How to make AAA graphics in Unity/Enlighten

I love it.

It's hysterical that this looks great even though there's a random tree stump in the foreground and grass growing on tile.

The color is beyond excellent.

I gotta know, did you hack some of the materials? The ceiling and right wall have the same color, whereas the left wall doesn't. It makes sense because the light is coming from the right, so the right wall would be brighter and light would reflect onto the ceiling. Did you just use different material colors to hack the impression?

The colors you achieved there are really really impressive man. If you hacked it or not doesn't matter. The result is beautiful.

I would totally appreciate a breakdown.

This is simply not accurate. Post doesn't work like a photoshop filter, it has access to layers and layers of additional data. You have depth, light info, normals, positioning, distance from camera, etc. Slap AO on the camera, and look at how it responds to changes in lighting intensity the angle of lighting.

Again, stuff like adding color bleed to AO is going to have as much or more effect than some GI on how lighting feels in some scenes.

What's more, stuff like tonemapping and eye adaptation or color balance can have dramatic effect on our impression of light and dark. Just check out how much of an effect tonemapping has on light, shadow and falloff. Depending on how eye adaptation and tonemapping is configured, it can completely change how strong a light appears and how much falloff it appears to have.

Any kind of volumetric post effect (fog, scattering, etc) is going to add depth and change the impression of distance and how light interacts.

Post is really part of the lighting process at this point. It's not cheating or faking, many of these things are designed to work together.

 

The difference between the big picture of rocks is incredibly subtle. You really have to look super closely to notice the differences there. It's in extremely fine detail.

I opened both and tabbed between them and didn't notice the difference at first. You really have to look at the fine details closely.

Maybe not the best demonstration pictures, the differences can be more pronounced.

Mips and ansio really make a big difference in game play - since a lot of times you're viewing things from angles and distances that may be more extreme. Especially at grazing angles like floor or terrain.

 

I have to say, I don't think this is the problem, it's not something that a lot of games do. If you look at the pic of Battlefront I posted for instance (as well as just about any other game) the textures come off very slightly blurry but it doesn't prevent the overall lighting from being fantastic.

I must admit though, for still images there's a significant improvement in the fine detail.

 

Not subtle at all to me in my 24" BenQ display -__-
My eyes feel some discomfort in the Box one.

This is what meant 'flat' to me every time I used it, so if Billy still finds it flat, then I am joining the train of finding out what is it 'flat' means for Billy haha!

EDIT: I asked my brother and he too does not notice the difference even in my display even after pointing it out to him so it really is me.
But I can notice it immediately whenever its present to the point it is bothers my eyes!

So this is the last time I mention the word 'Kaiser'.
Don't want to known as the crazy Kaiser guy

 

Just a final note on post. People constantly mention SEGI as bringing next gen lighting to Unity.

I would put money on the idea that SEGI is mostly just an outgrowth and extenstion of SE's old ambient occlusion.



Old SE AO used to have significant color bleed, this created soft reflections and as seen here dramatically changed the color of a scene.

In general though, I would hazard a bet that most of the stuff people find most impressive about SEGI is really the color bleed which he just imported from the work he did on AO.

 

@Billy4184 You do realise that Unity by default does not use physically correct inverse squared falloff lighting!

Unity 2017.1 beta Release Notes:

 

I said WOW.
I always looked at those screenshots and did notice a difference but didn't pay a lot of attention because 'It is just a SSAO asset', but now that you point out at color bleed, I can see it!

WOW.

 

Might have to try that out!

 

This art looks great, and there is some kind of enhancement effect that is very nice. It's still the same lighting as always though.

 

Yes, the lighting is the same as always, but if you stack color bleed with the current post stack, you are guaranteed good results I bet.
Now why isn't there color bleed in the post stack...

SonicEther said he would make SESSAO free. Just bumped the thread.

 

I really think you may not know what "lighting" is... you may just be mesmerized by buzzwords.

Slap different falloff onto lights and this kind of color bleed and you basically have SEGI.

With command buffer access especially, post and lighting are literally the same thing.

Interesting article for technically minded folks: https://petapixel.com/2016/06/02/primer-inverse-square-law-light/

Eye adaptation can really be thought of as dynamic exposure control and white balancing since that's... well that's basically what it does.

Some info on tonemapping and HDR: https://en.wikipedia.org/wiki/Tone_mapping

For final reference, here is one of the promotional image comparisons from SEGI.



