#virtual-reality
1 messages ยท Page 46 of 1
Figure out the forward vector of the HMD in a robust manner. Then use it to drive shoulder then hip movement.
let the HMD move independently within a range
but then have the shoulders follow along after
Repeat for hips. If the player moves forward significantly (30cm+)
match shoulders hips to HMD forward.
Or maybe you can use physics constraints
yeah. My code for doing this is super annoying
but it works for all orientations with histeresis
you just have to feel it out
also whooooaaaaa
I just extended my Vive cables by 50 feet
works like a charm, active HDMI cable and USB-over-Cat6 bypass
Both nice and dayum that's a lot of cable to reel up
Unless your play area and workstation are in very different areas
workstation is in the office, and the play area is in another room
office is too small to fit a vive setup, and my wife likes to work there too
Fair enough. So basically you increased your Vive play area a lot!
What size space?
and I just piped out through HDMI and active USB into the office
so I could have a terminal there
Ah yeah
that didn't fly when I got a 4k monitor, because displayport won't come longer than 1.5m and HDMI won't do the trick over that distance
Yeah... the short cable lengths of displayport are quite bothersome.
so now I pipe in the terminal into the vive room, and rearranged stuff around there to get approximately 2.5x3m
Had a hell of a time getting a cable to extend my DP range. In the end, I only ended going from 1.5m to 2m
Enough so I could awkwardly route the cabling from monitor to floor PC
now you're thinking with portals
That's a pretty good sized play space.
I've got 2x3 (or actually a slightly more ego bruising 1.9 x 2.9m)
ehehe
it used to be 2.5x2.6
squeezed the desk into the corner, and ignored the chair for the desk
so I've got to remember to put the chair next to the desk before I play
Friend of mine has a 4 x 4.5 setup and complains when he's here ๐
haha ๐
at the lab we have a crazy 4x4 space, it's a little isolating
feels so empty when I play there
It's great if you're playing a room scale only game
But I've found that the benefit goes down when you've got a good locomotion setup
mhm
Then instead of been the space in which you can move, it's just an 'envelope of active play space'
yeah, just some more void to get confused in and trip yourself
After all this time....
I still think the controllers I designed for VR (visual concept, not mechanical engineering) would still be great
I mean, look at that sum-bitch. Way ahead of its time. It's fully compatible with Vive, Touch and Gamepads!
You could even map Elite Dangerous to it well
... That's actually not the first time I've heard that joke
It does look like an iron though. lol
Doesn't matter what it looks like on the outside!
It should feel a bit like a skinny gun handle with the HMD over your head! ๐
Or a joystick
I think the vives being smooth helps a lot
plus the obvious reason for the redesign from the pre is that you could hit yourself in the teeth with it
not so much anymore
Hit yourself in the teeth with the pre??
yeah it has an edge on the back that angles up
sorry that's not the pre
it's the old dev kit
I guess the pre has the new shaped controllers and camera but is still slightly older than the current version
we had the ones in the picture at work for ages
if you put like an apple in the player's hand and it was set forward enough those games that make you push shit towards your face to eat it would cause you to lose a tooth
good times
bbl shopping
Hehehe.
@dry fjord That is what I am doing, but I want it to look as if it was played on a normal 1080p screen.
Isn't it wonderful how SteamVR starts when you start a VR Project?
Only to still not enable VR play, cause SteamVR started probably after UE4 checks if SteamVR is on
Such a useful feature
Imma cry as much a want!
๐
@neon egret I feel ya, man. Lets cry together like grown men.
@neon egret /waaa (crying with you)
Auto-launch isn't the problem.
It's that it launches and Unreal still doesn't pick it up.
I think they're fixing it in 14
closes off everything and starts up steamVR first.
Oh that'll be good.
Do you know when 4.14 is dropping?
technical director mentioned in a recent stream that he was lagging on implementing it, he's sorry, but it will be out soon. Probably 1-2 weeks
Couple weeks eh...
Hmmm, will probably be able to integrate it before I drop my own thing.
Hopefully it doesn't break too much stuff.
And hopefully forward rendering works well for me.
why not test it beforehand on the preview?
I'm using a plugin that...
last I looked hadn't been updated to 4.14
The plugin is fundamental to my stuff. Can't not use it ๐
which plugin?
well its not 'great' but it is way better than TAA's blurry mess imo
I've been using screen res 150 and FXAA.
Substantially better on text
Than TXAA
Haven't compared to MSAA yet
yeah, I just am not going to use text as much as I can, and switch to forward when it's all ironed out
I'm really impressed at the gains of fwd though
lol
I agree, I just like that it compartmentalizes all of the features that I don't need
but honestly deferred is better and is going to be for a while
if I wanted my game to look like that I'd just use Unity
I hope they release Robo Recall as a learning template
Then I can stick my own stuff in there and run around! muhaha ๐
TAA works best on extra screen percentage
put 150+TAA and it looks real good
but remember, TAA is a post fx, so it scales with the resolution quite a bit
and its already expensive
so if you budget for 1 ms TAA time, then go ahead
it also doesn't look too nice when you shake your head, because of the 'temporal' part
indeed
TAA best for deferred
but its expensive
the cool thing is that it works well on sparkly things
hey guys, anyone had any luck creating VR trailers on the Vive? haven't found a good workflow for generating 1920x1080 HMD footage :/
Shadowplay is quick and easy, so is OBS
ah
you need to alter in the render parts of the steamVR plugin
change U and V for the render rect to something else within
I attempted that but wasn't sure it was working for the Vive?
