Stay with unity or move on?

Hello,

I have a unity project which gets more mature.
However any amount of work does not seem to give good quality light and sharp scenes.

I now searched, I feel a bit like the lights were drawn by a little child sometimes.
Even the demonstration scenes and pictures look "flat" the light seems to ignore depth.
This feels like my project with unity can never look "professional" at some point i just wonder if i have to and can accept that.

I found a good example:

https://forum.unity.com/threads/pho...ing-between-unity-and-unreal-4.262181/page-16

In the top entry the Unity image on the left looks like you draw some light on a unstructured surface it looks like a Flat image on a Flat background.

The entry in the right looks "deeper" it looks like a real 3d thing.

I am very sad that the Unity staff decided to close the thread i assume its because of the bad publicity it brings to see so directly how much advanced the graphics of ue4 are.

I now wonder if i should rewrite my 10000+ lines of code in the longterm and try to migrate the project away from unity.
My project is more or less c# code which does control the unity interface with the help of some prefabs.

 

They closed that thread because it is a years old necro thread with outdated content. A lot of stock Unreal's advantage in its appearance is the by default post processing effects. Since that thread Unity has developed their own post processing solution, which is freely available but something you have to actively add to your project instead of being built into the engine. Also take a look at the new rendering pipeline options being included in the 2018 versions of Unity.

 

I would not migrate a project between engines if you are already pretty far into the project. A more effective solution would be to finish the current project with the current engine, and then investigate all available engines before starting the next project.

 

I cant do realistic lighting either.
That doesn't mean I blame Unity.
Ive seen Unity demos that look great.
Unity is capable of it, Its just probably not "drag and drop" easy.
Keep at it, especially if you have a sizable chunk already done.
Look into shaders though, learn how to program them, I think this is your bottleneck

 

It has nothing to do with publicity and everything to do with the fact that the thread is seriously out of date. You can very easily find newer threads if you take the time to search a little bit further (and tell Google to stop showing results from a decade ago).

https://forum.unity.com/threads/just-for-fun-guess-the-engine.434941/
https://forum.unity.com/threads/how-to-make-aaa-graphics-in-unity-enlighten.471528/

Additionally check out the most recent demos.

https://unity3d.com/book-of-the-dead
https://unity.com/madewith/adam

Here are some impressive third party assets made for Unity.

https://forum.unity.com/threads/rome-truly-aaa-quality-exterior-enviroment-pack-for-unity.389167/
https://forum.unity.com/threads/hd-...ing-box-2-7-2-next-gen-lighting-solut.475644/

Prepare yourself. You may have to accept an even worse reality. You may discover that even with the ideal game engine your game may still not look "professional". Realistically the game engine can only assist so much before it falls on your ability to make use of it.

 

In my brief experience, making your art look good is 1000x easier in Unreal. You create shaders by connecting noodles. Anybody can do it. There is also a ton of important editing you can do in engine, which is much more convenient than unity.

But I wouldn't switch engines mid-project unless I definitely had strong enough art to warrant the time it would take to redo all of your work. So basically, if you don't have AAA assets that demand a game engine finely tuned for AAA graphics, it's probably not worth the time.

 

https://blogs.unity3d.com/2018/02/2...raph-build-your-shaders-with-a-visual-editor/

 

I wholeheartedly recommend the OP downloads other engines and does tests etc. There is nothing stopping everyone doing this. Please go ahead and use other engines in addition to Unity.

Meanwhile I'll be using HDRP in Unity:

 

Nice. I am definitely looking forward to 2018. But, all things considered, Unreal has been doing it longer, and thus I'd expect a higher degree of polish.

I don't have enough experience/knowledge to know much about the engines beyond using them to display art assets, so the best judgement I can make is by looking at what the majority output of games from each engine are. Unreal definitely takes the cake for AAA graphics blockblusters, while Unity is popular for pretty much everything else. So I guess c# is a simpler/more user friendly language? Or Unity is more developed for the mobile platforms?