The difference here is almost entirely (or entirely) the color bleed (soft reflection of color).

 

That looks really good! I'm glad you joined the fun. Is this made in UE4 or Unity?

I was about to suggest that this must be a "you-problem", but it seems you already figured it out. In the on-off gif I can't even see the difference right away, I need to wait a bit till my eyes have focused enough to even see the difference. I'd never notice the difference while playing a game. I've overlooked far far greater changes in settings while actually playing a game.

I'm not sure how familiar you are with SEGI, but that sounds pretty ridiculous to me. SESSAO is a pure screenspace effect and SEGI voxelizes the 3D geometry of the scene and calculates light bounces in 3D space. The assets both want to achieve similar things (more realistic light behaviour), but technically that's a huge difference imho.

 

Heh, let me quickly skim it. I haven't opened it in like 10 years. Nah, it's basically entirely about how to do things in 3ds max, an old version of 3ds max at that. The only generic information in the book is learning how to approach recreating something and there's more than enough information in the preview I linked to on Google Books to learn about that. The book is about creating dozens of different effects, but the approach to each of them is the same. Just check the "analysis of effect" portions of the preview and you'll learn basically what you would from the book. Basically the book is trying to teach you that when you look at for example a candle flame you need to be able to describe it in more detail than: "it looks like a flame" or "it's golden and bright", for if not how would you be able to recreate it? Trying to create something that "looks like a flame" without knowing what makes a flame look like a flame is futile and something that's just "golden and bright" is certainly not going to look like a flame.

I'm still not sure what the goal of this endeavor is, but I've been wanting to recreate a real life photo in Unity ever since I saw posts like these on Reddit:
https://www.reddit.com/r/Unity3D/comments/69sdid/i_recreated_a_cool_image_i_found_to_see_how/
https://www.reddit.com/r/Unity3D/comments/63k1n9/finally_its_moving/

I used to think this was a really fun exercise in 3ds Max, but I've never really cared that much about graphics after I switched to game development and Unity. Gameplay first as they say, but something like this at least seems somewhat related to this thread, or at least my understanding of the thread.

And I would actually advice you as well, Billy, to try and match an image pixel for pixel rather than to create something different with the same theme. At least that was what I used to do in 3ds max, i.e. try to make an image that you could put side by side with the original and people would struggle to tell them apart. It will also make it a lot easier for you to compare the two in for example Photoshop as you're working and see what you're doing that's different from the source image and try to iterate based on that.

My workflow was usually to have the source image open in Photoshop, make an iteration in 3ds max, render it out and put in on a new layer in Photoshop, compare the two, iterate, compare again, rinse and repeat until I was happy.

Also, you should perhaps start with something less complicated than a fully fledged AAA game scene though. Something like this:


Or something even simpler, something that just requires you to create a simple mesh, a material, a skybox and then fiddle with lighting and post processing until you're happy. Basically a lighting study. If you manage to make a simple scene look good there's no reason adding more assets to it should make it look worse, unless they're horrible assets obviously.

 

I followed the progress so I can answer that. His SSAO bleeding solution DID inspire SEGI but SEGI is entirely different in implementation. He did a two bounces evolution from the AO then switch to voxelization of scene which is an entire new technique.

At the end there is no relation with the two in implementation.

 

I really think you have no idea what I'm talking about. Look at that image on the left in the picture you posted, take a look at the overall lighting and tell me if it looks realistic.

Now take a look at an image like this (hopefully one of the very few I post in this thread that aren't of my work):

Spoiler

Do you even think they're in the same ballpark in terms of fidelity and realism?

A bit of color enhancement and stylization is not what I'm referring to at all when I talk about deficiencies in lighting. I think we're on completely different pages.

 

Ha, if you don't think grass grows on tiles you should see the miracle that is my driveway.. I was hoping someone would catch the tree stump though hehe!. I will do a proper breakdown of the lot soon, I'm just testing some other stuff out and want to see if I can actually find the route cause of all o' this..

 

I think you're probably right, I'm going to have a think about this.

 

But is your opinion influenced by the assets in the scene or are you only talking about the lighting? I cant say I think the lighting is especially incredible here. The lighting in the image you're comparing it to is obviously not great, but much less effort has been put into that by someone much less skillful. There's no doubt in my mind I would be able to recreate this image in Unity, but I can't stand making assets so I would rather not.

