"Photo-realistic" quality rendering between Unity and Unreal 4

Daz 3D with Go-Z bridge is probably the way to go.. But that ain't cheap either.!

 

UMA2, but i don' t find it simple to use or make custom characters library for example. And UMA 2 is no more Unity support but a community project, that's sad.

Too much polygons until you use a Sympligon license or make your own LODs.

 

Nope, use the Daz decimator.. Usually cuts things down to about 20K Polys which is more than enough. You been doing this long? ..

 

I think its pretty clear for my game, that we would never dream of outsourcing characters. For me that's a crime, and horrific tbh.

 

If you display far characters you could have more LOD , this is more memory but better performance , so you could have 5000 poly versions and 10k poly versions. Or you only display characters on some near distances.

 

I think if you aim for a AAA look, you have the opposite problem: the Daz model has too few polys. Most games today have much more polys for the main characters than the Daz base model.

 

You're just waisting time if it's not needed, it's never poly counts I ever worry about on PC / Console (as 11 Million is fine) it's materials / draw calls and stuff like shadows / post. It really depends on the scene..

You do understand you can create LOD's from decimator too right?

@thxfoo

Then change the resolution to be higher? It's a slider, I mean seriously come on guys I'm beginning to wonder here with the amount of mis-information and random no experience comments what is going on?

 

Our characters have 50K-70K tris. Today's AAA games character range between 50K to 100K. Although Horizon main character is around 500K tris.

 

That's fair enough, but frankly that's the sort of moralizing that I think makes it hard for indies to reach anywhere near the heights of the big budget studios. Even these studios outsource this kind of thing, except that for them it isn't about cutting costs but rather getting the best quality. The problem is, you could easily extend the same approach to the environment, the animations, etc.

The way I look at games, is that I am the producer, director and screenplay writer. If we were making a movie, we couldn't spawn an 'original' Leo Dicaprio for each movie could we? Unbelievable that there are a number of movies out there using the exact same face! shame on them... no, I find the best quality stuff out there that enables me to construct the experience that I want to create in a reasonable amount of time, and I work with it and shape it to my story. There's no other way for me to do what I want to do.

 

Is that the mobile game, or the PC port?

 

Both. I should mention that this version is only for high-end mobile devices (minimum supported A7 core), consoles and desktop.

 

Yeah, was about to post "HOW DOES THAT EVEN WORK ON A 4S?!" but then noticed that you dropped support for 512mb devices.

Makes sense.

 

The current version on the Store (Alternate Realities) do runs fine on 512MB devices such as 4S and iPad2 but crashes a lot on iOS7/8/9 due to the OS taking more memory and killing memory hungry apps. The new version (Director's Cut) the one I posted few pages back is a more advanced version and requires even more powerful devices, that's why the minimum is A7 (even though we still playtesting it and will probably drop for A8 as minimum).

 

This thread needs more images and less talking. Can't keep up with all the posting.

 

Lol. All that does is turn on subdivision, it does not actually add more detail, it just gets smoother.

The main hero of todays games have many fine details that is why they have so many polys. Otherwise they would distribute the low poly version and just apply the subdivision on loading. The reason they do not do that is because they need the polys for the detail and not for the smoothness alone. QED.

Also see this:

You think that character is actually a 25k character blown up to that number by subivision? Nope.

 

I might start a new thread with the breakdown, I think this one has had its day..!

@thxfoo

Umm miss the bit where I said "Go-z bridge", y'know you can add detail until you're blue in the face. I'd never use default stock characters for an actual game..!

ahhh! LOL!

 

I agree, I don't think there's much more to say on this thread.

If you want to open a thread on how to make a solo AAA game, I'll be there.

 

This is what I was hoping to get out of this thread:

I think Unreal is superior in terms of graphical ability, I wanted someone to turn round to me and say, no Shadow you're wrong and here is the TECHNICAL reason why. With a massive breakdown of everything, I can do the UE4 stuff and do a technical breakdown and explanation of everything I do (based on something we all agree looks good).

But I need someone to fill in the gaps with Unity, because if somebody or (I) could do it. Then someone would have (or I would have) and this whole discussion would of been relatively short.

I'm talking in the context of a game as well, not just an arch viz render.

I can post, this is how you do it in Unreal but I'm pretty sure it'll be slightly one sided and not really for the Unity forums.