could you confirm that isn't Oculus only for some reason?
otherwise there is no good and easy way to make arbitrary cameras for this kind of thing
thanks for helping BTW
I can confirm people have gotten it working for 16:9 output to render window
for Vive
it's not good, it's a crap resolution but it'll output properly
still wasn't working though, I don't normally build the engine from source
I wonder if I actually rebuilt now ๐
hehe
how can I be sure my changes are running in engine? pardon the noobish question โค
once you've compiled the engine for the first time, and you recompile, if it mentions the files you've touched in the compilation output from Build you're good
other option is to record gameplay in a replicated sort of way, then replay and edit together in matinee or whatever they're calling it these days
ah, yes, so it had not mentioned the SteamVRRenderer file
right, I thought of that but really don't like that workflow idea, sounds hellish hehe
yes
will go back and make sure it is compiling now
I am not certain you must recompile the entire engine, I believe the plugins should be hermetic
unless the part is in the engine, then...
yeah, I when I compile other plugins they are discrete
pretty sure you're right
I did try recompiling the whole engine too though out of confusion
I haven't done this so I'm not experienced, but I think I get the idea of the plugins
because they'll each live in their own DLL
easy to swap in and out
pain in the ass for debugging because you'll have to generate symbols for them blah blah
good luck
taking another stab now, thanks ๐
I'm not yet in the trailer phase, just in the vlogging scene now, so I'm glad for any work done to further these types of things. Hopefully Epic will patch this up soon before I do anything formal. So thanks to you for experimenting
keep your commits tight just in case you figure something great out, we can piece together a pull request
in other news, Hans Zimmer's Interstellar soundtrack still gives me chills, keeps me going ๐
yes, the commits are aplenty ๐ hehe, and huge Hans Zimmer fan over here
he always does a fantastic job
sure,
Water Planet
lol, did not intend to take over the channel with an image
that is the music release cover
it also has the game counterpart
Steam Workshop: Greenlight. Water Planet is an audio/visual game that involves immersive first person exploration on land, underwater, and guided flying in an orb transport while listening to the music from Water Planet EP by VI
there is the Steam Greenlight link
0:57 in your video ๐ ace use of phsyically based shading
thank you ๐
we're really happy with that level and how it looks ๐
the others have come a long way too
that whole video has no additional post processing outside of UE, we've made some things look better since then too
cool. I'm assuming you added some type of wash around where the geometry meets the water?
because that would be my one and only critique
yeah, we experimented with a few things actually
but a lot of them were not VR friendly ๐ฆ
d'oh
kind of just "smoothed" things over there with fog ๐
yeah the fog looks great, definitely alleviates the transition
but yeah, it has been interested seeing what we can squeeze out while still trying to keep VR rendering a reality ๐
interesting*
we only have a 960 and a 1080 here, runs perfect on the 1080, and a little choppy on the 960, still have some optimizing to do
but cool, I hope you like the world and concept, I feel like if you like Interstellar, Hans Zimmer, scifi, etc, you will hopefully enjoy the game ๐ I will try and remember to ping you when it is ready ๐
great, let me know!
Weirdest crap ever right here.
The behavior tree I made for an enemy character works fine in VR mode when the headset is off my face and put on the ground. Everything works meaning the enemy NPC moves around properly, performs the required tasks with no issues, etc. As soon as I pick the headset off the the ground and stand up with it the behavior tree goes nuts. Starts looping tasks when they are only supposed to be used once and movement ends up being random rather than being scripted (just walks where ever)
FIXED: Usually according to UE4 logic the player is technically in the center of the play area which is always a pain for developers. AI Perception does not follow this logic and will keep updating based on the location of the player anywhere in the play area.
If anyone would like to try out my jam Vive game, here's a link: https://crackpotgumption.itch.io/flowstar
big thanks to @neon egret for his plugin, which enabled the game. ๐
cool! will check it out
@empty sundial I can't get any song to play?
No?
Hm
I grabbed both guns as well but dropping them made them fall through the floor. After dropping them I could grab a gun from my left side but not my right which kind of ruined it from there.
Nope, no songs populating.
That would be the reason. Hm. I'll see if I can recreate. Can you send me the path that the game is located in?
C:\Users\Chris\Desktop\CrackpotGumption_Flowstar
the orbs?
The things you shoot
They accelerate, but have a max speed
I only played one song (some Funk sounding one) but they didn't speed up at all, that's why I was wondering.
The funk one is pretty laid back
Ah
although speeding them up isn't a bad idea
Noted.
honestly 150% sp + TAA costs about the same on my scene as 100sp + MSAA with the new renderer
and the first option looks way better
I haven't looked at the latest preview but MSAA has a way to go
I think you may have to disable a lot of engine lighting features for it to look good
yeah and my scene is reliant on heavy lighting and shadowing
you're using r.screepercentage right?
hold on, is 100sp + MSAA supposed to be as expensive as 150sp +TAA?
it is in my scene
I'm currently at 130 + TAA... and I was expecting to get a performance boost with MSAA, should I adjust my expectations? :S
no
you will get a perf boost from forward rendering
but msaa is really costly
use TAA + Fwd
I'm currently using forward already in 4.13, but I'm using TAA
already noticed a performance boost of course
I guess I'll try 100sp with MSAA and see how that holds up against 130sp TAA
in terms of performance and quality
how does lighting affect MSAA performance, though?