There's obviously some volumetric stuff happening in this image so if that's important to you then you'll need an asset from the asset store to achieve the same in Unity. Or maybe some of the free assets on github are enough, but you don't have volumetric lighting in stock Unity.

 

What I'm trying to explain is that its not "color enhancement and stylization" there are more elements at play than this.

I think you can't tell the difference.

The picture you included depends immensely on volumetric fog, tonemapping, and bloom. Although it's possible that this is using bloom or some other effect to achieve the appearance of volumetric fog. One of the main qualities of this image is that the bloom from the windows creates the appearance of haze.

There are also specific hazing solutions, some very sophisticated to achieve this kind of effect, all of these things are attempts to mimic how our eye filters brightness and how particles in the air affect light, which is immensely important for telling us how far something is, which affects depth perception.

I'm not sure what effect its using, but some bloom/fog/hazing process is a huge part of that scene. In the end I think it may just be a boom hack.

Atmospheric scattering is another technique to achieve the same kind of effect. All of these are trying to give volume to space and change how light reflects. Even super simple effects like global fog have the same goal, just a more simple/naive implementation.

 

For anyone who wants to see what a game engine can do "without any additional plugins", I encourage you to take a look here.

In the meantime, I think there's a lot to do and learn.

 

@Billy4184, I think you may want to dig into shaders a little and maybe break down some simple post to understand some of the data. I think you may really have the wrong impression about what post is, or how shaders and light work in 3d engines.

I won't keep harping on the subject, but really, you may be under incorrect impressions about how these things work together.

 

That wasn't about mipmaps, that was the 100% texture resolution that looked entirely different because of the different downsampling approaches (I might have added an additional sharpen filter after downsampling, can't remember).

 

Look, I'm not under some kind of illusion that I know much at all about post effects, GI settings, and lighting techniques in general. But I can certainly tell when something looks realistic or not.

So I don't know what impression it is that I have that you think is incorrect, because I don't really have any when it comes to how this stuff works. I'm just starting to figure out some basic stuff. I don't understand what's the point of discussing post effects in general terms without examples that show how it gives the desired effect - and IMO the example you posted is far from realistic.

So I'm just going to keep chewing on this problem, improving my skills and getting a bit closer to where I want to go. I enjoy talking about stuff but in the end, what counts is whether or not we end up with the result to prove it.

 

Maybe the GTA 5 - Graphics Study provides some insights how they achieved that?!

 

For the record, I'm not trying to criticize you so much as trying to explain how these things work together.

For example, what exactly is GI doing?

This is from the CryEngine docs: http://docs.cryengine.com/display/SDKDOC2/Voxel-Based+Global+Illumination

There are two main results:
- Better AO (subtle shadow on small areas)
- Color being reflected in indirect light.

Different implementations will produce different degrees of quality, sure, but the goals and results are very similar. The more that you can connect shadow and brightness to colors bleeding, the more realistic looking the lighting is.

The value here comes from understanding how the different parts work together and figuring out how you can use your tools to achieve some goal.

A main source of difficulty is when we don't understand how things work together, and we're stuck just tweeking values endlessly, in a cycle that leads to frustration and wasted time.

The more we understand how different parts fit together and the roles that different tech plays, the less we are reduced to guesswork and endless tweeking.

Trust me, I've spent enough time on endless back and forth tweek fests, where I can never get each of the elements working together properly without unbalancing others.

The way out of the "tweek hole" nightmare is to better understand what each element is doing and what it's goals are. Then you can really start logically deducing values and improving your rig.

 

I agree.

GI and color bleeding are separated features. If you disable color bleeding in SEGI it just wouldn't be the same.

The problem in here is that SEGI encourages you to tweak with the lighting because its results are instant. Meanwhile, baking Enlighten takes hours.

SEGI being realtime means that unlike baked lightmaps you do not need to compress the shadowing as an image with artifacts (unless you go for a huge lightmap resolution) which greatly benefit effects like color bleeding.

Haven't tried the progressive lightmapper. Is it much faster?

 

NO! really? Understanding stuff makes you better at it? :O

Understatement of the year guys, come on. You owe it to yourselves to look up what the hell these things mean inside Unity's post effects stack

 

Chromatic Aberration is an effect that I enable to make 3D glasses work, right?

 

I've started to really start at the bottom and read about stuff like how our eyes respond to brightness and color range.

There are also very different goals...

Are you trying to mimic the eye?
Are you trying to mimic movies or photos?
Are you trying to mimic paintings?

