Lux-an-open-source-physically-based-shading-framework

Lux supports a specular reflectivity workflow, but what is suggested to obtain the information for those reflections? I'm seeing things ranging from cubemaps, box mapped reflections, SSR, etc. Does anyone have advice on what's best to use? For myself I'm focusing work on interior spaces.

cheers

 

hi bcoyle,

each material will hold its own information about its reflectivity stored in the specular color and roughness texture.
Lux is physically based driven so you just setup your material and the rest – no matter if you use box projected cubemaps or sss – is driven by the shaders automatically.

lars

 

Hi guys. Have a look here...

 

hi there,
unfortunately that texture guide does not have much in common with Lux...

introduction to pbs for artists:
https://www.youtube.com/watch?v=LNwMJeWFr0U

content creation:
http://www.marmoset.co/toolbag/learn/pbr-practice

lars

 

Hello all, I started using Lux a couple of days ago and have been lurking over these threads since.

Thanks so much for developing it, the results I'm getting on my first test seem super predictable and exactly what I was looking for - I might post some screens here later.

I've had a couple of issues that others have mentioned here,the 'cant compile' error on build (related to the cubemapper I think,) and also that the box projectors didn't appear to be working. I'll try again soon with the GitHub version and see if that corrects it.

But yes, thanks again and I cant wait to see how it develops!

 

hi there,
updating to the latest version from the repo should fix all your issues.

lars

 

Oh, i see. Thanks for the links.

 

Mobile support?

 

hi there,

as already said: lux has not been developed especially for mobiles. but it works on modern devices which support shader model 3.
even if i did some more extra optimization for mobile this limit will exist due to the fact that shader model 3 is needed to support specular ambient lighting.

lars

 

That question was asked the previous page I think, should probs do an faq on the OP

Also, i feel crappy saying i may have to move to skyshop as their update is coming way way sooner than I thought it was and the update has stuff i'm going to find pretty vital, HOWEVER i do feel contrary at times so if i make anything pretty happen and it's doable and images of lux looking nice are still wanted (lack of response to me has had me feel a tad unwelcome), when i have something considerable built up, i can adjust it to lux for the occasion and post it here in image or video form. There isnt a huge difference between them in many ways, skyshop provides a fresnel slider for control, i think Lux sports a larger specular range? It's just right now the gpu convolution,box zones and shaderforge implementation (big plus if it can do what I hope it can) would be a practical necessity, but i hope lux grows as it has been and i'll get back to it properly when the competition project i'm attempting is over.

For anyone reading this and new to the subject - it is worth noting that Skyshop is a paid for a quite costly product and Lux is absolutely free, and you can get absolutely great results from it, and ive no doubt its feature set will become extremely impressive, or even more extremely impressive. Actually to prove a point i'll do something now

 

hi lazygunn,

i am sorry for not having answered some of your posts. if so i probably did not feel personally addressed or what i have to admit did not really get what you meant (sorry, i am not a native speaker).
so just let me simply assure that you always have been welcome and of course still are.

as far as skyshop is concerned: the difference between skyshop and lucx is a bit more than just the larger specular range. but without going into details i would claim lux to be "more" physically based.
gpu convolution is not right on my list as i would call it important only if you have a lot of specular ambient probes in your level. but of course speeding up things is always nice to have.
lux already supports an unlimited number of different "box zones" in your level.
but shader forge support will not come in the nearby future as joachim is focused on bringing unity 5’s rendering features to it.
and that is the real challenge: the future of all frameworks will depend on unity’s approach, how flexible and powerful it will be.

lars

 

Ahh yep, i wasnt fully sure on the differences so please dont take it as deliberate ignorance. For my plans with the competition the issue is exactly that - i need not only a lot of zones made, which i know Lux can do, but its a lot of area that has to fit onto a webplayer for the competition - each zones env map will have to be made as you go along, i can't pre-make many, webplayer size would get crazy, and the plan is to have a day/night cycle, so for zones visible to that, a new env map would have to be generated reasonably regularly and not too noticably. This is why the gpu convolution is important to me because anything involving a dynamically lit or changing environment requires an appropriate environment map - so far pbr/ibl is limited because it cannot usually do this - gpu convolution is the answer (for now)