havent done tons of testing but my scene is heavily dependent on stationary lights w/ dynamic shadows
whole scene probably has 70-100
holy crap
how have you been able to get proper performance with stationary lights w/ dynamic lights?
the only dynamic lights I've been able to use are cascade shadow maps
but as soon as I turn on dynamic lights on a small point light... performance dies on me
even though I'm maybe only casting 10 shadows
tons of small tweaks and optimizations
tell me your secrets ๐
I've had a hard time, and I've also been unsuccessful in finding any information
i dont even know them off the top of my head but one really important thing is to set that max draw distance in lights
i have mine at 3000
but my environment is very dungeon-ey
ahh I see
I'll play around some more
I've had surprisingly good performance with capsule shadows
I'm waiting for this:
Hi, I'm Luoshuang. Just now I have finished DFIS for UE4. This idea is derived from Capsule Shadows directly, it extracts light direction information from indirect lighting cache, and generate indirect shadows by cone-tracing against distance fields.
GitHub: https://github.com/AlanIWBFT/UnrealEngine/tree/DFIS-4.14
Note:
- Currently tiled-culling is not implemented, so performance may be an issue.
- This technique shares the same drawback with RTDF: the umbra will always larger than
it's coming in 4.15 apparently
should be as cheap as capsule shadows
http://www.gameplanet.co.nz/news/g5828db5d55389/Vive-gets-wireless-more-expensive-with-new-add-on/
@dry fjord
imagine you've seen this but in case others haven't
but damn that's a hefty cost just for no cables, might be worth it still
Would be 100% worth, but I don't think it's legit
i know gp is legit, i imagine they investigate their sources
unless you're implying it will be frought with issues, in which case - probably!
I'm sure it works, but I'm also pretty sure it'd add 20ms+ of latency
yeah, no "noticeable" latency - for who, casual gamers?
Unless they did something revolutionary
well lets think about it
how much data is being sent vs how much can be sent
in other words, bandwidth
i imagine that would be the deciding factor, and without doing any research whatsoever, there is probably too much data being sent to that wireless attachment than you can send at a low rate
"HTC has quoted the Vive needs at least 15Gbps bandwidth"
"the company will sell a bigger battery for the add-on in the future. At the moment, it will only come with a standard battery that lasts for an hour-and-a-half. The bigger power source will last longer, but you'll have to carry it around in your pocket or in a bag."
tldr the battery lasts 1.5 hrs
if they have any sense vive2 will have that as standard
with the new bits on your belt of course, not on your head
Ehhhh, maybe
The problem is to solve a lot of the current gen issues, you need more data (better screens, better resolution per eye, still pushing 90FPS)
And there is nothing in wireless tech that can meet current demand for throughput and latency
Let alone even more data being pushed
@jagged vale yeah, I linked that here few days ago ๐
heh, so did HeadClot
also it looks like a real deal, HTC is handling the sales at the moment it appears?
but yeah, VR needs to go wireless, badly
wires are ok on seated experience
i look forward to it if it works 100% and when they have a better battery
but since room scale is what people are now focusing on, wireless has to happen
even if it comes at expense of other features
Don't miss out on @empty sundial VR game playthrough, maybe you'll learn something?! https://www.twitch.tv/victorburgos_
Anyone implemented in UE4 some kind of dynamically scaled render settings like that Steam Dev Days talk about perf in VR mentioned?
I would like to do this, but I need to know how fast I can change the settings, and if there's lag involved as well
Well you certainly can't use setresolution per tick
I've tried that. Doesn't work well ๐
yeah I did
@winged shale
I didnt go the normal route of putting a delay in or anything
I found what worked best was to keep a variable at 0 and every missed frame --, and every 5-10(whatever works best for your scene) frames over 85fps ++ the int. and do a check if its <-45 lower settings, if its >45 raise settings
@real needle Nice work. I like that!
thanks @clever sky
the oculus branch of the engine has a command for that
@empty sundial (: Will try it out later!
Does anyone know if there is a simple way to get replication to work on pawns and it's components? (hands and Head) because the way im doing now is really laggy :/
https://gyazo.com/e7a418d44185c4e257288964b064dcc3
for starters, don't replicate transforms, it's unecessary
I'd imagine their scale isn't going to change, and the network code has optimisation for replicating vectors / rotators
he is doing VR
transforms is the fast way
im doing it that way for now
but the better way its to do it location+rotation
also becouse rotators can be done in less precision
vectors are also done with less precision when networked
(being in VR is irrelevant)
I might also not use reliable server functions to do it, and just use repnotify
also important to replicate the "local" position and rotation
I don't think you want things on tick to be reliable
no reliable doens't work at all
you can put it at 10 per second
and interpolate
but that is very easy to fix
just do it later if you need it
no, reliable is not what he wants
what im doing right now myself, is to replicate 3 transforms with a RPC, then on Tick of everyone, i interpolate them
whats RPC short for?
reliable RPCs can cause considerable latency, because they have to arrive in the correct order
to set the final transform of the components
but here you do not care if they do or not
since only the latest transform is relevant
yeah, but then you have to allways get the latest
a simple way its to send a unique number with the rpc call
they can arrive out of order, thatz why i tell that
@frosty robin - RPC = Remote Procedure Call
Okey, so the way im doing now is completly wrong?