 

I feel exactly the same way, it seems to me Unreal has a fundamentally better rendering system but I would love to be proven wrong. If someone could do that, I would be the first one to sing praises. As I've said many a time, Unity is a fantastic engine, but you have to give Unreal its dues in the graphics department. And if Unity is lacking some Unreal-level rendering features I find it hard to imagine why, when AAA studios develop their own engines relatively quickly that outclass even Unreal, and I have no doubt that Unity has access to the necessary expertise.

If you want to go ahead with your comparison, I'm sure we would all learn something. But it seems to me that it wouldn't really serve all that much since we all seem to have a different idea of what we realism is, what we want to see in our games, and so on. You'd probably come up with a comparison, some people would say it looks about the same, others would say one looks much better than the other, and we'd be back where we started.

I encourage anyone who finds out about the details of the differences in the engines to make a thread with the technical details in the OP post, since just putting the question out there generates a lot of uninformed responses.

But I would like to continue this discussion somewhere leaving Unreal out of the question, just asking how good Unity graphics is and how we could improve it. Unity does seem to be putting in a lot of work in graphics lately and it would be good to compare experiences with the new post fx and keep the graphics discussion alive.

 

I think there's a fundamental misunderstanding here. You're not supposed to. You're supposed to understand you're humble, release a couple of great games and use the cash - presumably a few million, or even a few hundred thousand - to fund greater titles. Think incremental. Anything else is a get-rich-quick scheme that's a) pointless since your games will suck due to being only driven by greed and b) a flop, because if its only really driven by greed, it won't in fact earn much on account of being rubbish.

Now this is not directed at you, but whoever might be thinking of making a mistake that the first game needs to be brilliant and awesome AAA killing. I'm here to say, it shouldn't be the objective. But then, hobbyists wouldn't worry how hard it is to do, because it's just a hobby. There's no time limit.

 

Yeah, I agree it might be worth leaving it out.. Because I don't believe Unreal is just better than Unity graphically, as it stands right now it is THE best looking game engine out there (maybe Fox engine and / or whatever Squenix use is about on par). It's a bit of a freak of nature, whereas I like the CryEngine screenshots but I get the feeling with the updated post etc. there is nothing to stop you from doing that in Unity.

Then again bar a new experimental voxel based GI, not a lot has changed over the years in CE. I've definatley noticed as of Unity 5.X, CE can't beat Unity in terms of interior scenes.

 

For your first game, maybe. I cloned a few simple games when I started, but I wasn't interested in finishing them. I don't have any interest in them. I have a funny relationship with games, they don't have inherent value for me, I don't consider myself to be a 'hardcore gamer' or someone who derives pleasure from the experience of playing games as such. What I really enjoy are experiences. A game, a book or a movie that gives me a great experience is something I treasure. But most games, for me are frankly a waste of time. There are a few games which have given me a great experience, but they are few and far between. The reason I'm here is because the experience they've given has been more powerful than other experiences, from books or movies.

I want to create experiences. That is the artist in me. I want to communicate messages. And a large, large part of communication is the visual experience. For me it is extremely important, because a lot of my enjoyment of games has to do with the visual experience.

And lastly, I like the idea of doing something that hasn't been done before, especially when I think I can see how to do it. I need to challenge myself and make something beyond what might be expected from a single developer. If I was going to make something well within my capabilities I would probably find something a little more well-rounded in terms of job satisfaction than game development. I don't mind hitting the wall, I might find that I'm just not good enough to make the game I want. But I'm going to try, with a good, intelligent, well-calculated leap. I think I can see a lot of mistakes that people make, ones that I can avoid. Maybe I'll succeed, maybe not. But that's the goal.

 

No and I really meant 500K modeled, not tessellated on the GPU. It's not pre-subdivided neither, characters nowadays are sculpted or scanned in high-poly then retop to lower poly densities.

 

yes.

Well some games like Assassin Creed ot GTA needs good LOD.
The difference between 30 distant characters having 30000 polygons ( 900 000 polys total) and the same ones with LOD at 7000 ( 70 000 polys total) , this will make a real difference and animations , items attachments takes some CPU-GPU also.
This is a whole package, you don't have unlimited polygon budget , if the level is already high polycount on near view you can't allow distant characters using millions polys.
Until you want to run your game on latest hardware only.

And there is main characters and secondary characters , they don't have the same budget polygon.
Also character versions for cinematics in some games have higher polycount than the gameplay character.

SVOGI in Unity ? I doubt.
Unity as fast as CryEngine ? Impossible.
If CryEngine had C# , a high level framework, better workflow, it would be my number one engine. Because it looks as good as UE4, it runs lot more faster than UE4 on old PC , and the game size packaging is smaller than UE4.

