work!


It's interesting how a lot of commenters are writing "Relativity" instead of "Relativty". I also fell for this at first.

Makes it easy to mistype the url, if you were to visit it later on.


Google will probably auto correct any searches for it as well


Great brand name. Awesome write up.

"We made Relativty in my bedroom with a soldering iron and a 3D printer and we expect you to do the same: build it yourself." Soldering SMD components sounds daunting, so kudos to you for this.

Not sure why the pedantics about 2K and the price in these comments...


> Soldering SMD components sounds daunting,

It really isn't, watching some videos learning some techniques. Practice. The same statement could be made about most things that people aren't familiar with or haven't spent any time working to learn.


Just because you can get good at something, doesn't mean it's not hard :)


I know what you're getting at, but anything you're passionate about doesn't feel as hard as it does to people who are not.

There are so many things in life that I think are nonsensically complicated and I have a hard time grasping why anyone would ever want to waste time on them, while I know how to do a lot of things that "normal people" consider outlandish - just because I spent tons of time researching and learning about a topic because I was interested in it.

So yeah, if you're meh about VR headsets and "it would be kind of cool to have", just buy a Daydream - don't try to learn how to solder SMD's. :)

If you want it enough you'll figure it out.


As someone who has worked in an electronics lab and done lots of soldering, I can say your setup makes a huge difference. A great setup will make soldering SMD components a breeze, but a bad setup will make it very frustrating. Practice obviously helps but a nice soldering station with some kind of magnification will close the gap a whole lot faster.


Using a proper microscope for soldering was definitely one of the most useful things I learned as a high schooler interning in an electronics lab.

I am still surprised that this is not more common. Even a biology-grade binocular microscope for 100$-ish makes a huge difference.


It's also worth note that manufacturers like JLCPCB have SMT Assembly Service which might be able to deliver fully assembled boards. I haven't looked at the gerber+BOM to verify.


Also, it can be quite pleasant. After a while it almost feels like drawing pixel art.


Considering how Palmer Luckey was just tinkering when he made the Oculus Rift, this can produce some useful innovation outside of the corporate environment.


To the creators: this is a really incredible achievement, especially to have done at age 15, and really well communicated.

Please don't get disheartened by the mass of comments around this not actually being a $200 headset[1] or your unusual[2] spelling.

This is freaking awesome, please keep it up!

[1] It definitely would be cool to see some sort of parts list, but that's not really the point here.

[2] Clever domain hack, IMO


I think a lot of people are missing what I think are 3 main points before moaning about everything else: 1. It's freaking open source 2. It's DIY fun 3. It's the bare VR without being an over-engineered device that reports your farts to the manufacturer.

I can only respect and envy the guys for pulling this off.


Doesn't the Steam dependency mean the reports are going to Valve?


Why would this be steam (or anything) dependent -- am I missing something?


Most of the VR graphics code & engine integration AFAIK is implemented within Steam, not on the headset. So beyond just getting some basic positional tracking you're not really going to be able to access any content without something else.


It looked to me like it was Steam compatible, not steam dependent.

However, I can see where the idea would come from, because they do talk a lot about Steam on the site.


What would you be able to do with the headset aside from connect it to Steam? They could add support for OpenXR but OpenXR is just starting to get traction and that's on the application developer side. I don't know how widespread the adoption of the HMD agnostic bits of OpenXR are.


On the developer side, both Godot and Unity have some support for VR.

On the game side, there's also Windows MR, which supports some games. I don't know if this headset supports it now, but it could presumably be made to.


Godot and Unity have support for specific VR API's.

This headset is designed for SteamVR.

You can build for this headset by targeting SteamVR in Godot and Unity, but you'd have to do a TON of work yourself if you wanted to build your own VR runtime to bypass SteamVR for use on this headset.


Thanks for the additional info, I'll definitely have to do some more research in this area.


But to use it without steam you would have to develop VR software for your desktop. I think there is some open source stuff but idk how well it works. The PC software end is a much bigger job than the hmd.


The parts are all listed right there on GitHub via the How to Build link, either with prices or direct links to AliExpress where you can see the price.

The build comes in under $200. I don't understand the whinging in the comments.


The lcd screens listed are 195 right now. With a lower res screen it might get close. Pricing might be variable.


For 1. Cheaper if you buy in bulk. You could have mentioned that.


The bulk discount is 3% off, and only if you buy 5 or more.


Then you skip AliExpress and go to AliBaba.


Is that actually better than AliExpress for low quantity orders?


In my experience with Alibaba, vendors would sell a single unit for cheaper but the shipping quotes are higher than AliExpress, but that could just be the vendors I was contacting.


One interesting thing I noticed is that this project ALSO doesn't seem to have a controller. You can't really play SteamVR games without at least one! (Issue #50)

You can grab a Samsung VR for very cheap, put a phone on it and do VirtualDesktop magic, but you are still gonna struggle without controllers. Same for Google Cardboard and other similar projects.

There's some very cheap controllers you can get, with base stations for tracking like the Nolo-Home, but they are still quite behind the tracking precision and latency of some "top-of-the-shelf" headsets.

It's an interesting idea and I hope the developments on it go forward, specially since the recent Facebook/Oculus announcement.


There are many great SteamVR games that support mouse and keyboard or an XBOX controller. Trover Saves the Universe is one of them.

I don't think the lack of controller tracking is a deal-breaker for those who want to get an open source VR experience.


This has gotten me wondering whether there are any DIY projects for Lighthouse-compatible controllers?


I seem to remember that someone had made a Lighthouse receiver for drones.



Unfortunately, these projects have been discontinued.


I never knew they made these boards that take two LCDs and have an adapter board with a DisplayPort (port) https://www.aliexpress.com/item/32975198897.html


This is really great. I wish people would stop comparing this effort to commercial products, though. Like most open-source hardware, the proprietary competitor will always be the better value proposition, if that's all you care about.