its fine for a prototype phase
yeah, but im getting only 10 updates a second.
its easy to do interpolation, you can just use a lerp and you are done
works fine enough for prototype purposes
@frosty robin - I would look at using just local translation / rotation and handling it on repnotify
okey.. this is my first networking project though.. so im kinda off xD
not sure repnotify is good for something like this, that is sent on tick or nearly tick
im interpolating to the current replicated transform on tick
you can do it initially with just setting the loc / rot, then later you can optimised by having a desired loc / rot, sending it at whatever interval, and lerping as necessary
interpolation is going to be the best way to reduce network traffic overheads
but you may find you don't need to do it
I tried increasing the event tick to 64t/s but im still lagging alot.
rendering framerate != gameplay tick rate
well, in the same pc pretty much is
server tick rate is not client framerate
unless you are in ArmA XD
But how do i change the server tick rate then? because im using listen servers now.
I guessed that an increased event tick would lead to a faster update aswell
listen servers use game framerate as tick rate i think
@frosty robin - read through this if you haven't already:
http://cedric.bnslv.de/unreal-engine-4-network-compendium-released/
yeah.. im not getting that update.
im getting about 10 updates a second.
or maybe less.
listen servers will use game tick, yeah
at least i can see a massive slowdown in networking if the listen server doesnt have the headset on
thanks amber, ill read it.
Both client and server is using VR
make sure you have the second headset "on"
yeah both are on.
the headsets have a proximity detector and go into deactivated mode
if that proximity detector is off
alright
but the way i've setup the blueprints now.. it "should" give me a higher updaterate then 10/s
?
dont use repnotifty to do this stuff btw
im not
not sure, but repnotify its been buggy quite alot for me
and i mean REALLY buggy
to the point half of the stuff isnt called
it's strange though.. because if i use the 3rd person template as a third client + 2 VR.. nothing is lagging for that client... so the replication seems to be smooth to server
but if a VR pawn looks at another VR pawn, the client will see everything lagging
what ive done myself is not too different that what you have, and it runs fine
over real network
with a tester on the other side of europe and an american one too
yes
just a lerp
nothing strange
lerp locked at 0.7
i know its not optimal at all, but for now it does its job
allways find the easiest way to get it working
hmm you're just using the old value and lerping to the new? you're not driving the lerp alpha with anything?
hmm okey so i should set that when im setting the new values.. or when im changing the world transform?
and i should change to local transform instead?
when you change the transform
Okay
use the RPC to replicate the new transforms to the server
that function sets the Replicated transforms in the server
some time later those replicated transforms return to client
so then you interpolate the current transform of your components to that replicated variable
in fact, you interpolate the transforms of your components to the replicated variables on tick, allways
Hmm i can't seem to find any function called "RPC" or is it just a part of custom events?
It feels like i should read that book haha
the rpc is just replicated functions
okay, so basicly i should make my setting variables into a function instead.
also read the book
yeah im reading now ๐ thanks for the help
Hm
Tested PSVR.
You feel the lower price. Even if sold more (due to obvious reasons) it's (for me!) the one that gives the worst result
Rift and Vive are kinda equal. So idk. Let's see what the next year brings us.
Vive gets the 220$ wireless addon, which is a huge deal
IMHO, I prefer the Vive's roomscale features and the motion controllers just feel a bit better as they aren't so peculiarly lightweight
I don't have room for room scale experiences
i have all 3 side by side
Anyone have trouble with the touchcap events on the Left touch controller?
i think oculus is better, due to Touch and software such as ASW
tracking on both oculus and vive is similar
i run both on the exact same space just fine
Everything works fine on my right controller, but my left ones don't seem to have values != 0
Vive will have ASW soon enough
the one with the little robot and stuff
that one has fully implemented capacitative
so you can compare it and see if it works
turns out there are 3 capacitative
good call, brb
and the trigger
we do some right now
made my own animation blueprint
but for future stuff will be more robust
I pulled some last weekish, so maybe those were it?
i saw it had a demo level and stuff now
I don't know, the artists took care of that ๐
i had to align them with the motion controller
i just grabbed the zip and handed it off to them
ya, we have to translate our hands up a bunch compared to the motion controller root
i got it from a callum underwood email
I cannot send it to you (NDA)
but ive signed that nda too
Let me check
not sure if they have updated it since i downloaded it
but if you say its just .fbx and its offcenter
then its those
were you able to setup the material well?
yours is newer then
mine is OculusHandDemoAssets.zip
do you have the DL link?
i have access to everything
or link to forum thread
as i cant find it
my dl link to oculus handd demo asset.zip was sent on a "TouchCareAndFeeding.pdf" that they sent me when they sent me the things
@wicked oak hi! If you have time, could you check out the #multiplayer channel? I went with rep notify after conversation with eXi. But I'm getting a ever slower update from client to server now (like 1 every 10 seconds, very sporadic) I put up a picture there and didn't wanna spam the vr channel, since it's multiplayer related.
Neither him or me could make sense of the results :/
thats too odd
Everything looks smooth from the clients perspective.. but not from server
I'm using a listen server if that makes any difference
and im doing that too
Yeah feels very strange.