Also, there is a package on the asset store called Tropical Forest Pack. It's got a handful of really nice scanned environment models, but in the included demo scene the author has set up some better lighting and also provides a video demo to show how to do so for your own scene. I don't have any screenshots ready, but you get some super high quality assets like that with the linear deferred lighting, along with a few PP effects, and it looks pretty damn good. And that's just a noob (me) poking around.
I think some key things that are easy to do in Unreal but aren't easy for a noob to find in Unity is volumetric fog, scattering, stuff like that. But it looks like Unity is demo-ing something the same or similar with that video posted above.

Whatever engine you are working in, I think much of what we call "good graphics" is really just a skill in illusion. Like with characters, the more bells and whistles you slap on there, the more detailed it appears. You just give too much S*** for the eye to see, and then the eye doesn't notice mistakes. Check out characters for any modern AAA games. Like, who on earth would run around with so much S*** dangling from themselves?
Same thing with environments. The more doodads you squeeze in there, the more atmospheric noise, the more depth of field, it all actually works to kind of cover things up and fade out all the little CG give-aways.

 

One major downside to that approach is that Unity doesn't require a logo with the large studio games.

 

Everyone should move on then come back when they know what I know.

 

Yeah, seems kind of backwards to me. But I think I am mostly up to date with the best games coming from both engines.

 

For starters you don't post any screenshots of your game, which being the cynical bastard that I am makes me wonder.

Unreal specializes in visuals and latest gen rendering approaches. That's their thing. If you didn't understand that and make the right decision in that area for your game, that's on you entirely. What you can do with Unity at any point in time along it's development cycle is pretty easy to find out.

In the end engines are what they are at a point in time. Complaining that an engine doesn't do something you need, outside of say bugs, is pointless. It won't change anything. Complaining about where Unity put's it's development focus, that's a different thing and completely valid IMO. But that has nothing to do with your current game, which has to work within the context of what can the engine do right now.

One last thought is put this in context. The differences between Unreal and Unity lighting is real, but the chances it will have any impact on the success of your game is basically nil. You aren't competing head to head with multi million dollar productions, and if you think you are, you have bigger problems.

 

I hear the grass is always greener in the other engine.

 

And Unity's, if anyone's been looking at HDRP.

 

General complaints about graphics and lighting quality are pretty meaningless at this point really. eg if you cant describe specifically what aspects of lighting are lacking using the appropriate terminology, then its hard to have a meaningful discussion about it. And it leads me into the temptation that people not understanding how to get the best out of the tools on offer is a factor. Not that I am entirely blaming the developer for that sort of situation, it can be an indication of poor templates, starting points, documentation, samples etc by the engine provider.

Things are improving in Unity on these fronts anyway, both in terms of what can actually be achieved using the new HD render pipeline, and the scene templates and demo assets like fontainebleau. Not that demo projects that show what is graphically do-able in Unity are anything like a new phenomenon. Sometimes when I see people complaining about the results they've gotten with their own projects I wonder if they have studied some of these demo scenes for techniques used and resulting quality.

 

Elbows i try to describe some actual things i have seen.

i have seen the screenshot from this Video and thats one of the points where i feel thats closest to photorealism unity can get but in my experience it feels not real there are actual mistakes. I dont want to sound harsh towards unity but i wish it could do a bit better in some areas. Alswo i dont want to be harsh towards the demo but i think they are the best there is. Take following examples for "wrongness" from the demo. I did watch them with audio off.

Here the gras does move unnatural. It does not bend under the wind just move left and right.

Here i wonder long this image looks flat in some way. Thinking about it the Light does not scatter enough so much light should reflect and illuminate the darker woods more. I cannot really say if the problem is really that or there is something else too but it feels flat a bit.

At 58 look at the back of the trees when the sun is moving they stay exactly the same and it feels when the sun moves the light reflection should reflect on the trees and darker parts of the wood and this reflection should move on the back of the trees and create light patterns on it when the sun moves.