If it supports 5GHz Wifi, something like this could eventually work well with VirtualDesktop (An incredibly useful app which makes the Oculus Quest truly wireless and capable of playing all SteamVR games from your PC).

Keeping all the GPU overheads off the headset is the way to go. (And will also be seen eventually with future '5G' wireless systems (not to be confused with '5GHz' Wifi) which will put that 'grunt' in the cloud.)


OT but how is the lag with Oculus Quest + VirtualDesktop? Let's say I wanted to play Half-Life: Alyx.


On my Oculus Quest I have completed a playthrough of Alyx and a few other games of different genres, no issues and no delay whatsoever.

It feels magical, in a way that feels like it must be cheating and the spell is going to break anytime... yet it doesn't.

To pre-empt a few of the obvious asterisks:

- "No latency" is physically impossible. If you play twitchy music rythm games you will indeed manage to notice.

- You must have a good wifi ac router.

- Your gaming host must be wired to that router.

- To minimize network hoops, all of this should happen in the same room. Every milisecond counts.


> Every millisecond counts.

Not really sure that follows. For a 20ft room, it only takes a wifi signal 20 nanoseconds to cross it, and an ethernet signal around 30 nanoseconds. Your room would need to be over a hundred miles wide to add a millisecond of delay, disregarding stuff like repeaters.


I think the same room is to have good signal strength to minimize dropped/repeated packets. The transmission time won't be noticeable but the signal to noise ratio may decrease fairly quickly with distance and obstacles.


I think the lag is subjective. I find it noticeable compared to wired - however I'm fairly latency sensitive. I think if you've never played Alyx over wired, you may not notice the lag in VirtualDesktop. Playing something timing sensitive like Beat Saber it's very apparent and basically unplayable.

You also need ideal wifi conditions. I found the best option is to have a dedicated wifi card on your desktop and just make it into a dedicated AP for the headset. Without this I'd get occasional glitches/pauses which is motion sickness inducing.

Even with these drawbacks, I'd say VirtualDesktop is pretty amazing and shows the potential of what's possible. The image quality is far better than the wired Oculus link. The software is also far less buggy than Oculus link - it has nice, working integration with SteamVR that just works out of the box. I wonder how a single person could write a better implementation than Facebook/Oculus engineers, but there you go. And when it works, under ideal wifi conditions, it's really nice not to have a wire tether.

I imagine the future of VR will have dedicated wireless base stations. Nintendo was doing this almost 10 years ago with the WiiU - much lower bandwidth, but virtually lagless and resilient even in noisy RF environments.


It's great. But make sure you have a decent 5GHz router.

I use a TP-Link AC1900 and Half-Life Alyx works great.

VirtualDesktop completely revolutionises the Quest. It's an absolute must-have.

I wouldn't be all surprised if it got bought by Facebook at some stage, or they build something that does exactly the same thing.


It's a great start at a young age, congratulations and all the best!

I'm not into VR, but can this setup be used to create a 'Wearable low latency display for computers'[1] i.e a wired video headset? None of the current video headsets seem to work well according to the reviews be it $800 Ziess Cinemizer or $200 Avegant Glyph.

[1]https://needgap.com/problems/16-wearable-low-latency-display...


A typical Apple "Retina" display has 57 pixels/degree[0], held at ~10 inches. VR screens are roughly similar resolution but held far closer to the eye, so pixels/degree is proportionately less. All this mean that we'll need dramatic increases in display resolution before it's pleasant to read text in VR for long periods of time.

[0] https://en.wikipedia.org/wiki/Retina_display


I think a 'video headset' doesn't try to such a wide angle as a 'VR headset.' I would really like one for programming on, as it's more portable than a big screen.


There's a good one coming out soon, the Cinera Edge Elite.


https://www.cinera.net

It's a KickStarter, so I guess "coming out soon" is pretty speculative, but that's the link for it.

Interestingly, I googled this and the permalink to this comment was actually the top result, 1 hour after the comment was made.


It's aimed at watching movies, running android likely for the support of streaming apps; What I want is a computer monitor on my eyes and so this either underperforms for that need or is an overkill.

OP project may in fact solve this, I need to look deeper into the tech.


This is awesome. I'm glad we have have insurance in case no one fills the void of HTC leaving. You can argue that Valve is, but given the massive back orders of the Index and their discontinuing of their controller and streaming box; I'm not totally confident.

Now we just need controllers and maybe even base stations.


Valve discontinued their controller and streaming box because they did not sell - to get rid of their stock, they had to put them on fire sale. They're still supporting the controller and afaik the streaming box as well. The controller was innovative but often regarded as awkward, and the streaming box wasn't particularly special.

The Valve Index on the other hand has redefined the high-end VR market and is selling out accordingly. I've tried most of the consumer VR solutions on the market and the Index "knuckles" controllers blow everything else out of the water, it's not even close. Oculus has dropped out of that market, opting to produce more affordable but somewhat less capable headsets and controllers. I don't see HTC turning it around as they don't have the name recognition or the capital and their tech is already weaker.

Point being is that Valve owns the top-end of the VR market already and this project, while impressive and definitely promising, is not competing with Valve. It might in the future, with a lot of work, but it's not insurance against a Valve dominion. There might not even be room in this niche market for more than one competitor, which is awful but that's what we're looking at.


One reason the Steam controller didn't sell is that users had to figure out configuration schemes on their own instead of Valve being proactive with developers and publishing houses.

I agree that the Index, especially the knuckles blows everything away, but I'm already seeing the same issue that the Steam controller suffered from pop up with the knuckles.

You're right about Oculus too, and I hope that you'll be right about Valve staying in the game. I definitely want to keep using base stations instead of unreliable and heavy headset cameras. I want full body tracking to gain traction and that won't happen if external base stations die off.


