A different perspective on Unity development.!

fyi I think game industry is really missing an golden opportunity to bring cutting virtual grass down to the masses..... lawn mowers, hedge trimmers, scissors.... they should be mandatory working tools in any survival game. I'm totally being serious ok

 

The last on my list is Stingray, I've given it a quick look before but never tried to make a game in it.. Although first impressions are pretty damn good to be honest, especially as I got a free support contract in place when I got Maya.!

Opened it up, imported a HM and low and behold shadowmaps actually work properly.! I mean no flickering, no weird abrasive look.

I've had to dig into the code examples already for stuff like frustrum culling etc. but it's as easy as Unity.. For e.g. on the frustrum bit (Camera.set_far_range(camera, 10000) -- Needs long range.!!! Because it's a big terrain (duh)!!)..

It's not exactly rocket science to figure out most of it, the UI is clean and all in all first impressions are it's pretty impressive.!

Performance so far has been examplary, but we'll see once I've packed it full of "stuff"..

 

I'm guessing mostly, based on hearsay, but I think the publisher had a vastly different vision for the future of the IP than Techland, and so they gave it to Yager. And they started from scratch, based on the version of the engine that they've used for their previous projects. Techland then made Dying Light, wich is both in tech and in artistic vision, closer to being the "real" Dead Island 2, than Yager's Dead Island 2 ever could have been. I blame the publisher... Someone must have thought it's too bleak and depressing and needs more goofy stuff. Well, turns out people like me prefer the bleak and depressing zombie games ^^.

 

The thing I remember most about the originala game (and maybe the Riptide version too) was the modified shotgun. Someone took the saveGame data from a xbox360 memory stick and "hacked" it to give the shotgun infinity ammo and infinite power. Then people exploited the weakness that a player could drop a modified weapon on the ground and another player could pick it up and transfer the weapon to their permanent inventory. Almost everyone did it/ The game was fun as hell to play just because of that. I;m sure the developers knew what was going on but didn't care. Now COD on the other hand, you're account would be reset and probably banned from playing on xbox live if you did something like that.

 

The thing I remember most is spending 8 hours trying to fix my broken xbox 360 savegame with a hex editor. That happened before they patched that bug (their patch wouldn't have restored broken saves, just prevented it from happening in the future). I mostly succeeded (some sidequests reset, but I think I still had all my gear). Damn was I proud I managed to get that working again!

 

I'm very interested in what you'll find. Beast should be pretty mature by now.

 

Last grass post.

Spoiler

So this has some really dense grass, there are only a few spots near the camera where it's possible to see the ground. This has no special code or shaders, and it's running on a crap laptop with a ton of post, etc, all running ~40+ fps. Just good spacing, placement and settings on the terrain. Now to figure out how to make it bend properly

edit: without the grass, it runs about 5fps faster. The post and sky eat most of the integrated gpu.

final edit: worth noting that the grass here costs about 4 draw calls and 20 batches (deferred), so realistically I think you could do this level of density on mobile without doing anything fancy just through optimized settings on the terrain and well spaced placement.

 

Yeah, should be interesting.. I have noticed one "constant" most well known large "open world" games have been made in proprietary engines made exactly for the task at hand.. Elder scrolls (Gamebryo), MGS (Fox Engine), Dragon age (Frostbite), Witcher (Red engine)..

Even though cryengine is released to indie's it still feels proprietary in nature, not saying it can't be done in most engines although it is far easier with a custom lightweight build made for the exact task at hand.. That being said, the min specs (AFAIK) for witcher 3 was a 660TI and it's a bit laggy on console. On max settings I only get 45 / 50 FPS with a 390X..!

No matter what engine you use, you have to do a LOT of tweaking to get the performance right for base machines.. It's fine having it run 60FPS on a 390X or 980 but you're still going to have to squeeze every last ms for lower platforms (even if it's a 470 as your min specs).

That's why if I'm getting anything less than 60FPS on a 390X or 980 I get very worried..

 

Just a note, I submitted my findings on Unreal on their answerhub and asked Epic to look at it.. I'm even prepared to re-write some of the renderer and submit it back into the main branch if they allow it. If it's a no-go, I'm not supporting the whole engine and fixing all issues in the branch release, I'd rather go completely custom (or use something else).. I don't have the time or the patience so I'll just stop using it..

 

Haven't you been working on that something else already?

 

I have but it's far from ready as a production engine..

 

For outdoor scene with lots of foliage I would still use CE. Different tool for different job.

