Search Unity

Feedback I'd like this GI solution in Unity, thanks a lot :)

Discussion in 'General Discussion' started by hippocoder, Apr 6, 2019.

?

Would you like this?

  1. Yes

    94.9%
  2. Yes

    68.6%
  3. Yes

    69.5%
Multiple votes are allowed.
  1. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Yeah, indirect is one thing, but we are fully baked :D Sadly I have cleared the GI cache so I cant show you the Systems view for this level but its all over the place.

    upload_2019-4-10_16-41-33.png
     
  2. elbows

    elbows

    Joined:
    Nov 28, 2009
    Posts:
    2,231
    After years of mostly completely avoiding realtime GI, apart from the original pyramid demo scene and a brief flirtation with SEGI I have to explore this tuff more in the coming months. Can you point me in the direction of more info/examples of that problem, or any discussion about it elsewhere, or any other issues with it and HDRP? Example of problems that show up with the FPS sample would be handy.
     
    Last edited: Apr 10, 2019
    hippocoder likes this.
  3. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    And then there is the resolution issue with baked shadows, I mean 40 whocjh is the default is bascily the max for larger scenes and just look on how contact shadows looks.

    upload_2019-4-10_16-48-14.png

    At 200 it starts to look ok, but still some issues here

    upload_2019-4-10_16-49-53.png
     
    hippocoder likes this.
  4. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    https://forum.unity.com/threads/tho...le-project-feedback-on-content-design.574549/

    There are a lot of learning and blog materials how to get the best out of it. Expect to have to split geo, to prepare unique GI params (only relevant once you grow larger than sponza) and then babysit it forever. Finally you may fall upon your sword with lightprobe leaking. All things that I hope will be addressed or replaced.

    If however you're working in small scenes, it's plug and play fine... with fat walls if using probes.

    Here's an example of just how broken lightprobes are right now and even in this case, they will leak:



    This is for the small Unity FPS sample level.

    My own probe automation is far superior to this effort with far less probes but of course I wouldn't have needed to do this if they simply worked.

    The strategy of extreme quantity is not what all other games do so it's clearly a Unity feature that isn't working or scaling well.

    And probes aren't even that important to me. I just want decent lighting that's visible from inside and outside in realtime. I don't even demand lush resolution.

    Just something that works that's scalable, that doesn't need a renderfarm. I'm sure it's possible, I don't even have a quarter of the geometric complexity as the average dice level... for my whole game.

    Anyway I've been gently scaling up awareness of this, and I guess a lot of Unity's customers don't understand or know the depth of work they'd need to do in order to make their scenes look nice with realtime lighting. They just give up in the end and the result is that indie titles tend to look mysteriously S***e, and you get threads by them complaining that "engine x has nicer lighting" and I can't blame them even though I'm well aware of how to make "nicer lighting" in Unity.

    It's just not working with scale or fully working as intended (probes). I don't even care about streaming. It craps out long before streaming is actually needed (well inside FP limitations).

    I'd waited years for a fix and it hasn't come. I've waited since last year for Unity to be more forthcoming, and I've always gently reminded them about it, but I'm beginning to think "maybe I'm the problem, I should just accept indies should make beautiful tiny things or ugly big things".

    Cos I'm not inexperienced and I'm far from dumb and I've hit a wall, and that wall isn't lit correctly.
     
  5. elbows

    elbows

    Joined:
    Nov 28, 2009
    Posts:
    2,231
    Cheers for the info. lol, many more probes than that and it would start to resemble a point cloud of the level.
     
    hippocoder likes this.
  6. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    Haha! I hid the geo on the FPS sample project by Unity and simply showed the probes then took a snapshot.

    This is a problem Unity knows about for the longest time, since 2011 actually.

    Light probes in unity wouldn't even be needed with the OP GDC paper anyway (you do not need probes or even AO) but I'm open to using whatever works.
     
  7. xVergilx

    xVergilx

    Joined:
    Dec 22, 2014
    Posts:
    1,876
    I got the same problem with baking terrains (~2kmx2km, its a bit less actually, but still).
    Its absolutely impossible for one man projects. I've tried lots of properties and the end results are just S***.

    Plain utter S***.

    Neither Enlighten or Progressive work with it properly, both produce not even close results to those that realtime have.

    Now I really think that lightmapping and GI is way beyond hobbyist level, unless your map is literally a box of 2x2m, or an indoor scene.
    Because it's not the complexity of the thing, but the actual time budget that gets sinked by it.


    So, if anyone asks me, why my game won't run on potato - That's because of realtime lighting / shadows.
    GI, well, I hope there will be some kind of real realtime GI.