> fills the void of HTC leaving

Huh? I'm out of the loop here. HTC is still selling the Vive Cosmos. What are you referring to?


HTC is going commercial only. From what I’ve seen with other Chinese companies going that route, it makes it either harder or near impossible to buy a single headset when a company takes this route


Bravo les gars :)

Sincerly nice project, I would do one myself if I had the time or knowledge. What could be really nice is that make a DIY video to demonstrate it's easy and link to where we can buy different stuff (and put a affiliate link, please). Also suggest controllers that are compatible with steam if that exists.


Ever since Facebook bought Oculus I've really hoped something like this would spring up.

I really really want it to succeed and would be happy to back it on Kickstarter or similar.

How do we ensure that these projects don't end up being bought out by Facebook or Google?


I have mixed feelings about the Oculus and FB’s ownership. I would like an open platform but I appreciate both the more than adequate Oculus hardware and some excellent VR experiences created by both Oculus and many 3rd parties.

I might still buy a manufactured open platform system, depending on available experiences. I would also definitely buy an “Oculus 2” that had a higher resolution screen and slightly lighter weight.

I really like the recent Oculus software update that allows some games and VR experiences to be used without controllers by recognizing hand gestures. I hope there is a lot of exploration of new ideas based around no controller use.

Hopefully open platform VR headsets will be adopted by university research teams as well as commercial content creators. Hobbyists creators are to be encouraged but they are not enough to get widespread adoption.


You mean 3? We're already on Oculus' second generation.


Thanks, I meant "Quest 2" - I only kept my Oculus Go for 1 week, then gave it to my grandson. I get a lot of value from my Oculus Quest which I bought last year.


The Quest 2 is rumored to be releasing this month, but you'll be forced to log in with Facebook to use it.


I will probably get it. I just spent 20 minutes playing the Quest ping pong game - amazing game physics and general gameplay.


If we make sure it's licensed properly, that's not possible, right? They could hire certain people but if it's all open source, we can at worst just fork it.

According to the GitHub it appears to be under GPL v3. I'm not qualified to state whether this applies to the hardware design or just the software.

I certainly wouldn't want to put any effort in to a project where the IP can be bought.


Being under an open source (the term you’re really looking for is free software) license isn’t enough to protect a project. You have to maintain forward momentum. A company can buy the trademark, and the company or group of people providing that forward momentum, like Oracle did with Sun, and then, like China assimilating Hong Kong, it’s just a matter of time. Solaris died, Java got wholly corporatized and free software mostly moved on.

Stallman, of course, realized this a long time ago and started a group of people, in the legal form of a charitable foundation, to keep that forward momentum going. To his telling, his Hail Mary was phenomenally successful.

If what you want is to work on a platform that will stay free no matter what, your best bet is to align with the FSF. They have the most momentum, the fewest corporate threats, and a long enough memory to keep it that way.


> Java got wholly corporatized and free software mostly moved on

?... The Java community was completely unaffected by the Oracle transition and is as strong as ever. Of course, Java wasn't open source (or "free") to begin with.


Sun released the JVM under the GPL in 2006. Oracle bought them in 2009. Sun always intended to release Java as free software, they just needed to make their money first, and were a big user of dual-licensing, which is endorsed by the FSF. It's a pattern they replicated across their entire product stack.

Of all the big names in the technology space that made their buck from tagging along the free software train, Sun was far far far from the worst offender. When they got bought, I have friends that were personally extremely offended, they had built their professional lives around expertise in Sun tech.

And, as to "stronger than ever," if you're a new coder, are you going to learn and use and love Java, or Golang? Java is what you do for your job. The only reason it's still a thing is because of the massive existing install base. The company I work with, is already starting to treat their Java projects as legacy and most teams, including mine, are starting up, are moving, or have already moved to Golang.


Can anyone recommend an elegant method to 3d map a face? If I were to print one of these housings, I would at a minimum adjust it to my PD, but it would be great to also contour the fit to my face.


I did a similar thing to design a 3d printed mask.

I used an iOS app called ScandyPro which gave me a mesh which I then imported into Fusion360 and modelled some lines/layers of the curves around my face. Then I performed a 'loft' between them.

The result was a perfect fit.


I've been hoping someone would create an open source version of the lighthouse technology. I think that technology could be huge outside of VR.

Inside VR, I just want more custom hardware... I want VR feet.


> I want VR feet.

That exists, and has existed for a couple years now through the Vive Tracker [0]. It's essentially a puck that acts as another receiver for Lighthouse tracking. You typically strap them onto an object or body part and it gets tracked.

Search YouTube for "full body tracking" and you'll see it in action, though it's mostly just a bunch of silliness in VRChat since very few games actually take advantage of it. With each tracker being $99, adding both feet and your hips will run you $300.

[0] https://www.vive.com/us/accessory/vive-tracker/


Isn't the lighthouse system relatively open? I recall the analog receiver circuitry is relatively simple, and Valve cooperated with an IC manufacturer to get rid of the analog stuff (or, more precisely: move it into an IC). The IC was intended to be not be exclusively for Valve alone.

Of course that doesn't help with the base stations, and might still violate [Valves' or other's] patents if used in an actual product.


> Valve is now making SteamVR Tracking fully available to other companies, without licensing fees.

> There is no certification from Valve necessary. Ship the product you want to build, whenever you think it's ready.

I don't believe it's strictly open source, but it's certainly far more open than Facebook's ecosystem.

Source: https://partner.steamgames.com/vrlicensing


on a side note, I wish if there was a company like LEGO but for providing components for building DIY electronic gadgets especially small size home robots which gets some particular work done. for instance DIY wifi locks for doors, automated food dispenser for pets etc.


Isn't that AdaFruit and Sparkfun?


People have tried to make kits like this but they never work out because a LEGO-like form factor is not flexible enough for 99% of projects.

Electronics are already pretty LEGO-like conceptually; you just have to solder instead of snapping things together.