 

If I were doing it again, I'm not sure what engine I would use. My visual goals are a little different than most games, I'm shooting for a grassland style area, with very few trees. Since I'm targeting low end hardware, I need to avoid dense trees and plants, but I still want a rich environment.

The key elements are: hills, sky, grass. Some of the references I've been collecting:





The idea is just to make it look good while using as little actual stuff as possible.

 

So Stingray, a very interesting one..

Firstly the good, tons of cool middleware.. I messed around with Scaleform / Navmesh etc. etc. and it all worked as expected even if in some cases it can be rather tempramental. Scripting (APPKit) they use is a good starting point, although you can literally work from a blank slate and build everything up in LUA.. From animation controllers to camera "wrappers" etc.

Graphically it's not bad, sort of between Unity and Unreal and beast works great. Comes with tons of stuff Unity really should have, like deferred decals and a decent terrain system for one.. Didn't get problems with "noise" from foliage due to some rather decent temporal anti-aliasing. I've tried the experimental post pack from Unity, but they still ain't on par with Stingray or Unreal.

Cloth / character systems and material editor seems to be based around Maya with HumanIK support which is mighty cool.

The bad:

Performance actually wasn't that great when I started stuffing it full of stuff, actually far worse than both Unreal and Unity using non static lighting. The biggest issue is for a closed box engine it's simply not extendable enough, like for example they have a scatter system for foliage.. It has a frustrum distance, there are no controls (distance wise) for it and no way to access components of the terrain system (from the camera to scatter system) via LUA.

Also they tied the foliage to the splat paint layers which is limited to a base / RGBA setup (ummm wut?), there's no scriptable editor extensions etc.

The API / documentation you get from Autodesk is pretty damn good, although there aren't enough user examples laying about. It simply isn't mature enough at this point like Unity / Unreal and CryEngine is.!

Sorry but Unity does win, damn it I can't use that as an argument anymore...

Huhhh, I'm going to regret this.. Next up is CryEngine.!

 

Looks umm a little bit "artistically" touched up.. But I suppose in engine it ain't far from.! Also going to move the Unity one down here, because I want to see exactly "how much better" it is..

From what I can tell, I just need MOAR FOG.! And tons of god rays, maybe push the highlights / back off on shadows.. Y'know artsy crap.!

CryEngine



Unity

 

I totally didn't read about 90 percent of these responses so this may have been covered already - but about the guy saying he had issues with vector 3's and floating point problems - why didn't they just create a Vector3d class (vector 3 double) and just use that instead? I mean the vector class itself is not all that extremely complicated, for instance Vector2i is a great way to use integers and it extends most of the existing stuff vector3's and vector2's can do with floats as well... why not modify it to be Vector2d (or 3d) instead?

Maybe this guy is right though, I've yet to work on a "huge" unity project, and really only mostly stuff that by publishing time is less than a gig project file.... so yeah that and maybe working solo means I don't run into these issues...

EDIT: oh it is occurring to me though - internally the transforms of objects are floating point vectors aren't they... no changing that is there...

 

No, but depending on where you need the precision you can do your own work with doubles before setting it to the float.

 

At the risk of repeating what I said in the other thread, I'm just going to say, the thing that Unity doesn't do for me is give me a correct sense of depth with the lighting. This is a totally naive statement (i.e. I have no idea what I'm talking about) but it almost feels like the Unity lighting has a lower precision and kind of stratifies it. When I look at a lot of Unity shots, I can almost imagine that it's a bunch of 2D images set at varying distances from the camera, whereas other engines really pull it all together and there's a really continuous unbroken transition from light to shadow that makes my eyes relax.

Anyway, I saw this in the WIP thread and thought it's probably the best Unity terrain I've seen. It still has the same issue imo (looks fantastic from afar but up close is a little flat) but it shows what some extreme artistic skill can do to help things along...

 

It's only shaders chief, I'll have to cast my mind back a couple of years.. But the first things we did was change the lightings quadratic linear thingy to an inverse square falloff and we were doing GGX before it was "cool".. We had custom shaders and all sorts of goodies, now the shader frameworks available have far surpased anything I did..

These comparisons aren't really about "graphics", it's about if I shove it full of "stuff" what am I left with? You can improve Unity's pipeline no end if you know how, but you need a rendering budget to work with.

Time is if course a major consideration, you don't want to be spending forever doing this sort of stuff but this is what games dev is all about. Unity is here to save you time, not to gloss over the fact a dev might not have the first clue about what they're doing.