    Edit:
    There's also an issue with large spotlights on the medium / low-end GPU's. Large spotlights aren't even rendered, and I haven't investigated why. Idk, maybe it's a feature.

    Was fun to present a game on a medium PC just to realize there's no light in the scene at all. In a night scene.

    (and no, it's not culled, because on my 1050TI same build it has lighting, and the quality settings are the same)
     
    Last edited: Apr 10, 2019
  8. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    If light probes died tomorrow I would be a happy chap, hate them :) similar feelings to reflection probes, they work if your world is Minecraft square

    Edit: I heard you can rotate them if you can move away from built in
     
    hippocoder likes this.
  9. Player7

    Player7

    Joined:
    Oct 21, 2015
    Posts:
    1,383
    Can't stand baked lighting aside from the ridiculous setup and compile times, the entire thing just falls apart the moment you want dynamic lights and shadows in the scene, and forget day/night cycles... way too static like the old height map terrain systems that we still see games being made with..

    Voxel based terrain and Voxel based GI lighting please... needs to start officially somewhere, out with all this old style tech of baked lighting and height map terrain its so 90's

    7dtd like voxel terrain engine with realtime GI.. officially supported in Unity awesome.
     
    Last edited: Apr 10, 2019
    xVergilx likes this.
  10. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    This is what I meant with systems, not easy to batch this mess

    upload_2019-4-10_21-26-1.png

    And if we get down close its not better
    upload_2019-4-10_21-26-38.png
     
    hippocoder likes this.
  11. Adam-Bailey

    Adam-Bailey

    Joined:
    Feb 17, 2015
    Posts:
    229
    Good reasoning.

    There was a recent Twitter thread where a Unity employee was genuinely baffled why some people wish Unity made larger games themselves with their tech, and I feel this issue is one great example.

    The current systems work great for smaller demo environments, but when you try to scale up things become a lot more difficult. Every time a new batch of Unite videos go up I eagerly watch any lighting ones in the hope that a development team talks about their techniques for lighting on a larger scale using the built in tools, but they are always "the built in tools wouldn't work so we rolled our own unique solution".

    Stuff like the FPS project is a good start, hopefully they are working on building a larger level for that project to realise where the tools need extra work.
     
    daisySa, dadude123, drcfrx and 7 others like this.
  12. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    They need to make a big generic rpg, not a quirky fps. Big generic rpg™ can be boring as long it has all the system (and the pain of setting them), Big generic rpg™ is a great showcase because it has all the system needed for many recurrent format of game. Unity is making a push in "cinematic" rendering, well that's great Big generic rpg™ have many tier of cinematic, from the great performance capture pron that will showcase rendering like the heretics, except not hacked together for a presentation, but as a part of a solid and robust import workflow because scale change everything and hack don't fly anymore. Big generic rpg™ will also be a showcase of the timeline editor for lesser cinematic, and put to great use the future procedural cinematic director, which is now an industry standard. Big generic rpg™ also have huge world that need to stream everything neatly in the background, to get that seamless experience, and also a great packing system to store, debug and organize all those data at scale. Megacity is a start, but that's still a big hack, and you are confine to a single biomes, there isn't much system interacting with each other, no cinematic with choice where you need to freeze background enemy without interrupting background ambient animation. Big generic rpg™ will also a big showcase of lightbaking at scale, make it efficient to endure iteration during dev and allow to detect pain point, such as data size, streaming, and ease of use, don't forget Big generic rpg™ need a TOD, great way to show the integration of shadow system and various lighting like a probable GI solution like @hippocoder wish. Big generic rpg™ is the future, vote Big generic rpg™.
     
    xVergilx and Ruberta like this.
  13. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    Big can mean the environment has to be big, but the gameplay is small and content is small. Presuming indies should somehow not make games involving flight or somehow not make survival games (which are pretty popular among indies) is a (big) mistake.

    Also going bigger than Sponza isn't big... Enlighten starts to break down out on an average sized level, and needs excessive time spent making sure it doesn't. So doesn't even need to be that big.
     
    Ryiah, angrypenguin, Martin_H and 3 others like this.
  14. angrypenguin

    angrypenguin

    Joined:
    Dec 29, 2011
    Posts:
    11,983
    For this they don't event need to make a "big generic RPG". All they have to do is make an environment for one.

    Then, drop a 3rd person character in there, run around as a player, and think "from a developer perspective, what would we have to do to bring this up to scratch, and does Unity support tools/hooks to do that?"

    Then fly around in a plane and do the same. Then drop a car in it and do the same. And a first person character controller.

    I don't think they need to build whole games. I do think that some now-common use cases could use some attention.