Radio Shack could have done this.


Amazing work, congrats!

I think it is smart to build on top of SteamVR and PCVR, but...

I would love to drop my Oculus Quest for a DIY open-source headset. My #1 requirement is stand-alone. No cable, no wires. I don't own a gaming PC, so the Quest has been a great way for me to experience 6DOF VR.


I'll repeat a comment I made on Reddit:

I was scanning the page trying to understand how you were doing positional tracking. Took me a while to realise that's what you mean by "Experimental room-scaling"

I'd suggest that this isn't the best wording. Call it "positional tracking" or "6DOF" or something.

The whole "room scale" thing was confusing at the start and isn't used much nowadays (Is seated 6DOF VR "room scale"? It's ambiguous and irrelevent for most purposes)

I'd also say that I think most people assume that being SteamVR compatibile implies having 6DOF tracking of head and hands. Most software simply won't work unless you have this. How experimental is experimental?


I understood what they meant by "room-scaling" immediately and I've only ever used a VR headset one time about 4 years ago.

On the other hand, I can surmise from my experience with robot arms that "DOF" likely stands for degrees of freedom, but I would have had no clue what "6DOF" meant in terms of VR.


3dof. 3 degrees of freedom. These are typically rotation - yaw/pitch/roll. You can look around. 6dof is 6 degrees. In addition to rotation you add translation (x/y/z) so that you can walk around the environment. Generally you can't get translation without at least first having rotation because humans are more sensitive to rotation.


OK. Fair point about "DOF".

But what do you understand "room scale" to mean"? Genuine question.


Not OP, but in SteamVR room-scale is typically used to indicate you can move around within a play area, and your avatar/camera moves as you do.

Its the way the setting is named in Steam, and in SteamVR games.

Example:

https://www.vive.com/us/support/vive-pro-hmd/category_howto/...


I imagined it to mean I could move around a constrained physical space while seeming to explore a larger virtual environment (thus "scaling" the dimensions of the room I was in). Admittedly the reference to a webcam that tracks your movements was an additional helpful context.


That is also roughly my interpretation. But that is different to the meaning used in the post itself. See my other comment below.


Oh, I thought it was a cool mode that made your current room look huge or tiny.

A sort of "Alice in Wonderland" mode.


Yeah. The term arose around the time the Vive was released to distinguish against the Oculus Rift. It was a fairly subtle distinction - the Rift at that time was optimised for front-facing experiences because it came with two cameras you usually mounted on each side of your monitor. If you moved out of their cone or occluded them by turning around too much then it would lose tracking so most games assumed you'd mostly be facing forward.

The Vive on the other hand had tracking that was usually mounted in opposite corners of the room. Two trackers were more than capable of covering an average sized room and the only limitation on your movement was cable tangle.

The Rift tracking got a lot better eventually - even with two cameras.

The other meaning of the term is sometimes to distinguish between VR that can track the headset position (6DOF) as well as it's rotation (3DOF). But all serious VR is 6DOF nowadays. 3DOF is pretty much the domain of "slide your phone in this thing" VR or headsets that are intended mainly for viewing 360 video.

So - two different meanings of "room scale" that mean quite different things.


This might be a fun weekend project to work on, but cannot be built for $200. The displays alone (and their driver PCB that accepts DisplayPort) cost $195. They are also not 2K but rather 1440x1440 (or 2.0736M pixels). Maybe the displays were cheaper when they built this. They also hookup a 6-axis tracking chip (MPU-6050) to an Arduino board and feed it to a PC via usb. Even if tracking was working perfectly, this would be a cut down version of an Oculus Go, which additionally has smartphone electronics integrated, can run VirtualDesktop, and can be had for $150 as of this moment.


> Even if tracking was working perfectly, this would be a cut down version of an Oculus Go

Yes, I agree, this is a tremendous achievement and a great project.


Nobody disagrees with that, I greatly respect the project's authors, and even just successfully replicating it would be undoubtedly a great learning experience!

I think the comment is just trying to align the expectations better, given the context in which it appears and is communicated.

The $200 claim is plainly inaccurate, and unlike many open-source projects whose stated and successfully executed aims are to provide a hacker-friendly, cheaper alternative, this project does is not one of those; and its technology is in fact inferior. If one aims to experiment with VR, there's simply better options for an individual.

This is separate from the project itself being a great achievement and/or a potential base for further community improvements - and it would be appropriate to celebrate those aspects, which the website, and their ProductHunt entry do not communicate too well - as is evident from even this comment section.


They mention in the build guide that the display is sometimes available for $150, that's probably how it gets to around $200. MPU-6050 appears to be about $30, main PCB is claimed to be $5 or so, another $5 for the lens and foam respectively, and $10 for the strap. That'd bring it in at $205 provided you already had the capability to 3D print and solder and didn't factor in the material cost for those. It's unlikely, but you could probably make it happen.

I assume you could downscale the display to like a 1080p version to save costs as well.


On the plus side - you don't have to have a Facebook account.


Probably Facebook can sell it cheaper that the hardware costs, because they (hope to) make money by tracking its users and showing them more ads.


Also, when you are homebrewing a device, vs say, shipping 400,000 devices to users, you just don't get the same economies of scale.


1440x1440 is 2k, as is 2048×1080. It refers to the total number of pixels.

4K is 3840X2160, but you wouldn’t be wrong to call 2048x2048 4K, especially in a VR display where you have two displays, doubling width, and in the end what matters is pixel density.


I think the (repeated) note that you can downscale to 1080p displays is how you are supposed to do it under $200. They note themselves these displays are $150-$190 so it's not like they are hiding something.


High resolution(> QVGA) display panels is difficult to come by, especially at small quantity-per-months.

Those in the form of Raspberry Pi or Arduino parts are more accessible but still not cheap and not so high resolution.


Just buy a used vive for that.