I'll admit, since I've been doing nothing but art since I can remember I've lost the "mojo" and it's going to take time to get back into the swing. Although it's something I need to do, not particularily want to ..

P.S I do understand what you mean about Unity looking "flat" though, I noticed on the terrain especially that there seems be a distinct lack of normal depth rendering.

 

Yeah it may be just shaders, but I've literally not seen a Unity screenshot ever that satisfies my eyes the way renders in a lot of other engines do - except for maybe the Adam demo, so maybe good things are coming.

As an aside, it might be personal taste, but I really like the Cryengine aesthetic, that thick, rich, oily look ... I hope it's just down to post as I wouldn't mind a bit of that for myself, but I've also not seen a Unity screenshot with that look.

Anyway, it's not high enough on my list of priorities at the moment as Unity graphics are good enough for what I need them for right now, and there's signs that they are improving.

 

Since I'm back on with Unity, give me a couple of weeks and I'll have it on par ..! Seems I have to let bygone's be bygone's..!

 

Yeah, same for me. I really like that kind of art style. I guess to achieve more realism at this point we would need to have photogrammetric textures.

 

Well I'm not entirely sure that style is realistic as such, although it sits on top of CE's realistic lighting. But it's better than reality!

The Adam demo looks very realistic in lighting, but doesn't use that style - that's why I hope that Adam demo lighting + improved post = really nice game-style graphics, more like CE.

 

I liked that talk a lot - how they used photogrammetry in Star Wars Battlefront:
http://www.gdcvault.com/play/1023272/Photogrammetry-and-Star-Wars-Battlefront

 

Yes. Cryengine has this distinct look that I like as well. Usually, a trained eye can differentiate one engine from the other. Here is a shot that I like from CE. It doesn't look very real but is nice to my eyes:

 

While we're at it, I'll post this one I like (no idea if it's an in-game shot of Ryse or an impractical/marketing shot). I'll justify it by asking if anyone has a clue if/how it's possible to get this style in Unity?

 

After having recently read the background to you not liking Unity, I have to say your willingness to give it second chances is admirable.

I think you're on to something. Looking forward to see your results.
The main difference I see between the screenshots of the different engines is that Cryengine has less contrasts within the areas that are in shadow or lit directly, whereas Unity has too much contrast within both those areas. Fog and GI are possible ways to mitigate this. Since there is no practical GI in Unity yet, I'd experiment with tweaking environment lighting, Bloom, stacked SSAO effects that take lighting into account, and tonemapping. Also the translucency on the Foliage shaders in Cryengine looks a lot better.

 

Isn't the real difference volumetric lighting and effects? Volumetric effects have a just tremendous effect on visuals and give a real sense of depth.

Both fog and god rays are really just hacks for lack of true volumetric lighting.

 

Seems like we will have some of this tech in 5.4?

See the first post here:
http://robert.cupisz.eu/

Robert is Graphics Programmer behind Adam demo.

 

Those image shown lately, the first thing that strike me is not the tech, it's just masterfully composed image at a basic level. Be it the first unity landscape, homefront or the ryse. Tech do add to the final image (control of gradient, lighting, etc ...) but being able to diagnostic the image at an artistic level is the first key to get what makes them great. Unfortunately that's not a skill unity3d do out of the box There is always a way to cheat the tech to get a good image, any good artist can always punch above the tech capabilities by creating "cheat" to improve visual.

 

There are also some other guys working on it, there's an open source volumetric lighting thread in the wip forum. If anything close to it actually gets official release in 5.4 it will be a huge step up for unity. I mean, these effects are huge imo. Especially if you're trying to create a mood.

^^

(tbh, I don't think the ryse shot is artistically good. I think the composition is not hot. It compensates with noise, tons of lights, a million reflections, crazy godrays, etc)

 

Completely agree. Here is a nice shot from Unity game:

 

Well I did offset it later with HDRI IBL, but I got bored of waiting for Enlighten by the time I posted it. There's bloom, the new "SSAO" had some really odd artifacts and the old one isn't exactly great. So I'll probably use HBAO, they use SSDO in CE and it's pretty tasty. I was also using SVOTI in CE..

As for the rest, the lighting balance from the falloff (or lack thereof) is out. I'll re-do the lighting and / or use a shader framework, then slap tons of volumetric effects and sun shafts in it..

I'm using Aces tonemapping on that one but didn't do anything past a "basic" colour grade via unity's new post pack.. Ton's of stuff that can be upgraded / sorted.. Especially the shaders.!