    For what it's worth, this isn't a complaint. They can't do everything all at once. I think it's cool that the tools have advanced to the point where stuff like this has become a "common" issue.
     
    dadude123, Ryiah, Martin_H and 4 others like this.
  15. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Also this is not only a Lghting issue. For open worlds they also need to revisit the stone age old terrain system. Native streaming etc
     
    Martin_H likes this.
  16. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    As someone who does just that around, and worked on at least one big game with some people, there is definitely a difference.

    In fact a lot of people don't understand how to scale from a single scene they made that look like a game to a complete game. Youtube is full of half game made in 1h where character walk big expense of non descript landscape, that's super easy to do.

    In order to meet gamedev problem, you need to have a game dev. Scale has problem of its own, in fact anoter way to see it is in AAA post mortem, where scaling from vertical slice to full game is a problem in itself.

    That said, they don't have to make a fun game, they don't have to make a game that fit a market, and they don't to make an original game. That will remove most of the risk inherent to game dev, while keeping the making part okay. But also cut dev time by 10x. That's why cloning games take inordinately less time that making the game.

    Again it's just a suggestion, not a complaint either.


    EDIT: Also it's a good stress test in INTEGRATING all their tools into a single project and check if they synergized together.

    I mean lighting can reveal issue when used in close up cinamtic, or mayeb loading light specific to dialogue pose problem on top of the global lighting, etc ... there is so many small problem that pop up because real application will bring edge cases.
     
    Last edited: Apr 12, 2019
  17. OCASM

    OCASM

    Joined:
    Jan 12, 2011
    Posts:
    240
    Have you guys considered Image Based Lighting? It's what CryEngine used for Crysis 2 and Crysis 3. All you need is cubemaps and unlike CryEngine, Unity can update them at runtime.
     
    hippocoder likes this.
  18. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    IBL is the basis of all modern lighting now (read cubemap)
     
    Adam-Bailey likes this.
  19. OCASM

    OCASM

    Joined:
    Jan 12, 2011
    Posts:
    240
    It's basically probes that you can use for both diffuse and specular. Unity supports clipping volumes so no leaking.
     
    hippocoder likes this.
  20. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    It's expensive to update, it has limitation as it doesn't reflect the geometry correctly. Else it would standard by now, it's known for a long time.

    Though I have theorize a GI solution based on a variation of that, I just need time to test it and see where it actually fail lol.
     
    hippocoder likes this.
  21. OCASM

    OCASM

    Joined:
    Jan 12, 2011
    Posts:
    240
    Every technique has its trade-offs. That's why there's no standard realtime GI technique.

    Also it's not necessary to update at runtime. It's optional. You can bake the probes at certain times of the day and interpolate between them as needed. It's the technique used in MGSV (supports dynamic weather too) and that was on last-gen consoles:

     
    Adam-Bailey and hippocoder like this.
  22. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    That's baking not real time GI, that's what unity does out of the box, the starting point of the thread is beyond that. Also you have a huge overhead to manage to get that, as raised by @hippocoder.

    The premise of the thread is hassle free GI. The solution of that paper is like less than 2ms on midrange gpu, no set up, limited artefact.
     
    Gekigengar, xVergilx and hippocoder like this.
  23. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    Seems like engineering I'm not qualified to be working with :(

    I'll happily try this - HDRP has a "dynamic" probe option that covers everything outside so I'd just need to work out how to darken the inside to compensate... so far no joy.

    ---

    Re: re-rendering probes...

    Most of these techniques re-light existing probes at very low cost. A feature I asked for in Unity more than once, including explaining to Unity staff when they asked me for specifics a year ago.

    As it stands, Unity's probes need to be fully re-rendered. This is not something AAA likes to do, based on far cry and other tech breakdowns. They just re-light what is essentially a deferred probe, it's a tiny cost because they don't need to render anything. This makes lighting updates very fast but is not available for Unity by default so I'd have to roll my own probe system entirely.

    This is a common song at Unity events and shows "Unity doesn't natively do it but we threw a staff member at it since we're indies with big pockets live on stage so technically Unity can do it... but not for poor people, so there."

    --

    Though really if Enlighten precomputed was actually precomputed within a time frame that didn't require NASA level processing and didn't poop on the CPU whenever the sky changed and actually had lightprobes that didn't leak like the Titanic, I'd be so in love with it. As it stands, no :(

    There's nothing truly wrong with Enlighten in general, but the Unity side of it is an unacceptable situation in my situation (phew!)
     
    dadude123, xVergilx and Martin_H like this.
  24. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Also dont negelect the importance of Final gather that I have.