@wicked oak I got it from somewhere on that Oculus VIP forum. I'm 99% sure you've signed that NDA, but I can't go giving out other people's assets
did they change the vip forum adress or something?
it gives nothing on me
this using the same account that has the game creation credentials and that kind of thing
If you are logged it it should be one of the forums on the right hand drop down
VIP IN DA CLUB
but I'm pretty sure I just clicked a link in a PDF somewhere...
it was an AWS S3 link
Is there any way to use CPU and GPU profiling tools with GearVR?
turn on developer mode, run your app before inserting phone into Gear VR, open console, type in: stat startfile
put it into Gear VR, test your app/game, once done, take phone out, open console, type in: stat stopfile
close your app, copy stat file into your PC, open it with profiler and there you have it
you can also use stat cmds in the same manner (to get basic stats)
Awesome, thanks!
@spring pond Thanks for that screenshot. I just got my golden ticket to the VIP Room last week but didn't even notice it on the forums, lol
what's in VIP Room ?
It's just another forum where people can talk about Early Access SDK/Firmware/Touch
(ie. anything under NDA)
Has anyone figured out how to get the Vive camera to be activated/deactived with UE4 yet?
You mean the actual camera on it?
spigen armor case looks like it would protect the s7
gearvr seems a very poorly thought out system
why not allow an option to deploy over wifi so it can be left in the headset
@velvet cave Yup. The actual camera on the Vive.
I doubt that's easily possible @pallid echo
The Camera might just be a second device that UE4 doesn't even have registered
Its no where close to easy, I tried but only got so far before I was like, "Welp, this is beyond my knowledge ATM" lol
Search for something else
Check if there is docs about getting a simple webcam into ue4
cause that's what the vive cam is
@pallid echo
Well look at that:
Hello!
I had to access the web camera in-engine for one of my projects, so I though I would share it with the community!
https://github.com/Temaran/UE4Webcamera
I implemented it using directshow, so only windows support for now.
Hopefully it's useful to someone :)
And if you would like to contribute, please do! It still needs multiplatform support, Easy BP support etc.
And this ๐
Exposes access to the web cam video feed as a 2D texture
leave the vive cam turned off, strap another webcam to your face instead
Na seriouly
Why
the camera should never have been added, it can't handle the bandwidth of the feed AND do tracking
the number one way to fix vive problems is to turn the camera off
maybe you got lucky ๐
ah I see
the problem must be in steamvr then
that's cool because all i wanted to do with it was see what was going on around me
Yeah you can add it to your fav braod caster software
and add it as a webcam
and show both
awesome
I'm testing out my new MSI 1060 laptop, and playing games on both Vive/Rift works great, I'm pretty impressed. HOWEVER, in editor, the rift drops to 25 fps where the vive works flawlessly, and I'm having a hard time troubleshooting it...
I have both SteamVR and OVR plugins enabled, and I've never had an issue with them interfering with eachother before
Instanced Stereo on, config files set to low like I always have them
It's very apparent that it's not lack of hardware performance but something else. Could be the forward renderer...
you may have to disable the SteamVR plugin in UE4, or just disable the OVR plugin and use the SteamVR handles for abstracted headset/motioncontrollers
I could do that, but there's one issue: The fucking menu buttons on touch
fffffff
They have their own handles
I don't need the capacitive events, but without the menu buttons it gets hard
Imma try anyway and see if it fixes it
I'll report to Oculus if it does
see if you can find a way to thunk in the menu button stuff externally for now, and bug Valve about it as well as Oculus, and Epic
AutoHotKey might work, but it is a specific OVR SDK delegate
does this issue persist on a desktop?
I'll test it
In a bit, packaging a project built on source...
Gotta love those rocket builds
Rocket packaging speed!
๐ ๐ ๐
you actually knew that one
๐
what is it?
First Contact!
ohhhhh
Same team that did Toybox
right, that
do they have a tech whitepaper or anything, they included OFR with it right?
something that talks about how they slammed things together to make UE4 ideal for oculus forward pipeline, etc.
ideally a github repo
From when they released it in 4.11
yeah....
I haven't seen anything else
Runs great without OVR
But it doesn't seem to respond to "Set Tracking Origin"
The guardian system seems to override Chaperone
Which I don't mind because it's actually pretty good
It highlights around your hands when they get close to the boundaries
Worth to point out that the Oculus Library can be used without the OVR-plugin, but Oculus Input plugin doesn't work without OVR
when they fix that bug with chaperone in packaged games you'll be able to do a custom cage in UE
I have a multi step cage that hilights near the offending part, but goes from weak, to mid, to STAY THE FUCK AWAY RED
it has diagonal "warning" signs on it
sweet
#safetyspikes
The Oculus Guardian is really nice though
It's more smooth than chaperones hard edges
guys can we do a quick public announcement
this is not what VR is for
thank you, please continue
wow
Tragic if the first released product from the SteamVR tracking licencees is a doll
it'll almost certainly be a gun controller
@dry fjord Highly disagree! That's the main thing that VR is about! fondling mannequin boobs!
ok but only if you invest equal time into fondling MANequin crotch
please represent all demographics
much more cost effective to just surgically implant neural waystations in motor and somatosensory neurons
who wants to buy a bunch of lifelike, eerily warm squish mannequins anyway
Tens of thousands of people!
I'd love to squeeze Mr. Incredible's foam thighs.