Automation of changing an objects referenced env map is very helpful (the visualisation for the skyshop zones seem to be essentially box collider triggers), as i find box correction very helpful - it's a great approximation where I hope to use a screen space reflection image effect to 'fill in' the center of the image with more accurate but in many ways limited reflections - i see box corrected reflections coupled with screen space reflections as a good marriage

For right now, the shader forge implementation is needed purely because of another problem I have with getting everything into a webplayer - taking a huge amount of hand-made geometry and textures, maybe one or more gigabytes worth in its untouched state, and reducing it to fit into a sensible download time as its a competition - im sure the judges will not be impressed having to wait forever to download the game, so the plan is to decimate everything heavily, author the textures into a more stylised NPR look so they will look decent at much lower resolutions, and the decimated geometry look a lot better because of this

but! I would like to have tessellation available to dx11 users to rebuild some of the information on the decimated geometry or at least improve its appearance, and that means getting tessellation into the shader - if shader forge is implementing skyshop (and vice versa) and I can reconstruct the needed skyshop shaders properly in shader forge - then it also has tessellation available, and has the tools to give the option of some basic tessellation to, say, bevel the edges of an object with that material, or use a very simple displacement map to 'reconstruct' an object with that material, or use gpu generated perlin or fbm to generate geometry detail, and this would still look stylised and not realistic but would improve the fidelity I think

I'm not, not using Lux out of ignorance I assure you, i've been testing lux out in different situs for a week now and I really do like it a lot, but it's completely unreasonable of me to ask for feature inclusions like i've just described when i can do them myself in an alternative manner for the time being

I will be using Lux again, as I say i really like it and if it keeps evolving as it is, especially along with Unity 5, then it will become a fearsome and very strong package. I don't see Unity 5 as taking over any of the pbr shaders in usefulness, Lux least of all, because Unity's pbr will surely be good but it can't adapt to every need and cannot move quickly, the paid for pbr solutions have the same issue but to a slightly lesser degree - Lux does not have this problem whatsoever as the way you've developed it and given it exposure means it can stay extremely agile, be developed and updated extremely quickly, and with more people coming to support it, one of the real assets of Unity

Sorry for the rant, i just wish to explain myself, and i'm currently looking for a suitable scene to adapt to Lux to hopefully give a good demonstration of what Lux can do and what it looks like when you're actually playing a game with such a system in place. I should probably be working more obviously for the competition but I think i'll still be learning good things anyways (I'll take the opportunity to use a few other things i should be learning too) and hopefully it'll be a nice promotional thing for Lux. It's mainly a matter of finding what uses the broadest range of Lux shaders to be fully representative

 

hi lazygunn,

so i hope skyshop will allow you to create and convolve cubemaps at runtime. has this been announced as regular feature? that would be a great surprise.

and i do not know how shader forge will allow you to reduce the size of the webplayer. lux shaders are pretty huge as they come as uebershader.
but most of those uebershadert feature you will not need: linear AND gamma? blinn-phong AND cook-torrance? so you can simply strip that off using defines instead of the multi_compile pragma and make the shaders much, much smaller (and compile much faster btw).

tessellation is a dx11 feature only requesting one more texture at least. but it should not be that problem adding this feature to any lux shader.

nevertheless, skyshop textures are compatible with lux which should be compatible with unity 5 (especially if you use cook torrance).
so you won’t waste time working with one framework and switching to another.

lars

 

I think skyshop can make cubemaps, but its a Pro only feature because of the way it's handled, as is using lightprobes. Though it seems like they are updating it to work with Free. Though I imagine that none of that is during runtime.

 

that is what i think too. convolving cubemaps at runtime is – pretty challenging but probably not impossible.