    Final gather
    Untitled.jpg

    Indirect only
    Untitled.jpg

    No indirect
    upload_2019-4-12_17-58-44.png
     
  25. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    BTW my theorical solution is basically a varient of that, except I use the cubemap as a query structure (box projected scene UVs), and the lightmap is the gbuffer, once the lightmap has compute the direct lighting, I use the cubemap to query for each fragment the indirect bounces (ie box projected adresses of the points above the fragment, can be spread over frame), then store it into an accumulation buffer that spread bounces onto multiple frame. That accumulation buffer is also a lightmap and is used directly by the geometry as a regular lightmap.

    However it's a crude approximation that have the same problem than box projected lightmap, and probably light leaking. The second paper (non real time god of war GI) shared above seems to have some neat idea to fix some leaking artefact, in theory.

    The neat thing is that it doesn't involve mesh at all, could be process from texture alone since lightmap texture is basically an abstraction of the scene, and once process you don't need to run it again because it's essentially "baked". Which mean it also work with dynamic objects (no contribution) as they can just read the accumulation lightmap by querying the "cubemap adress". You can also have dynamic scene modification by re rendering cubemap's adress when appropriate.

    It was designed to work with low end open gl es 2.0 gpu (mali 400) assuming I can actually read and write to a texture buffer without having to do rendertexture in screen space (which is why it's a theory). It has some limitation like all the scene must be within a single lightmap, on open gl es 2.0 it WOULD also suffer from precision issues that limit size of lightmap potentially.

    I have been thinking about way to get dynamic object contributing (pointless due to the rate of update on low end machine) by rendering dynamic onto a cubemap and then merging the result with static cubemap, that would mean storing extra data (depth) and you will have artefact if dynamic intersect the cubemap FOV ranges. In theory.

    There is other gotcha, but that's the best I could find for the cheap hw I have.
     
    hippocoder likes this.
  26. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Another aspect is also rendering pipeline, real-time light in unity today more or less requires deferred rendering. Something that's not very practical for VR. A real time solution that worked with forward rendering or a entire new pipeline would be awesome
     
    hippocoder likes this.
  27. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    HDRP is equally performant in most cases with forward+ due to the architecture. The forward+ renderer dramatically reduces overdraw (with depth prepass) and seems to handle more lights than I will need at similar performance to deferred. For the hundreds and beyond, it can be switched to deferred.

    It's also switchable at runtime.
     
    AndersMalmgren likes this.
  28. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Nice
     
  29. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    I love it, after a 2 hour bake, Unity F***s up and it looks like this

    upload_2019-4-12_21-23-22.png
     
    xVergilx likes this.
  30. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Even large companies like Nintendo have problems with reflection probes, look at the poor result on the garbage bags :)

    IMAG1316.jpg
     
  31. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    I'm pretty sure it's on purpose, nintendo isn't known to stick to realism and favor other type of design since forever. For example they generally don't use cubemap that reflect the environment, but that give the idea of "metallicness", also use most visual trick to give a toy/jewelry/fabric vibe, is emphasizing tactile shyniness than reality. They also tend to follow design rule like readability, so they use effect to make things pop up visually to direct the eyes to important things, while realism tend to blend things in their environement. Entirely different philosophy.
     
    angrypenguin likes this.
  32. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    I think for me it's just optimised so that it runs with room to spare on switch, so the cubemaps are probably not unique to every area.

    Not sure what it has to do with Unity's global illumination though.
     
    angrypenguin and xVergilx like this.
  33. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    That ugly glow they have around their garbage bags, its realy common with reflection probes. Like hre from our game

    upload_2019-4-13_23-23-58.png
     
  34. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    Yeah, but it would also had been applied to every objects too, character aren't affected, and I'm interpreting based on their old game, as I'm mostly playing nintendo game. Which game is this anyway?
     
  35. Murgilod

    Murgilod

    Joined:
    Nov 12, 2013
    Posts:
    5,756
    That glow around the garbage has nothing to do with reflection probes. Nintendo rim lights loads of S*** and have ever since Mario Galaxy. You can even tell that they're just using a standard cube map because the reflections aren't environmental at all.
     
  36. angrypenguin

    angrypenguin

    Joined:
    Dec 29, 2011
    Posts:
    11,983
    True. I wasn't thinking about them making "a single scene that look like a game" though. When I said making an environment for such a game I meant everything that's involved with that. What they don't need is content, or any of the non-technical stuff that's involved in making an actual game.

    Even there, though, you have a point. The non-technical stuff has an impact on the technical stuff, because workflows and data all need to support it.
     