I would snuggle up to one like a pillow
which is what it may be
actually those legs are way off
I might jack it to that later on
yeah NOW I'm keen
I don't believe that claim is even possible.
i dunno, but its a shit ton of data to send in 2 ms
unless the bandwith of the channel is absolutely huge, im not sure they can do the 2 ms claim
the 10 ms claim was already hard enough
It's not 2ms "motion to photon", 2ms is the transfer delay. There's other stuff like compression delay etc. Their 15ms statement is closer to the total delay of the system, but probably 15-20. It might be totally fine but I haven't heard anyone I'd trust to make a judgement try one
I think they're trying to be market first, I'm sure there's better stuff in the pipeline from unannounced products
that second gen its going to be really cool
4k displays, wireless, inside out tracking?
@wicked oak That sounds like a second gen that's 5 years away... according to Michael Abrash ๐
Personally, I think we'll prbably see second gen HMDs in a couple more years. A lot of competition in the space, even if the tech is hard.
So manufacturers like HTC will want to press their advantage in building a market while they can
Mainly I'd expect to see in second gen, improved HMD design, better res, slightly better FOV, improved controller design, possibly an interesting 'big' feature like camera based motion tracking (hands/body) or eye tracking, or wireless, but not all of them at once.
15ms total delay is just about doable
20ms is the point where things start becoming unplayable
Better all around and would make most of us want to upgrade. But not the Abrash 5 year prediction spec (4k/eye tracking/inside out/wireless)
i think the controllers like the Touch are real good
the step over those is basically just going with full gloves
and gloves WITH feedback
@wicked oak Touch is trying to be what knuckles is
Getting rid of that grip button makes it so much easier to design for everyone
Believe it or not but people actually have issues with the grip, even on touch
It's not entirely natural
Where knuckles are, since you're actually doing the same motion you've done a million times to grip
The grip button is extremely intuitive if you use is, especially from what I've noticed with our user tests.
Like... I hate that some devs think that any amount of learning is too much for users.
There's a difference between taking 5-10 minutes to get acclimatized to something
and a few hours to come to grips with something.
what do we gain from 4k displays in the HMDs if no one has GPUs to handle that at smooth fps
I think the display tech is really not the problem, displays get better extremely fast, but GPUs only improve really slow
it will take like forever before GPUs can handle 4k 90 fps with a game that doesn't look like some mobile game, so has stuff like dynamic shadows at an acceptable view distance etc
With 4k display, we assume decent foveated rendering is available
and in VR it would be really nice to have something like GI too, but even without VR dynamic GI already really slow
Also ASW is widespread
At worse, it'll be an option for developers to consider
Do I take advantage of frame rate or resolution?
And for gamers to decide if they want to build beast machines
with SLI or something crazy
ASW is nice but the better the resolution of the display the more you see smaller artifacts caused by something like ASW I guess
Well, you'll see more of all artifacts in more clarity
But the overall visual is still much better and nicer than the current displays
At least that's what I'd wager would be preferable for most people.
and even if 45 fps would be acceptable for the majority, GPUs would still be too slow
How far out are you thinking?
I'm using Abrash's 5 year timeline
4k in 5 years is pretty conservative.
well I think in 1 year you can easily get a 90 fps OLED 4k display. but there will be no GPU that could power that in VR
Eh. Worst that'd happen is you'd have better fill rate and some people building beast machines to power it.
I'm just talking about that the displays won't really need to evolve any more because GPUs are so slow
the screens exist, I think the cables/transfer protocols are more of an issue ๐ but 4k HMDs will happen within 2 years
I mean I don't disagree with you; that GPUs are lagging displays
I want to see more innovation in optics and FOV though, and HDR! Resolution isnt the all important thing here
GPU can be solved with new rendering methods ๐ esp. when we can do foveated rendering
But I think someone might make it happen anyway just because some people will have use for it and for the hype marketing factor
and even with 4k displays you will still see the individual pixels
Especially if decent eye-tracking happens within that time span.
if my research goes according to plan, by May I'll have made a prototype HMD that is capable of rendering 4k by 4k per eye on conventional hardware (including cable)
@winged shale really?
Dont forget all the VR rendering improvements we've had over the past 2 years! instanced stereo, hidden mesh removal etc
yep, the only issue right now is sourcing an appropriately dense display
@winged shale You work on VR hardware for your day job?
๐
at what refresh spychiatrist? DP 1.3 or HDMI 2?
90Hz
it's a few things, a compression algorithm that is my thesis + eye tracking + neurobiological insights
4k per eye would be nice
Dayum. You got access to some pretty bad ass hardware to pull that shit off.
to not see pixels you probably need 16k per eye
Also you can probably say good bye to your game. You'll be too busy working on next gen VR HMDs!
@royal anchor it has much more to do with the subpixel arrangement
most people say between 8k and 16k yeah
now for the acuity of the fovea, that's a different story
16k would be 20/20
@clever sky I worry about that
@winged shale arrangement? lets just say amount of subpixels, isn't that more accurate?
I mostly meant area
space in between pixels
@winged shale don't you hate when the money diverges from the fun ๐
black gaps cause screen door
just moving to hexagonal arrangement will have extreme effects on the visibility of the screen door
Has there ever been a hex arrangement?
not that I'm aware of, no
it would be hard to render and then transform appropriately for it
I see. So... it's gonna be quite the feat to pull off
do you think the chromatic aberration stuff will ever be improved much, so can the lenses still get much improvement? I think thats what annoys me most when looking at the vive
but breaking up the longitudinal and lateral lines that make typical displays look like screen doors would aid your perception in forgetting
So how come people hate the pentile display so much?