No one is claiming that you can't get something cheaper, either.


Great, is the oculus go open source?


Not a valid retort, IMO, to a comment that was only making an observation on the prices of the parts.


Counterpoint: the comment was not only making an observation on the prices of the parts.


If you only use one variable for comparison then the comparison is hardly valid.


It is a valid retort and trying to hide that fact when comparing doesn't do it justice.


The fact is well known and obvious to anyone so I don't see how anything is being hidden by anyone.


I'm not sure where the confusion is coming from. The fact still stands that it's an achievement for an open source DIY project to come close to Oculus Go specs. The original comment was asking a rhetorical question to make that point.

If Oculus Go was open source, it would cost you much more than the retail price to manufacture it.


The people in the know who believe it's obvious will be the same people who know the price of an oculus.

As for the rest of us. We neither know the price of an oculus $150? or that an oculus isn't as open source as other products.

Valid point for the rest of us.


> Even if tracking was working perfectly, this would be a cut down version of an Oculus Go

Oculus Go doesn't have position tracking. I don't know what you're talking about.


Neither does this headset. Oculus Go has at least a 6D IMU, that gives it 3DOF (no position tracking), while this headset has a 6D IMU. The Quest and Rift have cameras on them for positional tracking. If it was possible to have reliable positional tracking just from an IMU, John Carmack would have delivered it on the Go.

This OSS headset has "experimental positional-scaling" using "any camera". On github, they say: "The tracking is still very experimental and can only be run on NVIDIA GPU due to the usage of CUDA. The tracking makes uses of only a video input and an Artificial Neural Network AI trained to estimate a 3D body position. This method is nothing close to the precision or freedom of movements of a dedicated sensor however we believe that the model can be trained and improved by orders of magnitude."

Again, this is a fun project. Probably as fun as the cardboard VRs Google used to hand out for free.


So it does have position tracking. I guess it uses an external camera directed at you - of course that will be far from perfect, especially if you only have one. But it has potential. With more cameras and an improved model, it could possibly even do FBT in the future. At no or marginal upgrade cost.

I see more in this than just a weekend project. Even the fact that you can just buy a new lens when one gets damaged means this might actually save you money. (This actually happened to my Quest; lucky I could save it with some polish.)


Saying it "has tracking" is a bit of a stretch, given that the tracking isn't a feature of the headset, but rather a separate system that can theoretically work with any other non-tracking headsets.


i'm pretty sure the $200 quote is an "it's possible" type thing, like where car companies quote prices "from" the lowest-spec version.


Can you ever buy a car at advertised price? Excluding taxes... Im trying to buy a motorcycle and the dealers tag on 20% in fees


In the US, if you buy a car at the advertised price you're probably getting massively ripped off!


I figured THE base configuration are for corporate uses. Plain iron exhaust, no hubcaps, flat paint, manual air conditioning, multi band AM radio and the likes. Sales reps worldwide are driving in that configuration but dealerships that don’t expect business customers probably don’t stock them.


You can in most countries that have proper consumer protections. Cars in Australia are often advertised with their "drive away" cost which includes all state and federal taxes etc.


It is not possible. Not now at least.

Even if they manage to get the display+motherboard under $200, there is a number of small parts that will end up being quite expensive if you can't do the economies of scale big manufacturers can do.

There is the 3D printed enclosure, lens, cables, straps, padding...

Also it is just a headset, you probably want some sort of controller too. Also, even if though they mention experimental "roomscale", it involves a camera that is not included. And as with any DIY project, you a probability missing a few tools you'll have to buy...

I'd say it is affordable as a hobby project, which, in itself, is really nice. It is the kind of thing you enjoy making even more than you enjoy using.

For cheap, you can't beat cardboard. And if you really want VR and not a new hobby, commercial offerings are hard to beat.

EDIT: one way to reach the target price could be to start with one of these cheap "cardboard+" headsets you can find for about $20 on Aliexpress. They already have most of the parts you need except for the display and motherboard. The optics may be tricky though as they are optimized for smartphone screens (if they are actually optimized for something...).


I am sure you can build a version for $200, but if you are going to says its $200 you should provide an exact list of parts you used to make it for $200.

I did not take it as each eye was getting 2k when I read it. Your total screen space is 2880x1440 with those displays. Even with two 1080s you have 2160 total. But then again the screens are advertised as 2k on aliexpress. So it is all a bit weird.


You forgot the cost of Mark Zuckerberg‘S eyes in your bedroom.


Sounds like you could sell this pre-assembled for $300 and it'd still blow all the other headsets out of the water. Taken at face value this sounds pretty great.


We know nothing of how performant it is. The reason ALL VR isn't at 2K 120fps isn't because other companies don't know how to do that -- most of the technological hurdle comes from the PC producing those outputs -- it's just that there are clearly performance standards companies hold themselves to in order to consider the product fit for a general audience.

In regards to the room scale tracking alone:

> Precision and freedom of movement are still very far from dedicated sensors, however, we believe that the model can be trained and improved by orders of magnitude.

Whereas, say, Oculus, already has room scale tracking down-pat.

The only way I see this as "blowing all other headsets out of the water" is if the way they're able to achieve 2k 120fps where others could not is Relativty simply being willing to cater to a much smaller audience of high-end PC owners that a mass-reaching product like Oculus wouldn't limit themselves to.

Anyway, this whole thing is really cool, and I want to see more of it, I just doubt that the correct way to think of it is as something already capable of blowing competition out of the water.


The problem is roomscale, although if you bumped up the budget by $30 dollars I bet you could put the sensors needed for 6DOF in there. Then it’d be great! (Without the sensors, most SteamVR games would be janky until their model is improved)


As others have pointed out the displays aren't 2k and they aren't 120hz and they cost almost $200 alone. So none of the marketing is true.