@neoshaman

Some green grass / a couple of tree's with ivy meshes painted and some rocks doesn't exactly require magnitudes of artistic talent.. That is purely lighting / post and effects, nothing on this earth will convince me otherwise.. It's about as basic as it get's yet it still looks great.

@Adam Sowinski

It looks nothing like that in engine (gameplay), although as far as Unity goes it's one of the best..

 

@ShadowK

You underestimate composition greatly, composition don't need to be complex to be on point, and it don't need to have fine touch to be great. Reducing it to fog and ivy mesh painted is clearly not what I'm talking about, I think in artistic terms of plane, focus, line, balance, etc ...

The Ryse shoot IS good, I don't like this style for the reason frosted mention, but it's way above average. It has a clear focus, clearly delimited plane, it break the rythm smartly, the hierarchy works, etc ...

It's very important to separate personal taste from analysis.

 

In terms of post, I've been looking at Prism, wondering if anyone has experience with this?

Consolidating the post stack has been on my list forever, especially since I'm working on tight performance margins. The extra blits are definitely costly and there is a ton of redundancy in using 7 layers of post effect.

I've got a couple hand rolled post effects that I would need to integrate here as well, but I figure starting from a prism or a scion instead of stock unity post would save a huge amount of work and certainly way better results.

Any experiences?

 

I think it is confused in terms of narrative, which is why I think that it's objectively bad. Although it looks 'nice' the message it conveys is muddy and confused.

In the foreground there are a bunch of corpses and bloodstains, the light is tipped over, but the effect isn't carried out anywhere else and the carnage is limited to a very small area in the corner. Were these guards that just defended the throne room from invasion? If so, why aren't there other dead guys on other parts of the floor?

I'm not sure what the 'point' of that image is, other than to look nice. If you check the other examples, tarkov, homefront, etc, they have a very clear message and the composition of the scene and the message work hand in hand to communicate a clear feeling to the viewer.

Although the ryse shot has a pleasing visual composition, I think it fails to clearly communicate it's message. Although yes, it looks "grand" and sorta roman gothic, the message is muddy and noisy.

 

I thought the message was grandeur and decadence contrasting the glory with the massacre, anticipating the game story (the fall of the roman empire). There is a clear blue/red composition where the red is the base of a triangle composition with the status, certainly alluding to that glory is based on massacre. It's notable that the blue part is having a radiating composition with the light shaft that break it but still participate in this image of glory, with front and center is the statue bath in light, protected by two shadowy guard whose lance direction emphasize the statue. the massacre is literally frame by the ground that act as an imbalanced composition in opposition of the radiating composition, there is flame and it's oblique in a way that suggest shaky ground to the triangle composition.

So maybe it's too subtle, but it's very classic in its composition and reminiscent of renaissance painting with a very modern take!

 

You understimate the big picture greatly, I did a break down here: http://forum.unity3d.com/threads/how-to-make-an-aaa-game-in-unity-or-fail-badly.390947/page-17

Also look at this:

https://moritzweller.wordpress.com/2014/09/25/dissecting-koolas-ue4-archviz-magic/

In 3D games at least, the rendering pipeline has a massive impact on the overall look of everything. The only people who seem to "get that" is arch viz guru's and tech artists that have to work deeply on both sides of the window.

It's very important to seperate personal "opinion" from analysis.

 

The problem being that there is a bit too much incoherence in how it's presented. My guess is that the corpses were thrown in as an after thought at a late stage, after the artist had already chosen the camera angle and composed the basic scene. Because of the stairs and the shape of the environment, he couldn't really blend smoothly between the massacre and the grandeur to make the scene fully make sense.

The problem is that the gradient between the massacre and the grandeur is so sharp that the message loses strength.

There are many images of corpses leading to the throne, it is a classic image. I don't believe that this one was executed as well as it could have been. (although admittedly, it's still way, way better than I could do).

 

I never used Prism. As the cinematic effects by Unity get better and better I use less and less of Scion. Just comparing features on paper, Prism can do more (e.g. SSAO, fog, sun shafts). If it isn't urgent, I suggest to wait a few months and see how the cinematic effects from Unity progress. There was talk about combining them for increased performance once they are polished enough. So in the end, both paid assetstore solutions might become obsolete.

 

When I tried CryEngine years back, that thing was a mess. Dated editor, broken example project. Was deleted off my harddrive in minutes.