 

You can see env maps being generated at runtime in their gdc demo, there's a slight pause between them entering a new area and the blend between the maps, whereas when objects go into an area that has already been obviously given a map the blend is not that noticable at all. I don't really see why processing the maps at runtime should be a problem either if it's so fast and can take best use of gpu parallelisation (As in everything else it's doing doesnt grind to a halt)

Shaderforge has tessellation and its displacement as one of the main inputs of a shader in its graph, as I explained i needed tessellation, or well, i wouldnt die without it but having the option there is almost vital to how my 'vision' of the project works. It should be compatible with dx9 and have fallbacks for each dx11 feature i use but i'll be aiming for an appearance that at its most accurate uses quite a few dx11 features. Shader forge implementing skyshop is currently the only way I can think of (I am not a good programmer and my shader writing skills are very lacking) to conveniently get tessellation implemented into a pbr system I recognise.

If you are genuinely willing to make me a set of Lux shaders with the features i request (And yes increasing the speed of compilation would be extremely handy, i can get very specific regarding what I need) or have specified, then i'll use Lux, simple as that - As I could continue to use Skyshop to create my maps and use Lux for the shading as I have become familiar with it. I would never have thought to ask you to do that because it's a lot to ask someone out of the blue but i'll gladly use Lux, and see no reason why not, if I can get the features of skyshop i was needing running just fine with Lux.

If you were willing to do that then i'll be happy to go with that, but don't be surprised if i send you completely mad aha

 

And yes I was under the assumption it was realtime or.. why make a big deal of it i suppose - particularly if it was via dx11 and a compute shader or something similar - the method i'm using to 'treat' the texture maps is implemented in CUDA and it can process hd video in realtime and I don't see, unless i'm being extremely ignorant, how the convolution process could be a great deal heavier than that, because that itself is a very intense process

I'm fairly sure ive read white papers regarding gpu convolution in the last month or so - That could possibly be another thing to look into. There's a general agreement on these forums that devs should stick to directx9, and I understand that because they want to make money with what they produce. I however do not and havent owned a 3D card that didnt support dx11 pretty much since dx11 became a 'thing'. I think the market for dx11/cuda/opencl has opened up enormously even thinking back a year ago, and with the new gen of consoles now here, hanging onto dx9 isa strange and archaic thing, after seeing what gpu computing is capable of. Maybe its missed because calling them 'gpus' is pretty misleading

 

hi lazygunn,

env maps being generated at runtime: i do not know... but we both are just guessing i think.

tessellation: i might have a look into this.

customized set of lux shaders: you should be able to do it on your own according to the documentation.
but probably i will not have any problem to do that for you. nevertheless my personal experience is: you do not need that much customized shaders but have to know how to use the given ones properly.

but back on track: i guess it is time to get know more about your project. so post some infos here or send me a pm.

lars

 

I sent a lengthy PM (Although there's probs loads more to say) to avoid clogging up the thread but if you're interested in what I proposed it could maybe benefit Lux a bit, and me a bit, which works for me

 

I modified the Lux Env Probe script to make it a little bit easier to use for my work. it now, automatically assign the meshs inside the Probe's bounding Box, and generate materials instances for every room at runtime. making the box projection workflow a lot easier. (you'll still have to bake the cubemaps in the editor)

you can grab the scripts from here : https://googledrive.com/host/0B3n1sS2Im_HhZ3kwRU0yNTVncEU


And here is a video showcasing Lux BoxProjection in action, the environment is still early WIP though.

 

Hey cheers for this! Perfect! This was another thing i was dreading having to rely on my shonky programming skills implementing, and something i'd need to hand really soon, when i have to swap between programming and art creation it seems to take 3 days to do something that should have been far quicker, you've saved me a ton of hassle so thankyou

As i'm mentally defunct as messing with my desktop goes but i'll try do something useful and go on the great gpu convolution hunt

 

Waah, that´s just so cool it hurts! Have to give those scripts a testdrive and see how they fit into our workflow.

 

Here's another shot I made:


This tool is absolutely amazing.

 

holy S***, that looks so damn real!
incredible.

i want a webplayer!
;-)

lars

 

nice!
but where do the reflections on the ground come from?
pointlights you have placed above the ground to simulate the spots?
i guess you might need real spot lights placed below the ground...

lars

 

thanks to Lars who made this awesome shaders available under open source license, expect the community to build a lot of nice features on top of it.