  37. angrypenguin

    angrypenguin

    Joined:
    Dec 29, 2011
    Posts:
    11,983
    Sure, that's one possibility. To me it just looks "Nintendoey", though. They use that kind of effect in their imagery a fair bit. Example:
     

    Attached Files:

  38. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Thats exactly how it looks with probes. For example when you have a probe in a shadowy place but you got lit parts in the probe. Then you get that unnatural glow to it

    upload_2019-4-14_10-56-43.png
     
    Last edited: Apr 14, 2019
  39. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    Unity very much think its part of the lighting process
    upload_2019-4-14_11-3-21.png

    And with PBR it has a huge impact even on static items

    Here with a probe in the shadow.
    Untitled.jpg


    Here without
    Untitled.jpg

     
    xVergilx likes this.
  40. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    @AndersMalmgren We are saying it looks like it, but probably isn't because "nintendo", and also it's localised to a specific object not all object. It's probably not even a cubemap but good old spheremap (think matcap), I ask the name of the game because ALL nintendo game are ripped and you can check the asset on the internet "unofficially".

    Sure you get the same artefact, it's just that specific case isn't a good example, because they choose to do it that way.

    EDIT:
    let's get technical here, your effect isn't the same actually, nintendo has even lighting below the asset and on the side, despite the scene being block on the right. Your asset has globally light loosely coherent with the scene BUT the lighting seems floaty because cubemap are either box projected or can only represent far field lighting, ie they either have a loose concept of proximity (which affect direction) or none at all.

    In the second paper in the early post of this thread (the god of war one) they propose a solution to mitigate that a little.
     
    angrypenguin likes this.
  41. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
  42. SunnySunshine

    SunnySunshine

    Joined:
    May 18, 2009
    Posts:
    663
    Dunno if it was brought up already, but apparently the Heretic demo was lit using solely real time lights. I'm not sure if GI is implicated when they say "The Demo team is working on a probe-based lighting solution to solve use cases currently not covered", but it could be.



    https://unity3d.com/the-heretic
     
  43. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    Yeah not in 2019.1 so far though unless you mean probe de-ringing option... actually that looks like marketing bollocks - probably just means HDRP and dynamic outdoor probe or something.

    I am developing heavily in 2019.1 so I'm not sure about that marketing claim above... and for a film where you control the shots, nah. It is a film though and I want to see part 2 :p

    @robert is my grand sage and will lead me to an age of enlightenment.
     
  44. Ryiah

    Ryiah

    Joined:
    Oct 11, 2012
    Posts:
    14,302
    Two to three years in development behind the scenes. One to two years in experimental. We won't have completely gone to raytracing hardware on the desktop, but by the time they make anything worthwhile we'll likely have affordable raytracing hardware on the market from both AMD and NVIDIA. Oh, and Intel too, if they don't just make another Larrabee. :p
     
    hippocoder and AcidArrow like this.
  45. Lurking-Ninja

    Lurking-Ninja

    Joined:
    Jan 20, 2015
    Posts:
    4,187
    Well, Intel has hired half of the AMD/Radeon engineers and heads and now they got Tom Petersen, I think they play serious this time, but will see.
     
  46. neoshaman

    neoshaman

    Joined:
    Feb 11, 2011
    Posts:
    4,190
    I heard the nvidia guy left because his boss roasted him on stage once, is there any sources of the events?
     
  47. Ryiah

    Ryiah

    Joined:
    Oct 11, 2012
    Posts:
    14,302
    GTX 1080 Unveiling (starting at 37:40).

     
    GameDevCouple_I likes this.
  48. AndersMalmgren

    AndersMalmgren

    Joined:
    Aug 31, 2014
    Posts:
    4,684
    I love lightprobes. Like when they use probes from the other side of a floor

    Untitled.jpg

    upload_2019-4-16_19-19-49.png
     
    Last edited: Apr 17, 2019
  49. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Posts:
    25,356
    There are a number of previously undocumented GI features in todays 2019.1 release: https://blogs.unity3d.com/2019/04/16/introducing-unity-2019-1/

    1. they claim they have fixed the async readback for enlighten (sky etc) - haven't tested
    2. they have a hook in shadergraph so you can plug your own indirect stuff into the gi backend...

    So the backend thing is interesting cos it means by default they can plug different GI into all their shaders (not just shadergraph, obviously) so either they fix enlighten or they use it for raytraced gi or have a better overall approach...

    Lets see? Still no word from Unity's lighting team yet.
     
    Martin_H and neoshaman like this.
  50. AcidArrow

    AcidArrow

    Joined:
    May 20, 2010
    Posts:
    5,760
    Interesting. Is this something different than simply writing a shader meta pass?
     
    joshcamas and hippocoder like this.