Now I most probably didn't spent the time I should have to really evaluate it, but I got the impression (also going by the fact that most big studios using CryEngine seemingly customizing it a lot) that CryEngine is brilliant if you have a team of dedicated engine and tools devs that hack that trainwreck into something usable for actual development (as opposed to "theoretically good engine if you manage to build your game with a spoon as your only tool").

Which ties into the big overarching thematics of the thread. If you are a big AAA studio with dedicated engine engineers, you can pick an engine based on its rendering power.
If you are a smaller, more constrained studio, you have different needs. Like a usable editor that allows you to work at max efficiency. A simple system to prototype new features quickly. Easy visualization of your data in the editor without having to write new tools all the time.

I gave Unreal Engine 4 a spin this winter, and while I'd say it is ahead of Unity 5 when it comes to runtime features for PC dev, its editor drove me mad multiple times. How the heck would I quickly throw together a charactercontroller and get it to run in the game? A quick on hour job in Unity, took me way more time in Unreal, granted I was "unlearning the Unity way of doing things" at the same time, still, as much as Unity gameObjects and Monobehaviours might hurt performance, I find it the quicker way to prototype a function before sinking the needed time into the thing to produce production ready code. At which point you might need to abandon the Unity way of doing things, granted.

There is also the Unreal way of doing things however. If you don't like the Blueprint visual scripting system, and having to create everything as a Blueprint, mix actual code with visual scripting, you don't need to apply.
Unless you are "working around the Unreal Engine 4 way of doing things"... which is easier given theres the source code. But if you have time to do that, you are either a big studio (thus could go with any engine under the sun really), or not really developing a game but wasting your time hacking an engine to do your bidding.


What I find more concerning is that bugs sometimes get not fixed for a long time, or that features take way, way, WAY longer than originally planned. I also find that Unity stretches itself too thin by supporting all these different platforms, but given that Unity started a Trend here, that is true of most of todays engines.
I specifically abandoned a smaller engine I migrated to after Unity 3.5 was proving rather limited in many areas, BECAUSE the dev of that engine started to move into different directions that where no longer adding to the original target... slowly the upgrades for devs targetting PCs were getting less and less. When nonsense like a mobile editor rose to the top of the roadmap, I knew it was time to jump ship.

This is not limited to Unity by far. Seeing how things are on the Unreal side, it looks different, but same. You are better off as high-end PC dev, and worse as mobile dev. You have more high-end features, but less choice (people are still waiting for a forward renderer last that I checked).
There roadmap is long, and some features get pushed back just as in Unitys case.


Unity should weed out the supported platforms and see which ones are REALLY asked for (I guess Win Phone, or Windows Apps could go). Concentrate on a few platforms that are REALLY in high demand, instead of trying to cater to every platform under the sun.

 

I'm a little concerned about the cinematic effects, specifically about working with them from code at runtime.

In order to not trigger a full rebuild on the tonemapper when changing saturation, you need to do something like:

Code (csharp):

1.  
2. var cg_copy = colorGrading;
3. cg_copy.basics.saturation = s;
4. m_ColorGrading = cg_copy;
5. material.SetVector("_HSV", new Vector4(colorGrading.basics.hue, colorGrading.basics.saturation, colorGrading.basics.value));
6.  

Otherwise it will crazy rebuilds each frame where the saturation changes and absolutely destroy performance. If I need to deal with this kind of thing for runtime changes to anything, it'll turn into a total nightmare. Admittedly, the results seem quite good - and the newest SSAO removes most of the weird artifacts I got in deferred.

But if I need to dig through all their code in order to make changes runtime it's going to be a nightmare.

 

We are not talking about the same thing I apologize if I come off as rude, I'm not very good at that. But anyway the discussion is good I think something positive can come out in crossing our misunderstanding.

First messing with the setting of an image render is not the same as messing with the composition, because you mess with the elements the image was composed with. For example most modern game, if you put their quality setting very low, ala "lowspecgamer" youtube channel", look way worse than a ps2 game. The reason being they used different tool to achieve composition, ps2 game use to bake lighting into vertex colors, removing texture mess with lighting composition in modern game. Lighting composition is tech agnostic, if you can't have it with a technique you use another, ie you compose.

In that same thread that's why I purposefully used WII game (to negate the tech impact) to show how composition matter. IMHO Mario galaxy is one of the best composed game artistically, I still confuse at how they achieve such finesse of composition out of few things (the no man's sky's artist also had a shout out to them). I'm not talking about the style, or the color, just how every element relate to each other and balance out composition in a way that's almost surnatural. I'm not sure you understand that part.