These three things are connected, related, but very different. Personally, I think that people focus too much on trying to replicate photography and film, instead of paintings.

I believe that really excellent paintings come closer to replicating the feeling of how something looks in real life. Photos have very different properties.

Witcher 3 is probably the best example of the "painting" quality instead of the hyper realism of archvis or Battlefield.


Really great color:


This, to me, feels much closer to a painting than a photograph. Although it may look different, I think this "feels" closer to how we experience visuals in real life.

 

I'm all for it, what I'm not really sold on is the idea of just talking about stuff in general terms without examples and details. I'm poking around a lot of things at once, looking through a lot of information and watching tutorials etc. There's plenty to find out but I want to stick to specifics, or examples, or otherwise it just becomes a vague jumble of words where nobody really learns anything.

Personally I'd rather see a few simple technical terms or processes discussed with examples, than hear a verbal description of some deep and extraordinary relationship between a host of different effects at the same time. I don't really know how to learn from that, or what to take away from it.

Just listening to myself trying to describe 'flatness' in 200 different ways is bad enough.

 

The "Tweek hole" illustrated by @Frpmta earlier in this thread...

Spoiler

This is, IMO, the real battle when we're trying to work on lighting, although these are a little extra extreme. You end up going back and forth and back and forth.

What's "Good"?
What's "Correct"?
What exactly are my goals for the visuals here?

Whats worse, you finally get the "perfect" balance but then you move onto a different part of the scene and suddenly everything is a disaster because you hyper optimized for one location.

 

Little test scene I was playing with yesterday to look at how enlighten works. Just a bunch of cubes. Threw kind of random textures on different cubes so I had some normal maps and stuff.

This is baked gi, everything static. The biggest problems I was having really had to do with light bleeding and stuff (shadow bias).

This has light bloom, tonemapping, ao.
Here's the same shot with no post.

It's worth noticing that the post is:
- creating light fall off where there is little to none
- hiding some of the problems with shadow bias
- hiding other problems with weird halo under the cubes on the left (indirect lighting bug?)

Worth noting: some/most of these artifacts are because I just slapped cubes together and some just flat intersect others.

 

If you want baked GI then use the progressive lightmapper. It's waaaay better. If you want semi-dynamic GI then use Enlighten.

Take a look at this official Unity project to learn the best practices of using Enlighten:

https://www.assetstore.unity3d.com/en/#!/content/73563

 

My project is on 5.4 I can't upgrade right now.

In terms of using any sort of mixed mode - I think the results are horrible. Real time and baked elements look so different that it ends up being kind of awful looking regardless of light probe configuration.

Ends up looking like 1980s cartoons, with hyper detailed backgrounds but the foregroundis super simple, layered on top.

As a "whole" in an actual game, I tend to think everything looks better as straight up real time since the lighting treatment is more equal across all elements. Am I crazy?

______

EDIT: The tutorial link has some excellent docs:
https://unity3d.com/learn/tutorials/topics/graphics/realtime-resolution?playlist=17102
^^ pure text, goes over everything

 

Maybe you are since most AAA games use mixed lighting

 

Yes but what's the gameplay like?

 

Well I was hoping to finish off the one in Unity and throw in a comparison, yes the one I did was in Unreal.. I spent 4X as long in Unity (just tweaking as I already had the meshes / mats) and had so many issues trying to balance it.. I specifically chose and indoor / outdoor scene because they cause the most amount of headaches, I have to agree the progressive lightmapper is simply miles better than Enlighten but it's things like the skybox influence that cause little issues..

It doesn't cast any shadows and to get a hint of colour reflections like you would IRL, so you have to crank it then the tonemapper causes it to look odd. So you either have a somewhat dull scene (I suppose like the Adam demo) or it's completely imbalanced when in conjunction with tonemapping.

There doesn't seem to be much influence in terms of shadow offset, the grading system is limited (you're far better off using a LUT). A lot of glitches in the post processing stack and finally Unity kept crashing constanly, yes it's a Beta so I of course won't complain.. I might have another crack at it when I'm free over the weekend.

 

This kind of thing is what I tend to have the most problems with, over and over.

It's crazy hard to hit that sweet spot.

It could also be little things like preferring "intensity" scales vs using real world units, so figuring out where to set everything just ends up having way more guess work.

Light intensity and tonemapping are so deeply connected that any change to one really needs a tweek to the other. Eye adaptation should handle this, but in my experience, it tends to blow out the color instead of balancing it.