The developers could add 6dof roomscale tracking relatively easily by adding an intel 265 (https://www.intelrealsense.com/tracking-camera-t265/). The unity integration is very plug and play. The 261 can also be used but must be bought in trays of 8.


Just their suggested displays cost $195 (2 displays with driver pcb). Otherwise, very nice.


That's for 2k @ 120fps. You're free to use any display pair.


This is really awesome. People doing things like this just makes me happy.


From a financial aspect, I think the soldering ion and a 3d printer would cost more than $200, plus $200 on parts. So with $400+, i would consider buying an Oculus?

From bottom of my heart, hats off to those kids.


> the soldering ion and a 3d printer would cost more than $200

That sounds odd, a bit like saying a car and a smartphone holder together would cost more than X. A soldering iron is, what, 15 euros? And a cheap 3d printer like twenty times that?


I actually love this but I wonder if inside-out tracking using webcams might be better (although you may need more oomph on the processing side).


This looks great. Is there anything on the horizon filling in the more inexpensive market that Windows Mixed Reality headsets were occupying?


Daily reminder: 2K doesn't really exist. Yes by definition anything with around 2000 pixels vertical would be 2K but that would mean FHD is 2K because it almost 2000 pixels wide. The display in question here is 1440x1440 That is not 2K. Sure there are two next to each other that would mean 2880 pixels horizontal but the image you see cant have this resolution. Also that would be 3K not 2K. So pls everyone stop the nonsense with the Ks unless used properly like if you actually mean DCI 4K then its appropriate to say so. Everything else is marketing nonsense. /endrant


2K was/is a real specification that defined and used in digital cinema well before consumer TV's got near to these resolutions.

DCI specifies a few 2k resolutions based on aspect ratio: 2048x1080/24 full Frame, 2048x858/24 for Scope, and 1998x1080 for Flat. Basically any crop of the full frame image with either the maximum width or height is still 2K.

There's probably not a lot of value in beating that dead horse anyway; I have a sneaking suspicion this is all just a typo. Most devices containing a display like this are usually digital cameras or imaging instruments which almost always market the raw pixel count ahead of screen resolution. This is probably generally advertised and specified as a 2MP display not a 2K display.


Can we just go back to calling resolutions by their height?


There was a brief period of sanity in the late 90s early 2000s where resolution used actual numbers (ex: 1024x768)

Otherwise it is a mess. There is the letter soup (WSXGA+, WQUXGA,... ), the lines (1080p, 720i,... a throwback to analog CRTs), the HD (Full HD, HD ready, ultra HD,...), the K (4K, 8K,...)...

Is it that hard to use 2 numbers?


Both existed at the same time.

You had VGA (640x480) for example.

But I do agree we should have called it 2160p not 4k, but marketing guys are like that.



I remember when makers were advertising "8514/A!" You know, 1024x768 interlaced 43Hz.


You missed Megapixels too, which is usually used for cameras, but I have seen used for screens. 1080p is around 2 megapixels, 4k is around 8 megapixels.


That doesn’t really help, as while typically it’s assumed the aspect ratio is 16:9, that’s not always true, making such a scheme ambiguous. Listing vertical resolution with aspect ratio is fairly sane, though. Or use standardized specific resolutions (e.g. FHD, QHD, 4K UHD).


These specifications don't mean anything for a lot of people. To me they sound like marketing buzzwords that companies use to try to sell me inferior products. I'd much rather have them say 2k than any of the "FHD" abbreviations.


I actually like the abbreviations because, while exceedingly optimistic if you sound out what each letter means, every vendor maps them to the same number of pixels. For example "QHD" or "Quad HD" means 2560x1440. "That's not 4 times HD, what a dumb name," I hear you scream. But in this system, HD is 1280x720, not 1920x1080 as everyone in the real world understands it. But it's fine, because anyone selling their monitor as QHD will be giving you a 1440p display. They get to mislead you as to the specifications... in a standard way that everyone agrees is the maximum amount of misleading they can get away with, so you actually know what you're getting! (If they did 4x "real" HD, then they'd call it "UHD" or "Ultra High Definition".) It's all genius and I love it.


Perhaps they don’t mean anything for a lot of people, but ambiguous terms mean nothing precise to anyone at all, so how useful are they really? At least a term one doesn’t know they can look up and learn, rather than be perpetually misled by ambiguous terms.

If anything 2K and 4K and 1440p are the more marketing terms because their ambiguity allows companies to apply them to products in a misleading way to market their product as something better than it is.


At least FHD almost universally means 1920×1080 and is googlable. This is less true for QHD / QHD+, unfortunately.


I thought QHD was almost unambiguously 1440x2560 (4xHD), and QHD+ was wider than that.


I think QHD+ is a little more ambiguous, as it could also reasonably be used to refer to taller displays (16:10 display at 2560x1600, for instance).

Also, annoyingly enough, QHD itself can stand for "Quad HD" or "Quarter HD", the old smartphone standard of 960x540, which admittedly has pretty much entirely fallen out of use


Isn't that qHD?


Today, it is. But humans are usually not very good at paying attention to the case and machines are too often very bad at preserving it (the best — or should I say worst — example being search engines). And that's without even taking into account the fact that the need to distinguish both didn't exist before QHD was a thing and that most people have no idea there's ambiguity at all.


It's exactly the opposite. Whenever you see 2K you have to check what they actually mean with that. I can be ~2000xAnything or they can blatantly lie and name 1440x1440 2K. FHD however is defined its exactly 1920x1080 and nothing else.


I've always wondered why we can't just use the pixel count.


That's how you'd end up with a billion pixel monitor that's 100x100000


Camera manufacturers often advertise the number of "dots" in their displays/viewfinders, but they also tend to count each subpixel separately so it's 2-3x higher than the pixel count.