That's where we diverge, in my opinion, you confuse the tools used for composition with composition itself. Let's say it's an opinion. I will be glad to be false, I'm trying to point to a valuable direction, not try to make people look like a fool, I hope it's not the feeling I'm conveying.

@frosted
I don't invalidate what you say, I'm just saying the composition is effectively highly skilled and that participate of what the image is good. We can always nitpick and do thing differently, I know I cannot do a composition this good either (I suck at colors and luminosity balance, I always overdo lol)

 

Don't they generate a LUT upon parameter change, to then have better performance in future frames? I think it's just not meant to be changed at runtime. You should try baking your desaturation settings into a separate LUT, use that under the LUT option of the tonemapper, and see if tweaking the LUT contribution slider at runtime still triggers the rebuild (I don't know).

 

I don't "confuse" anything, it's a collective summation of everything that equals end result.. From mesh flow to level design to texturing to lighting to post etc. etc. etc. In a 3D games overall "composition" (I'll use this loosely as it can mean many things) is relative to the technology used and artwork as opposed to something like a 2D game which is purely artwork composites.

If you don't understand that you will struggle (that's why super mario is confusing you), how do you think they made the Adam demo for example? It's a mix of everything (including tech) built from the ground up to display that style. Unity and all these "AAA" companies don't spend time and money researching / implementing this stuff for a laugh. It all serves a purpose whether it's atmospheric / compositive / blended etc. until in terms of "realism" or "hyper realism" for e.g. we can all use brigade.

Lighting tech is "agnostic" is it, well turn it all off (including shadow mapping, HDR IBL etc.) and see if you can make something better then the Adam demo. Silly statement.!

Not that I agree that mario galaxy is "well composed" compared to say that escape from tarkov picture, which covers major key elements.. But y'know we're back at opinions here.

The whole point is, that CryEngine "tree" scene should be relatively simple to replicate in Unity with some more stuff setting the mood.. Ultimatley it's the user / artist(s) that define it but many even screw up that bit.

 

This is a nice and inspiring article:

http://80.lv/articles/student-feature-ue4-interactive-scene-in-20-weeks/

 

Well let's agree to disagree, the way you comment on mario tell me you don't understand fundamentally what I'm saying, we are not having the same reference.

Composition is not relative to technology at all, resulting image quality is (as in level of details and rendering). Mario is not confusing because of tech, but because the artist is mighty skilled do a lot with a minimum.

I'm saying, focusing on composition will lead you further than "image quality". I'll refer you to this: http://artisaverb.info/GDC2013.html

Look at this

Nothing you cannot do technically, yet this is master class at composition, placement of stuffs is incredible. Even though many stuffs are actually copy pasted.

I'm talking about the stuff you learn there:


It doesn't matter if it's hitech or black and white image, good composition is mighty difficult and completely overturn an image.

It doesn't matter if you are limited to 2048ppoly max or million like in modern game


It give richness to the image beyond what tech can provide you, just by the careful uses of every elements. No matter what the edge flows is, you make the most of what you have, no mater how little it is or flawed they are in isolation.

I'm moving out the discussion if there is no new element because I tend to get stuck in circle, we had enough iteration already I feel like I'm derailing

Let's just make stuff we believe are good, I was trying to highlight a particular point. It actually does not invalidate other points, I just feel this part is grossly underestimate.

 

I actually think @neoshaman brings up a good point. Especially for those of us on a very tight budget. Good composition is much less expensive than high quality assets, and can really help to close the gap. Also, if you are trying to build a strong "mood" composition is a very important tool. Although I don't like the Ryse shot, I think they did a better job of creating a mood and theme than I gave credit for.

 

I won't say it's just about "mood", composition make an image solid, period! I'm not saying tech is useless but without good composition fancy effect will lead you nowhere! They work together!

To illustrate a last time, let's look at a ps2 game (32mo of ram, 4 of vram, ~300mhz), ie that could not have the fancy effect we have in modern game. Now how many basic unity user can get the quality of a single scene there? And we have massively more power!

Once you have the composition down you can move on details and image quality.

(emulated graphics for resolution)

Spoiler: more image example

 

The majority don't have level design and composition skills.A lot just copy what they see in other games.
There is lot of research sketching or general level shapes study and theme work people don't do.
If you don't have inspiration or try to innovate you will just take ideas from what you played or found in other games.