When the hobbyst want to make the game in a reasonnable amount of time and plan to sell the game , then the limit is no more infinite and the game goals become more humble and realistic.

 

No longer hobbyist. It's never both. Once you start thinking that income is important, then you start choosing. You can't have massive production values and be indie with no budget. If you want to use this for money, you should be sensible about what can be achieved.

Big mistake, it's a bandwidth cost, so it's not a hard rule. Also, you forget all these meshes need to be drawn again for shadows. Your 11 million became 22 million, or depending on the game and FOV, your 11 million became 100 million as you see more and more. Not only this, dense triangles can be more costly to render.

Infamous second son is probably highest poly count on current gen consoles, and it doesn't run that well. Basically if it looks like it could look almost the same lower poly, it is only beneficial to lod it. Simplygon is amazing and virtually foolproof.

 

I'll start with this first..

1) I said it's dependant on game, who here is trying to make the next GTA? It's not a concern I have to worry about.
2) Of course, I said an allowed poly limit for our game right? I'm not sure why you're even saying this.
3) Who said anything about Unity and SVOGI? There is such a thing called Enlighten, so not sure what you mean there. Some interior scenes as displayed in other photo's show Unity is more than capable, as fast? Well if it's not tweak it.
4) Nah, doesn't look as good as what UE4 can produce, will agree it's more performant though.

@hippocoder

Again, quite a bit of noobishness going on there too ..

You understand you'd never try to widecast shadows across a whole scene right? How many polys you have in your scene and what you have in view at any one time is completely different. Even if it's 11 million in view, generally the smart idea would be to bake shadows widecast (which has no impact) then overlay something like CSM's at short distance. Remember if you're doing it properly, for a huge scene you'll use a render farm / compress and stream smaller LM's but of course there will be a trade off in quality..

Even if you wanted wide cast shadows and realtime, it'd be wise to use distance field generator information to pre-calcuate distance mesh fields for long standing shadows to cut issues with performance dramatically. Not sure how console would handle that, but it has worked find on PC tests..

Even Umbra does shadow caster culling if your scene is designed correctly:

http://umbra3d.com/press-releases/umbra-3-visibility-solution-powers-up-unity-4-3/

The only situation where you'd have this sort of problem is if you stacked 11 million polys right in front of you (like in a corridor). Or of course, if you didn't have a clue what you was doing..

I understand nothing in games is a hard and fast rule, I'm telling YOU I've had a scene with 11 Million polys in it that ran absolutley fine on console.! Ok?

 

I can't make decisions based on anything except Unity, which I currently use, and so my working practises will align with what Unity is designed like for best perf.

 

I've spent a good amount of time in the last month or so looking at some of the details in few AAA titles. Little things like how AI responds to you and combat in Assassins Creed and Shadow of Mordor. How they make their environments feel so good.

AAA is outside of the reach of small indies. Forget about the Uncharted 4 car scene (which is some of the most impressive stuff I've ever seen in any game), even the more mundane stuff is so polished and so varied and complete. The labor required to come close just requires too many hours on details that are too small.

As a small, ambitious indie, you simply cannot bog down on the little details to that degree. You need to invest yourself in the things that produce the most game play for the least development time, so that you can free up enough time to really polish up a few of the most critical elements. Modern AAA doesn't make sacrifices like that, the level of polish throughout is just crazy high.

It's also really worth noting that that polish that I'm talking about is not poly counts. The real polish is extremely cross disciplinary, and involves really impressive code, audio, level design, artistic design, animation all coming together - each contributing really, really solid work. Again, you could lower the visual fidelity in these games many notches and they'd still be incredibly impressive because of the contributions from other disciplines.

 

Glad to meet sane people for once

 

I hate to say it but what @frosted is saying is (or should be), slap bang in your face obvious. It's kind of like saying the sun is warm.. But it's not the purpose of this thread, we're trying to find out what us small little indies can do to push our boundaries, not that of a 200 man AAA team..

With tools given today, if you can't within enclosed spaces match a high end graphical AAA looking scene then it's your fault not the other way round. Because it's done time and time again in various arch viz renders or showcases from other small indies..

But where we get caught up is semantics, you know how long it takes to model an entire city? Have you every tried to do it? I'll save you some time, too long for an Indie. It's not art that's generally the specific main issue either. It's trying to build a living world..

 

You know the difference, SVOGI is Gi and Zero baking , no lightmap artifacts. Enlighten is a lightmapper and very very long to create a small lightmap.