That was never a thing so we can't go back to it. A resolution is 2 dimensional height + aspect ratio works. But for all the popular resolutions we have a standardized short form like FHD/QHD/XGA/VGA but these obviously cant cover every existing resolution. Like for example the one from the article 1440p 1:1 there is no standardized name for it.


At least for me, the short forms are hilarious and obnoxious. I can get an understanding of what 3840x2400 could mean in terms of aspect ratio and pixel density. But WQUXGA? I have to look that up.


When widescreen LCDs were starting to replace 4:3 LCDs and CRTs, the screens were described as 720p, 1080i, etc. It wasn't until 4k was a thing that screens were described by their width instead of height.


The "p" is kinda interesting because it used to convey useful extra information (presence of Progressive Scan), but now that's simply a given in almost every context. So the "p" has sort of degraded into nothing more than a signifier that someone is writing about resolution.

"c" could mean Colour. Bob uses a c2160p monitor!


i stands for "interlaced" that has nothing to do with display resolution. p stands for "progressive scan" technically also not related to resolution but it makes clear that the number before describes the vertical resolution.


My main monitor is 5120x1440. Is it 1440p?


Yes, but without aspect ratio you don't give all information. It's also 5K but again missing aspect ratio so best is to name the pixels.


I think it can be difficult to compare VR Headsets by stats, since the fields of views are different and stereo overlap is different. I think the metrics people are looking for are something that proxies for space, clarity and responsiveness. Ideally we'd have figures for these that can be compared across different headsets.

Although lots of different things can affect these numbers (e.g. size, shape and technology of the pixels) probably the best proxies would be diagonal fov, pixel per degree and refresh rate.

It'd be great to see everyone just standardise on reporting something simple, meaningful and comparable.


You missed diagonal. I believe that is how VR displays are often measured.


Having 1440p screens would have been great two years ago. The baseline currently is Index with 1700ish vertical pixels, and competitors such as HP and Pimax have pushed past the "4K" (2160+) pixel mark.

Disclaimer: I'm a sim game enthusiast and don't play any titles where the game designer can just change the gameplay mechanics and UI to work on screens with a low PPD.


> Disclaimer: I'm a sim game enthusiast and don't play any titles where the game designer can just change the gameplay mechanics and UI to work on screens with a low PPD.

I'm sorry but that sounds like pure snobbery. With the same PC running the game, and only changing the display, any game can be adapted to anything by changing gameplay mechanics and UI.


Then it's no longer a systems-level simulation but an arcade game with a sim-like physics.

However, you are correct in the sense that most racing and flight sim folks tend to be entitled snobs who have spare money to throw around. Many have spent money on a RTX2080 Ti card when the non-titan edition could not run games in VR at high enough frame rates. Most are going to be upgrading to a 3080 as soon as it comes out this fall.


> Then it's no longer a systems-level simulation but an arcade game with a sim-like physics.

How are you defining a "systems-level simulation" in a way that it would no longer be one if run at a different PPD? I feel like there's something I'm missing here.


> Most are going to be upgrading to a 3080 as soon as it comes out this fall.

You mean in a week?

Looking forward to replacing my GTX 1070. Such an underpowered card to sit next to my i9-9900K and 32 GB of RAM.


The website makes it clear you can swap out for other panels.


Any sim game recommendations that fit your criteria?


Digital Combat Simulator (DCS World) has some really tiny text on the cockpit dials and screens, and distant targets are often only a few pixels in size. Those 2000+ vertical pixel HMDs I mentioned don't have any issues with cockpit details whereas my older Samsung Odyssey+ required constant use of the zoom button.

Furthermore, the performance hit when upgrading to higher resolutions is negligible. Smaller pixel densities require heavy supersampling to look good whereas higher density displays can be run at or near their native resolution.


DCS is so cool. I feel like I need to quit my job / be single again in order to learn it though.

I got the rift in order to play it only to realize that learning DCS from the ground up requires a lot of keyboard input / referencing cheat sheets so need to be pretty proficient with it before you can't see your hands and as you mentioned the resolution is not quite good enough in rift to really see things with much detail though it is still a fun experience to try out.

Looking forward to playing it on a future headset.


Exactly. The original Rift and Vive were 1440x1440 (iirc) and the Vive Pro, Valve Index and most WMR headsets are 1440x1600.


It's an open source project, you can modify it however you want


It's like Oculus redux.

Give it a few months and Google/Microsoft/Amazon will probably buy it up.


If the product turns out great they might buy it as an aqui-hire.

What has Facebook gained from their purchase of Oculus? At the time I thought it was a pretty shrewd move by Facebook but six years later I'm not sure they got much for the $2 billion they spent.


Judging by the demos[1] they've shown and changes they've made to the Oculus software like homes, it seems like their long term goal is for it to be the basis of a google glass like system that's more environmentally aware and collecting lots of environmental mapping data for FB, and a VR facetime/The matrix but with Xbox Live avatars-ey thing. I think adoption is probably a ways out, but if people wind up wanting more than Teams/Zoom/Discord/Facetime and so on it sets them up.

[1] https://www.youtube.com/watch?v=JTa8zn0RNVM

https://www.youtube.com/watch?v=Ijmnk5L767U

https://www.youtube.com/watch?v=0JD6F9XYrMQ


By the time any of that is ready, their purchase of Oculus will be more than a decade in the past. I think the acquisition was probably a mistake, but for Facebook blowing $2 billion isn't a very big mistake.

From my perspective as a consumer and somebody who was very interested in VR 6 years ago, I wish Oculus had either remained independent or had been purchased by somebody that was actually going to do something with the technology before a decade passes. I think Facebook's purchase of Oculus is one reason VR has lost so much energy and interest (Magic Leap's hilariously bad execution in AR hurts VR too).


If Relativity ends up with a competitive design, I agree that it's a risk. But I'd rather hear from Relativity's creators before assuming it could go the same way.


Where do you source the lenses?


There's an AliExpress link in the README of the Github repo.