Doesn't that operate on a similar principle?
yeah, that will be improved with better optics, we need better manufacturing tech to make them non-cylindrical
it's hard to optimize for all points on the flat screen
pentile is using 2 subpixel per pixel, so that means you have 1/3 less subpixels than with a same resolution RGB panel
ah
having the ability to curve the screen + having the ability to manufacture arbitrary diffractional masses will make things a lot easier
its cheating, just like AMD with their bulldozer cores basically, they aren't real cores and pentile pixel aren't real pixel
skips?
which is part of why if you see straight blue on screen it looks... offset. The other part is perceptual and to do with chromatic aberration
hold on let me see if I can find the image
no I'm pretty sure that was a seeeecret image, but basically the overlay pattern of blue pixels is sized up to like 1.4 times, so per square area one blue is skipped, but the blues are much larger as well
has to do with visual acuity and shortwave cone distribution
Galaxy S8 will have 4k screen, so whoever has Gear VR can see if 4k is enough for VR or not ๐
Valve has been tweaking the chroma abb correction in the past few udpates
^^this, I barely notice it anymore
but that might also be because I'm getting dev legs
and/or becoming more complacent
VR legs are BS ๐ lame excuse for bad design ๐ I still get ill right away in the old Tuscany demo and its been uh.. almost 4 years since I saw that first
I didnt have any demos when I got my alpha DK1 ๐ it wasnt great, but still blew my mind at the time
also poor drift correction, IMU sensor fusion algorithms, etc.
blah blah, but yeah DK1 changed the track of my career
yeah tuscany stairs fucked me hard
I did positional tracking with DK1+Hydra like many ๐ it was WAY better than IMUs but not exactly accurate (damn warping)
@heady parrot I saw the steamvr beta changelog with those chroma abb improvements, did it really help?
for some reason valve has only released beta versions for like over a month I think, thats unusual for them
I never used the beta stuff
I actually haven't tested it yet ๐ do most of my dev on the Rift
yeah I swapped from beta at home after having some issues that cost me a few hours. In the end I think it was actually an USB issue though
for some reason, the Rift's ergonomics are shit for me
really? thats rare. What specifically?
I only have a vive so I can't choose ๐
I can't get the device tight enough on my face to get it to focus properly
I find the Vive so uncomfortable compared to Rift
yeah they messed up with the glasses
I'm 6' 5" and have a large head
@winged shale You gotta pull them on from the front
so I guess I'm at the > 95% ergo distrubtion
Then slip the back over your head
I got a large head too, so confused why you cant get the straps tight enough? I'd have though that would be the last problem
I've got a big head too
And glasses
If you don't slip them on from the front, they suck shit.
did you consider getting those lens inserts or a separate pair of "VR friendly" (Harry Potter lookalike) glasses?
it's either the straps aren't tight enough, or the optics are focusing the collimation plane too far for my naked eyes
what I find super annoying about the vive is that damn sound the lighthouses make... thats so annoying. I know its not loud, but very noticable.
and when I have my glasses on it distorts it so I can see the flat edges of the display
the lighthouse itself does not make noise
John you got the Pre or retail? I hardly notice it on the retail ones, but I got them on tripods not walls
Pre
if you're getting noise, it's vibrating on something
they have motors inside
lol
the retail ones can shut off when not in use via Bluetooth
not much louder than the VDK1
oh they are louder than retail? I thought hardware didn't change
I have the pre, and mine are mounted firmly on tripods; no noise at all
@hard light quite sure its your ears
I can hear noise from mine when it starts up
they tweaked it slightly. It implements the same thing but I've noticed my CV lighthouses are much quieter
I recently moved my computer to a different room and piped in Vive stuff to the Vive room, KB, Mouse, display, etc.
the Pre does not have the bluetooth control, that was added in retail
and without my computer noise I can hear them loud and clear
I doubt I'd hear anything from mine unless I were stood with it right next to my ear
Only in a totally silent room do i barely notice the vive lighthouses. They are pretty quite
also the pre houses might have had more use on them too
probably about twice as much as the consumer versions by now
and if you went from the VDK1 mounting solution plates with the god-awful 3m strips to just screwing on the Pre lighthouses, there will be sway
therefore there will be torsional stress on the high-RPM motors
with VDK1 they included these weird plates for mounting with 3m tape on the wall, that they also included
the squish allows for an ever-so-slightly noticeable amount of sway from building movements, etc.
VDK1 is the one before the pre?
yeah
well I never saw that
the sway probably causes some bearing wearing in the motors, it's not good to move your lighthouses when they're in full swing
I can take a picture today of them, they're still up on one of our setups
well I just hope valve somehow will be able to remove such high frequency motors from the lighthouses
are there more than 1 motor in a lighthouse atm?
ah that reminds me, didnt finish watching all the Steam Days talks ๐
there's 2, orthogonal
two
well is 1 less loud?
veeeeery likely
my solution atm is just to have music playing, then I don't hear the lighthouses
I just usually find it annoying to have music playing when programming
with the old VK1 lighthouses, sometimes the vibration was caused by bad sync so turning them off and on again fixed it
but less annoying than a constant 4400 hz tone
my solution was to keep my computer in the same room, and keep the 1080 running at full-tweak with 100% fans all the time
haha
now it's silent in there >_>
yeah that works too
Headphones? ๐
but the GPU fans are also quite high frequency noise
already a PITA to put the headset on and off to try out a tiny feature you just did
yeah
I can slip the Rift on in like a second, but I dont wear glasses in the headset
and then you're just sitting devving around in a silent room with squeals
yeah
any of you got the new Vive cable yet? I was pretty bummed when I bought a Vive like a month ago I didnt get the new one
theres a new one?
ordering a few to replace fraying pre cables
yeah thinner, not a ribbon
I'll let you know about it (hopefully) soon
ok
In tech headlines today: Apple's $300 book contains 450 photos of Apple products
sounds like the Onion
To be fair, it looks pretty swish. I'd buy it for like...
$25
I dislike Apple so I'm a pretty bad sales target
so basically a catalog you pay for? well done Apple
and I don't really want to turn my living room into an Apple store
Yeah, I get that. I'm just a design fan
totally fair, that's a thing to admire
I dislike apple too but it would be nice if they would build a VR headset
So setting aside everything else I don't like about Apple, I do admire their designs.
From an aesthetic viewpoint.
Functionality is shittttt
IDK how they would lock the apple VR down to Apple hardware, their computers can't push those frames
unless they're going to go the route of mobile VR, which is increasingly flooding the market
If Apple does VR, it'll be mobile.
++
Just like Google
I don't wanna see more mobile VR
++++
agreed
Well, if someone manages to do good positionally tracked mobile VR
probably Oculus given that's what john's been working on the entire time.
It'll be more interesting.
VR needs extreme GPUs and on mobile you don't have those
so that leads to VR being a 360 video thing
Only if you want good visual fidelity
the damn things are overheating trying to push framerate at 60fps, they sure as hell aren't going to be able to add SLAM on top of that
and after 30 minutes your battery is empty
And you know... a lot of experiences are ok like that.
untethered VR is super interesting and so is "standalone" devices.. but using a phone for it? not exactly optimal
mobile VR would basically need a fan to work good and having a fan on your face would be strange
yeah advancements in mobile are gonna happen in the headset
good "mobile" VR you only get with somehow having the hardware on your back or as a belt or whatver
that makes sense
including localized tracking, processing, etc.
eye tracking
and at that point why not just make it a standalone device that doesn't have to compete for compute resources
computing devices are getting smaller, but battery tech and power utilization needs a giant leap in the next few years. Things just heat up too much!
thats something nintendo should do, build a nice standalone VR device with awesome games
Im hoping that nano battery tech will take off in my lifetime ๐
thing about batteries: the tech is THERE but manufacturing is the issue
Quick question: Can you rebake lighing for selected meshes in Unreal?
rather than the whole scene
well the problem is now that progress with CPUs/GPUs slowed down so crazy much, now would be the time we actually need something like a 10x increase in performance because of VR
there's this one company that is able to hit 2x energy density, and is going to be successful ONLY because they're producing parts to make their batteries with existing Li-Ion manufacturing techniques, just drop-in-drop-out materials
there was a good MIT Tech Review article about them
and yes Moore's law is dead, RIP
Well... li-ion battery tech has been improving energy density steadily over the last few decades.
unfortunately there's only so small that we can shrink the anode and cathode
It's doubling every 7 or so years.
the tech is about the surface area that the lithium electrode contacts with the polymer substrate
and the size of Li atoms is not decreasing
well battery tech is "just fine" for most stuff now, I worry much more about CPU/GPUs
Volta is promising, as NV always is, but after that... IDK
Yup. I'd rather have proper VR "at home" than worrying about mobile
Eh. We're just going to have to accept transistor stagnation for a while.
they need to make graphen CPUs work
We're going to get sidegrades for the next decade
until we move onto graphene.
Whenever you decide to get your butt here graphene!
Starting to sound like the fusion of materials.
yeah, like the phone industry lol, "die size is better as it gets bigger!"
already seeing that trend
I'd be happy enough if they can do multi GPU properly ๐ but that is as much a software problem, but needs better buses too
multi gpu isn't the solution
On the flipside graphene research is getting super funded
while fusion research is... really not.
the industry motivating force is computer architecture, not nanomaterials
Confusingly and terrifyingly.
a single high end GPU will basically always stay at ~300W I think
how can we optimize all of this spaghetti VLSI
it's coming down to layout now too
maybe we'll see 'EZ' 3-d ASIC and silicon fab
what about 3D designs? HBM is quite promising
imagine a CPU that is a cube, not a thin wafer
it still eats more watts
that's pretty much what I just said, so yeah
the watts are the problem, not the size
Some CPU dude that I talked to a while ago seemed to think that the possibility of 3D CPUs was been overstated.
at 3d-manufacturing, density of transistors can decrease, which will decrease power usage
Science/tech journalism and all that
why would reduction of density mean decrease of power usage?
how do you transfer heat from within though? heat pipe mesh lattice?
less ambient temperature enclosure, less electron-tunneling etc. but then the whole thing is insulated so it's gonna get really hot
reduction of density in 3-d
So John what you are saying is we need to drastically reduce watt usage to be able to scale futher?
but wouldn't it be strange if silicon would really be the best material to use for semiconductors? they use it for such a long time now and usuall the thing thats used 40 years ago should long be replaced by something better
more resistance nearly correlates with more power, etc.
@heady parrot yes
often the reason to use something is cost more than anything ๐
increasing die size is easy, but you can't cool a GPU with 2000W even if it would be cheap to manufacture
hasnt AMD been doing a lot of work lately to reduce wattage?
they're crazy behind though