 

Will deliver, soon

 

@Botanika: this is simply amazing! I downloaded the latest version of Lux the other day, and this script should do nicely for my cubemapping needs! wouldn't mind a webplayer to test out! and your video could be mistaken for unreal engine, IMO!

@Lars: how are the foliage shaders coming along? and with the SSS shader tex: is the texture a combination of roughness, SSS, and spec across colour channels? i am not quite sure how to combine them!

I was also considering Lux to give a graphical boost to a project i am working on, so if it goes well, i will post screenies! and BTW, i reckon this framework could rival not only Unity 5, but Unreal Engine! (You may have seen evidence of this already by dnnkeeper!)

BTW, i am thinking of doing a new water shader that has better fresnel, and may make a lux version for you guys. Once i have something, i will post results here.

EDIT: the hair shader won't compile: it gives me "undefined variable: LUX_CAMERADISTANCE at line 29" Please help!

@Squared55: Ooh... A DARK SCI-FI CORRIDOR!!! MWAHAHAHA!

Anyway... Great shot. Also wondering how you got the light spots on the floor...

 

Don't think I've ever seen Unity look this good. You guys are doing great work! Love the LUX Box Projection, UMA asset with the skin shader, Picture above (that dark corridor), and that UE4 scene.

 

I'm... not really sure how I got the light spots. I'm using Blinn-Phong, and my metalness map has no transparency. I just need to put the lights close to the ground... but I can't for the life of me replicate it exactly elsewhere, for whatever reason.

 

Looks great!

Unfortunately, I can't figure out how to work it... clicking Find Closest Objects doesn't seem to do anything (even when the object is scaled up to encapsulate the entire environment), and nothing changes at runtime. Any ideas?

 

so those white spots are the direct specular reflections from the point lights.
but instead of using point lights (which do not represent your kind of light source anyway) you should use the "correct" lights which would be spot lights:
just place them a bit below the ground, uncheck "cast shadow" on the ground and the lights geometry (the white cylinders (using the self illuminated shader)) and everything should look fine:



lars

 

I figured that the wide radius combined with the low range and high reflection would be best approximated by a point light (Plus, I like the way the spots look). I'll remember that for next time, though.

 

- the foliage shaders are on a good way: most features already work. now it is about testing, testing, testing – and updating the docs.

- sss: Thickness texture: this texture only holds thickness in the green channel. all other channels are unused.

- hair shader: ups, i have forgotten to update the hair shader to make it work with the latest version of the custom fog include, sorry.
but i have just updated the repo.

lars

 

@FuzzyQuills Thanks man, I'm working on a webplayer.

In order for the probes to automatically find encapsulated assets, the meshs materials have to have a Lux/Box Projected Shaders.

if for some reason , you wish to use it with standard Lux shaders, you'll have to comment the line 284 and 288 in LuxEnvProbe.cs

 

Err.... I REALLY should have been able to figure that out myself. Thanks.

 

@Lars: Thank you sir! I may need the foliage shaders later for another project, but for now, i will stick with weaving shader magic with lux bump spec... Also, just an idea: would it be possible to do a sphere projection shader like the box projected one?

@squared55: yea, i suggest you don't use the box projection shader with a sphere or another shader though, sine it came out really weird on my computer... better to use box-projection shaders with cubes!

@Botanika: thanks for info and you are welcome!

 

I find box correction quite useful on a variety of shapes so long as you're not wanting very polished surfaces, they're nice approximate reflections for rough surfaces that work well with screen space reflection (so far, haven't tried them with Lux yet although since it uses specular and gloss the 'regular' way, candela's ssr effect should work. I should test that when i stop arsing about on the internet

It actually occurred to me the other day i was trying to approximate box correction and glossy surfaces (using amd's cubemapgen) ages ago, a year and a half ago in fact, according to my youtube video, so i'm maybe more familiar with box correction than many, which is possibly why my mind made the connection with them and ssr quite early after realising the ssr effect would just look 'wrong' (it does really look wrong on its own, a bit like too-sharp corrected cubemap relflections look wrong on their own, even in rectangular spaces) without something filling in at the edges of the camera view