Though, I'd imagine you could use ones from cheaper 'phone-to-face' Google Cardboard-style ones.


Lenses are actually not that irrelevant for this application. There is a huge difference between some cheap plastic and camera-grade glass. You don't want [excessive] chromatic aberrations, or [excessive] "God rays" (the Vive has that effect), or other optical artifacts.

Optics is similar to electronics: With high school physics it's all quite simple, but with high school physics a transistor can switch nearly infinitely fast ;-)


That's true. I actually have a few cheap card-like gadgets. And they're certainly not fresnel lenses. ;)


On a slightly related topic: are the open-source VR controllers?


This is extremely impressive and well-communicated!


Can I point out, that's a nice domain name.


Did you notice it's missing the final 'i'?


Ha, I did not.


The screens are 1440x1440 and square format makes a lot of sense for this application. They're best described as 2MP instead of 2K.


Jees, pls dont. MP should not be used for display resolutions. It's 1440x1440 or 1440p 1:1 But yes definitely not 2K.


For a screen that you're placing in front of each eye, giving the total resolution actually sounds pretty appropriate. Particularly since although this is technically 1440 lines it only has as many pixels as 1080p resolution, which won't be obvious to most. Given all the different aspect ratios if screens started being quoted as megapixels and a format ratio we'd actually be much better off than with the current mess.


Since head-mounted displays put the screen a mostly fixed distance from the eye, I wonder if resolution is best described in terms of solid angles[0].

[0] https://en.wikipedia.org/wiki/Solid_angle


You're confusing resolution and virtual image size.

Virtual image size of an immersive display system (video headsets) is the equivalent of the diagonal of a standard scren.

Virtual image size is typically quantified by the Field of View (FOV), and is typically expressed as an angle your eye closes with the diagonal edges of the virtual image. Pretty much what you describe.

Another way people express virtual image size is by saying e.g. "a screen with a diagonal of X cm (in) at a distance of Y m (ft)". This may seem more intuitive at first, but problem with that is you can use different combinations of X and Y to express the same virtual image size. This is why FOV is a far superior way to quantify virtual image size.


No, not at all. Solid angles could describe how big the display looks, but absolutely nothing about its resolution.


What I meant was to provide both of the following measurements:

(1) Solid angle of each individual pixel. (Assuming that typical screens and optics are used, I'd expect this value to vary depending on the location of a given pixel. So something like (average(solid angle of each individual pixel in the display)) could be advertised.

(2) Solid angle of the overall virtual image.


In the enthusiast community, I've seen reference to 'pixels per degree' to define fidelity.

This changes a bit based on how close the eye is to the lens/screen, but is a useful general measurement that also takes into account lens effects.


I think it's a good idea to use megapixels for resolutions.

720p - almost 1MP, 1080p - 2MP, WQHD - about 3.7MP, my monitor (advertised as "4K") - more than 8MP

Also we should use area instead of diagonal measurement to convey the size of a screen.


It's not. You give number of pixels not resolution. Resolution gives number of pixels and aspect ratio the total number does not. array[3840][2160] is clearly not the same as array[4M] Also your assumption of 4K is wrong. Its 4K not 4M it means 4000 and it means approximately number of vertical pixels. 4K UHD or just UHD however means exactly 3840 x 2160 thats whats written on the box of your monitor. People usually think of that if they say 4K.


Why isn't it RelaiVRty?


why not just called it OpenVR? Shorter and more to the point IMO.


Well OpenVR already is a thing. It's an API by Valve for interfacing with generic VR hardware.


test thread (please ignore, will delete after)


test thread (please ignore, will delete after)


test thread (please ignore, will delete after)


test thread (please ignore, will delete after)


Looks very cool, but you are going to run into problems with the name because people are going to confuse it with https://www.relativity.com/

That company sells a major legal document management platform, and has the standing, incentive, and capacity to drag Relativty into court for trademark infringement and easily win. The sheer number of people getting the word wrong on this HN thread is itself evidence of potential consumer confusion. Trademark infringement doesn't have to be wilful or malicious, and trademark holders are required to defend their marks or risk losing them through a legal doctrine called 'constructive abandonment' so it's probably only a matter of time before Relativity-the-legal-software-company sends a finger-wagging C&D letter to Relativty-the-bedroom-hackers.

tl;dr find a new name or get sued.


nothing brings out the haters like a job well done. great project. people will benefit from it.


Good luck to them, and I've bookmarked it to keep tabs on it in future, but that is an infuriating and in my opinion - very poor - choice of brand name.


Relati-"vee"-ty?

They could brand it like "relatiVty" or "RELATIvTY", I guess.

Or maybe "relativTY" (relative, thank you [open source community])?

Either way, by searching for "relativity vr headset" you end up on the proper website. So the point is kind of moot in my opinion. People might remember it as the "correct" spelling and end up at the right place anyway.


I agree that "relativity" isn't a great name: it's a mouthful at 5 syllables, and I don't see a logical connection between the name and the product.

But I don't understand why the name would be infuriating. Can you explain?


The name is "Relativty" without the last 'i'

That's the issue. It's an awkward 4 syllables.


"Relative-T"?


The fact that you thought it was "relativity" is exactly what makes it infuriating, I think.


Not the OP but I thought it was miss-spelled at first.


Infuriating? Let's not go there here please.

https://news.ycombinator.com/newsguidelines.html


Really? Wow.


Getting infuriated by something as shallow as a "brand name" inevitably leads to shallow comments.

Perhaps I'm overreacting to one word there, but fury doesn't make for curious conversation.


Can you give more details on this? Is that name already associated with something else?


‘Relativity’ would anyway be generic, non-descriptive for this product domain and cliche, but then to mispell it just to 'get a domain' (or whatever the logic was) seems poorly considered - every time I see 'relativty' it reads as a typo.

Choosing names is Hard though, I appreciate.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: