Relativty – An open-source VR headset
relativty.com
relativty.com
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.
(Plot of later part of Sword Art Online - really.)
Makes it easy to mistype the url, if you were to visit it later on.
"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...
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.
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.
I am still surprised that this is not more common. Even a biology-grade binocular microscope for 100$-ish makes a huge difference.
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 can only respect and envy the guys for pulling this off.
However, I can see where the idea would come from, because they do talk a lot about Steam on the site.
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.
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.
The build comes in under $200. I don't understand the whinging in the comments.
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.
I don't think the lack of controller tracking is a deal-breaker for those who want to get an open source VR experience.
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.)
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.
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 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.
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.
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...
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.
OP project may in fact solve this, I need to look deeper into the tech.
Now we just need controllers and maybe even base stations.
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.
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.
Huh? I'm out of the loop here. HTC is still selling the Vive Cosmos. What are you referring to?
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.
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?
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.
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.
?... 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.
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.
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.
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.
Inside VR, I just want more custom hardware... I want VR feet.
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.
> 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.
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.
Electronics are already pretty LEGO-like conceptually; you just have to solder instead of snapping things together.
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 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?
A sort of "Alice in Wonderland" mode.
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.
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.
But what do you understand "room scale" to mean"? Genuine question.
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/...
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.
If Oculus Go was open source, it would cost you much more than the retail price to manufacture it.
Yes, I agree, this is a tremendous achievement and a great project.
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.
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...).
Those in the form of Raspberry Pi or Arduino parts are more accessible but still not cheap and not so high resolution.
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.
I assume you could downscale the display to like a 1080p version to save costs as well.
Oculus Go doesn't have position tracking. I don't know what you're talking about.
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.
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.)
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.
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.
From bottom of my heart, hats off to those kids.
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?
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.
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.
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.
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.
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.
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.
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.
"c" could mean Colour. Bob uses a c2160p monitor!
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
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.
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?
You had VGA (640x480) for example.
But I do agree we should have called it 2160p not 4k, but marketing guys are like that.
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.
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.
Give it a few months and Google/Microsoft/Amazon will probably buy it up.
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.
[1] https://www.youtube.com/watch?v=JTa8zn0RNVM
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).
Though, I'd imagine you could use ones from cheaper 'phone-to-face' Google Cardboard-style ones.
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 ;-)
(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.
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.
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.
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.
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.
Choosing names is Hard though, I appreciate.
But I don't understand why the name would be infuriating. Can you explain?
That's the issue. It's an awkward 4 syllables.
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.
Perhaps I'm overreacting to one word there, but fury doesn't make for curious conversation.