For games CryEngine looks as good, anyone will see great lightening , vegetation , water and characters.
Players won't see if it is Lightmass or Svogi , and while playing you won't notice small differences.

You can't compare Achi Viz and a game like Uncharted 4.
I forgot you talk about Skyrim indeed, yes Skyrim can be done, i think @frosted is talking about the big AAA boys like Uncharted 4.
It is our fault indeed if we can hire 100 guys to make Uncharted 4 within two years.

 

Nobody's disputing that. As I said before, the last 5-10% of polish is what you have to do without. There are some things that look easy and are hard, and some that look hard and are easy - or relatively so. Audio (at least for me) looks easy but it's hard. But AI is the other way around.

We've got to get as close as possible, and really look at things technically to find out what we can do, how far we can get. And a good rendering system is absolutely fundamental.

 

I just know indies will not be able to make big games like Uncharted, or it will be a really inferior version in all points.

 

Actually you posted a great example of the sort of area where I think indies could innovate a LOT. What's a mocap system? Just a bunch of sensors that continuously transmit their position in space. Nothing mind-blowing there. I really don't know why it costs so much, I've considered trying to make one myself.

Yes, these mocap systems are polished. Yes, the software isn't as easy to make as it seems. Yes, the sensors are somewhat expensive, although it is an area (robotics) where the technology is evolving extremely fast. But I can just imagine a Raspberry Pi mocap system turning up somewhere

What is really hard to get/make are good actors for the mocap, voice acting, high-quality audio, things like that. All the things that computers can't produce, things that take people with skill. Not to mention the ability to write and direct a good scene.

 

We're not going to move forward if we're constantly stating the obvious, stop talking about what we can't do and more about what we CAN do. I've still to see anything useful in this thread, we're squabbling over basics you learn in the first month of games development and it ain't progressing.

Are you going to show us something AAA worthy at any point? Explain in depth "technically" how you did it?

 

This.

 

Ok, so I recon we've got a decent setup for the most part.. Here goes:

Audio

MOTU 1248 X 2 and Apogee Symphony (AD / DA (audio interfaces), Daking Mic Pre-amps, Pro tools, Spitfire Audio Albion, East west strings / symphony orchestra, UAD plugins for mixing and TC brickwall for mastering (along with UAD hardware compressors. We use Barefoot MM27's for our main mixing speakers..

Mic's are Mojave Fet's, Shure SM57's, Shure SM7B, Neumann U87's.

Instruments are Midi keyboards, have some six / seven / eight string guitars, Drumit 5K (electronic kit), bass guitars etc. and so on.

(Just to note, a lot of this stuff is from a studio I used to own as I was big into music).. But it's kind of expected audio quality level.

Mocap we use a mixture of OptiTrack prime and flex's..

Game development software:

Of course you know we use UE4, but in terms of Art we have Maya LT, 3DSMax, Modo, Substance suite, Quixel, Topogun, Z-brush, Mudbox and 3dCoat.

Development systems:

We have a bunch of dell poweredge servers, which we did fritter around with the idea of adding an "on-line" mode but as of the moment we use them basically as a lightmass client and fileshare.

Desktop:

We use a range of system setups for testing, our dev PC's are I7 5820K's / 32GB Ram / 1TB SSD's and all have GTX 980TI's, the test platforms (all twelve of them) contain various things from I3's to AMD's all the way to our Celeron / GTX470 base test system.

Point being that the initial investment for A or AA titles (not AAA) is still rather hefty. But there is technically nothing they can do that we can't.. Well besides the thousands of man hours..

 

Nice! Always good to see how people are setting up their stuff.

Are you using mocap actors/voice actors, and if so, how do you go about finding them? People you already know?

 

Some of them yeah (actually a lot of them), with being into the music scene quite a lot. You'd be surprised at how many of them have ridiculously expensive home setups and are well versed in recording. So I asked some favours and got quite a few to help out, I do outsource bits and pieces from time to time but not often.

 

Small parenthesis. In the case of Unity yes, Unity do not give you access to the already available depth buffer, instead it re-renders the scene when you need it. Why? short answer because it's bit tricky (you'll probably lose the ability to render cameras with custom shaders). Also, the cost to access camera depth/normal (or whatever Gbuffer RT you need) is free if you are in Deferred Rendering because it's already done when constructing the Gbuffer. That's the main reason why DR was invented.

 

Your project is huge

What do you think about workflow to make a good terrain outdoor between Unity and UE4 ?