Its worth noting that there are very real gamer complaints about eye adaptation in AAA games like battlefield also, this isn't limited to Unity.

https://forums.battlefield.com/en-us/discussion/104922/these-lighting-effects-are-hot-garbage/p5
^^
A massive complaint thread about blown out lighting. Some interesting examples in a massive AAA game out in the wild.

 

Tell me about it, I'm like a moth to a flame so I did finally get Unity to stop crashing and I did at least try and get somewhere with it as posted below.. But I am done with it now, I can do the proper comparison now at least.!

Tweaking huh!, once you pop and an all that.. Lets see if it's up to Billys standard of lighting ..

 

Thank you so much for making the same thing in both engines! Not many could do that.
I actually like the Unity one more, with exception of the bright rimlight in front of the shadow on the left.

@frosted: that's a sick collection of screenshots in the BF forum. They really do look terrible!

 

Well The unreal one is really good technically. The difference does show, it has much more range of nuances than the unity one ...

But that's clearly not the same, the unity one has new source of light with the torch, that throw any comparison out already

 

Unity 2017 beta - WIP but trying for something like this...

I've probably not got the scale, angle or FOV right, just yet.

Is it me or is the Unity lighting in 2017 looking better?

 

Lol, lets not be silly they are area lights pointed at the wall which have little to no influence on the rest of the scene ..

What is unfair is I have spent around 16 hours more on the Unity scene..! I mean that's bound to make a difference.

 

Another way to view it is either intentionally or subconsciously @ShadowK just showed how making little changes to the design itself are more effective at making a great scene than just the engine tech.

 

I don't buy it - you can do better than that in unity. Where's the STUMP?! It needs that foreground stump.

But seriously, there's no indirect light on the arches or ceiling! I'm pretty sure you can get better results in Unity. The only ceiling light is coming from the point light on the torch!

You had very light bloom running in UE right? Or is the blur off the lit areas from something else?

Clearly the UE scene is dramatically better, but I call foul!

 

Yep it's because of the stump of course, I'm using the progressive lightmapper and I honesly stuck the bounces at 4 (max) and increased indirect intensity, so err what else can ya do? (besides increasing the damn skybox influence but it's bad enough as it is).. I'll have another look at the mapper and post them side by side with an updated UE4 one with all the "real" tech like DFAO etc. switched on..

I mean I am completely open to suggestions on either side to make it better, so keep it coming. Just to remove confusion I'll get rid of the old one's, but again advice = good.!

 

Possible to share the scene? I'd love to poke around with it. I wrote out a list of like 30 different problems and differences but it started to feel obnoxious

Three major points though:
- Focus of the eyes are ceiling and arches. Arches in Unity example have weird normal or noise, major problem.
- Light color is entirely different (orange in Unity vs Yellow in Unreal, look at the tree color on the right)
- Tree in rear needs to be dim. It ties the scene together and makes depth feel correct.

I'm 100% sure you can get better color on the arches and ceiling, although no clue with progressive light mapper. I would have used directional light with extra bounce, but your rig may be very different than any approach I would have used.

Can you share some of the details of the rig in UE? It seemed weird to me that the left wall had so little light on it but the ceiling was affected by so much bounce even near the edge. I thought you just used a different material.

 

It's going to be very difficult to get it like for like, especially due to trying to adapt for colour grading and trying to play to an engines strength with completely different lighting methodologies and technology..

Y'know I did try to make the tree in the back "dim" but the tonemapper wouldn't seem to allow for it, if it's going down that path and it's going to take this long to try and sort out then what's the point? Why not use the engine you believe you'll get the best results out of instead of pratting around for weeks on end? You could of learnt so much more about how to improve the graphics or whatever in your own project ..

I really don't have that sort of time to spend on this where I'm quad checking every single little thing. In fact I've already spent way too much time on this.

 

The art is super nice, and the Unity scene looks really good, but I would bet that your heart wasn't really in the Unity scene (or you got annoyed at it taking too long).

It's clearly not in the same ballpark. But I still call FOUL!

Unity can produce better bounce lighting than that example is, 100% sure. Maybe not as good as the Unreal example, but certainly much closer.

EDIT: Also, thank you for doing the comparison, I'm sure it took much more time than you should have wasted - trust me - I'm familiar.

 

The "bathtub" in the unity version look detached from the ground, just that show how different both lighting are ... but also it mean that the bake resolution is superior in the UE. Is there post process on the unity version? No ao?