 

hmm, i am not quite sure what you would need this for.
so can you please give me an example and go a bit more into details?

lars

 

I can see the logic behind 'hey spheres might fill spaces better than boxes' but unless you're in.. i dont know, a big dome, the distortion would make most regular rooms look completely bizarre, capturing sphere mapped HDRI is old news, applying perspective correction on it i dont think would have the imagined effect. I was just looking at filling an interior with large polyhedra, since they interlink so well, so get more information out of a scene, but that would be quite a leap from a cube (which can be a product of a polyhedra) to a possibly randomly grouped set.. buh i should sleep and stop talking crap

 

Did you mean if you wanted to use it with standard shaders? Will this work with non-Lux shaders? I'm looking for a reflection system that is separate from my shaders, where the game objects will pick up certain box-mapped cubemaps based on their location. Is this possible?

 

Another problem: Lux shaders add these weird bars to textures. This ruins both the diffuse IBL maps, as well as the box projected shaders:

 

i doubt that lux adds any stripes...
but i can imagine taht lux makes them much more visible ;-)

so those might be texture compression artifacts...
please try to use an uncompressed texture.
if you use linear please make sure that if a texture is not a regular diffuse map you have checked bypass srgb sampling.

lars

 

nope, boxprojection needs a shader that supports it...
picking up nearby probes however should not be that big deal (but that would not work with box projected probes of course (!) (how should that look like?: you project a box onto a box; that is all about box projection...)). use regular cubemaps on moving objects.
lars

 

You're right, It's a problem with my roughness texture. Can't seem to figure out why it's there, though... Oh, well, I'll keep looking.

Do you have any tips on getting box projection to look right? I can't seem to make it work properly. I always end up with what should be the floor reflections on the roof, and what should be the roof reflections on the floor.

 

I'll get a chance to test this is in a sec, but the box correction has no hidden magicks to it, the box projection relates to the dimensions of the area the probe took the skymap at. Give the box correction inputs those values, and i also assume you can use the normal transform tools to tweak

Looking forwards to checking botanica's work, what a champ, i keep meaning to sleep but then i find other stuff to do! Have a scene in mind that i can put up super quick, Lars might appreciate the visualisation anyways, and i'll use the opportunity to bung ssr in there and turn it on and off a few times. This is the longest going to bed ever. Gotta get some form of website back up first though, its embarassing seing the godaddy intern-art on my doman

 

get the latest version from the repo: i have completely rewritten the shader and now it should work like charm.

lars

 

Ahh haha more reasons to stay awake, stop doing this

Lars, out of general interest, are you doing requests or are you very busy with more fundamental aspects of Lux and your other work? I don't want to hassle you, so i'm not going to, im just interested for the future, theres particular materials i've been thinking of because I keep reading endless supplies of white papers, i'd rather leave you to what you have planned that's important really, but if you feel alright with suggestions theres some really interesting work in a few areas you might be able to implement into lux.

Although firstly i'll ask if you've read any of the gpu pro books or material like that. Given your dedication to your foliage shaders, i think theres some good stuff for nature lovers as well, especially coming up in a couple of weeks with gpu pro 5, although i may mention it to tom if you're not so interested as it's a grass rendering technique. Ive been looking forwards to this book since january. This isnt self motivated however and i dont want to bomb you with links to pdfs.

Haven't got much done today except thinking as i've been trying to get a new website up, i feel naked without a link to my stuff but kind of dirty when the link goes to godaddy

 

Hi,

Why not simply remove noambient

Code (csharp):

1.  
2. #pragma surface surf LuxDirect noambient
3.  

?

 

Is this what you are referring to? https://github.com/larsbertram69/Lux/commit/7fcc7024516162bb6fda35057cb27e9cb442d5bd. Is there another script I have to change too? Must be an issue with my set-up; I seem to be the only one having this problem.

Here's an image with box projection (same dark corridor as before; there's a flashlight used in this photo, though)


And the old one, with normal cubemaps for reference:


Also, the box reflections right themselves if I De-focus the Unity editor window, and the focus it again.

BTW, thanks for all your help here; I really appreciate it.