 

The workflow between the two isn't actually that much different, use worldmachine pro to spit out tiles so you can effectively stream the world. In UE you'd use weightmaps instead of splats, the major differences came from shaders like in UE there are bump offset nodes and parallax mapping setups you can use which tends to add a little bit of depth. So you need to be pretty handy with Unity and shaders.. Obviously POM has the biggest impact on performance, so again a fallback is a good idea..

Dependant ont the size of your game, you could look into sculpting the tiles in mudbox or Z-brush to add a little more character to it. If you really want to go to the next level, 3D photoscans of rocks that are re-topo'd gives that "next gen" look.

Adding depth (variations of hills etc.) and spending a little time doing variable decals really does help give it that polished look, rivers with flowmaps are always a good idea. Paying special attention to colours of foliage is a must, even with stock speedtree I'll still open some of the textures and de-sat etc.

World composition in UE4 is very good, on the asset store there is a metric boat load of terrain segmentation tools.

Apart from that, it's just a matter of filling the world effectively with art in a way that looks good but also adds enough occlusion to keep performance up. Instancing tree / grass meshes goes a long way in terms of performance, heavy LOD'ing etc.

Mess around with the position of directional lighting, in Unity offest the amount of IBL in Unreal use minimal amounts of skylight influence (also I usually mark lower hemisphere as black and modify the position (height) of the skylight then blend).

Decent AA and colour grading is a must, in UE it's actually much more tricky to get it right. If you want top notch terrain I'd recommend using DFAO..

 

Ye the technical side of structuring terrain in Unity is pretty much the same as UE4 as long as you use full featured list of Asset store tools to get there. So lets say that your aim is smaller terrain and as long you dont have to do the terrain stitching in Unity which is messy then its all fine. The differences in look are going to come down from things like quallity of vegetation shading/shadowing, lighting quallity which has the largest impact for terrain in my opinion or anything outdoor related, performance of what you get in, DFAO (if there is a lot going on rocky wise or bumpy wise)... UE4 is default better choice, i just couldnt get any scenes look good in Unity to my liking in Unity for outdoor to what the intent was in first place, or its perhaps just my lack of art workarounds to do that, in UE4 usually everything turned out fine rendered. Forest for me is solid game as in terms of outdoor terrain and thats the only one that i liked from Unity render side.
Very nice feature is Artomatix software, not sure if there is Unity release yet, for now i use it only on per-texture workflow, and from their video on Unity/UE4 landscape it does amazing difference, simply procedure yet large differnce on such basic thing as tiling.
By the way what did you mean by instancing the foliage? As long as your using paint vegy of terrain function and textures are atlased at largest part you are pretty much there already, or did you had something else in mind?

 

Still haven't finished Unity->UE4 conversion. Getting there (need proper mesh conversion, because the old tool couldn't do that properly). The only suitable package was Lab demo.

Checked out (unity) Courtyard demo, was very disappointed by it - it is a massive custom shader rewrite, not a standard unity shader. Ye mythical average indie won't be able to pull off anything like that. I didn't exactly had the patience for it to finish baking too.

On Unity side apparently dumping certain mesh data into file crashes unity somewhere deep in C++ json-related code. JsonUtility also inherited all the flaws of standard serialization mechanisms of unity, meaning it does not understand null, can't serialize List<int[]> and recursive structures are a big no.

On Unreal 4 side, apparently all the code that is responsible for creating assets within content browser was written by Satan. I can honestly say that I would have hard time remembering standard package creation ritual if anyone asked me. Also, UE4 can't load tif files. Of COURSE the lab scene was only using tif files instead of png despite them having inferior compression. And of course all of those were marked as non-readable, so I couldn't automatically correct them through a script via PackToPngOrWhateverItIsCalled function

Well, I think I'll post update once I'm done.

 

Is it just me or Unity roughness are lacking fresnel?


Edit : So it seems in unity you have to set the roughness/smoothness around 0.05 or 0.1 to get the same roughness as UE4,
hmm maybe i'm wrong, but idk

 

I am a noob, so don't listen too much to me.

But I have to admit that at first glance, I thought the left image was Unreal 4 (because I thought it looks better).

Until I saw the logos

 

Don't worry i won't judge

 

Can you attach bigger images? Hard to see with tiny ones.

 

Well the problem is i cannot find the same scene from the UE with bigger images, hmm i might just need to remade it with the same assets that i have. stay tune

 

Were the images from marmoset or substance designer or something similar?
If that's the case, it would be a better idea to recreate them. The engines keep improving, so something from half year ago might produce different results.