The tree movement seems random in a way and the branches and leaves do not bend.
And some seconds later the light does again not reflect into the wood.




Comparing it with the adam demo in that one i did not see that issues so strong but because it does not feel like putting so much effort in photo realism.

Still Adam has that issues @ 3:40 the gras does not bend.

And in the whole time inside from the start the light does not reflect properly so only the back is illuminated by the light and the front seems to have a constant light.

I think light is one of the biggest issues with graphics unity has.

Is there a way / are there settings to increase reflected diffused light from a surface?

Is there a way / are there settings to increase leaked light from a a surface hitting light to spread more to the surrounding environment to make softer shadows? I tried it with soft shadows and decreased the shadow sharpness but still it does not seem to mix properly with other lights on objects if i have two different lightsources from different angles it seems to get a sharp line between.

 

Grass is also a bit greener in other engines.


Joking side, I think you'll experience similar problems elsewhere because it's your lack of experience with the engine, I'm sad to say. Note: this is because I haven't seen your work and you haven't apparently asked for any graphics assistance so I could just be saying really silly things

 

I believe its quite well known that the closer you get towards realism in some areas, the more the brain may notice issues that spoil the illusion. This probably gets most attention when it comes to realistic humans and the 'uncanny valley' stuff but it shows up elsewhere too.

Honestly, even though our ability to do more things in a genuine 'accurate' way in realtime improves over the years, doesnt mean any less skill is needed to composite everything into the perfect scene. If anything, more skills are required, especially with the art assets and the lighting. And its a time of great change on the Unity front in terms of the scriptable render pipelines and how the HD one evolves over time. Now is a great time to start experimenting with this stuff and some of the assets that are already available for it. There are things missing, but in the meantime you can still learn how to harness what is there already properly and become familiar with how to set these things up and what settings are available. It's very exciting to me, about as exciting as checking out a completely different engine. Which is certainly something I recommend too, really get into the detail of where other engines are at and what aspects of their approach to graphics yields particular impressive visual results.

 

Well i have indeed my own battle with (the) light but i did look a bit around to what to expect what i can work towards thats how i came up to look into the demos and pictures that are floating around and illuminated scenes from unity always felt "flat" so i came to wonder why and now i think i know why because the light only comes from one side so to say.

But that essentially sucks it then more like oh if that the best can be maybe i just take only a global illumination and skip any point or other lights...

Just let me ask this way:

If i would want the light to illuminate and spread more into the forest and reflect on the trees so when the sun would move afterwards the backside of the trees would change in light in this one:

How would i do that?

 

I must admit i dont know much about the new features. I dont stay informed about new things.

 

I have here one example of how i tried to play with light and it ended with a strange look atleast for my taste. The first lights have a high intensity around 3 and the second has a lower around 0.8
However the background of the first image seems to take light less.

 

Neither in Unreal or Unity is there a single lighting solution for perfectly realistic lighting. Also, I have never seen grass that bends as separate blades. That would require a ridiculous amount of geometry and/or super fancy shader work. I don't think that exist yet. And who cares?

I don't have much experience with Unity lighting, but in Unreal you typically use a combination of lights to get the best looking scene. It's not just one light and global illum turned on = perfect realism. At the minimum you use an ambiet light, a direction light, global illum, and lots of point/area lights for specifics.

And, IMHO, looking cool is way more important than looking realistic. Who evens know all the intricacies of realistic lighting anyways besides tech artist? Certainly not the average game consumer. We want pretty/interesting/awesome as gamers, and as artist we want control, options, and convenience.

In your own example above, why waste time fretting over the engine when you don't even understand 10% of it? Just put in another light to get what you need, or read the documentation, or read some lighting tutorials. You can definitely achieve what you need in Unity, I have no doubt about that.

 

Book of the Dead is a nice work of art and looks pretty good. But if you put some wax in your ears, take a single screenshot and compare it to a range of stuff from polycount and artstation, I would say it's not as much of a leap forward as some might think, and doesn't compete all that well in a variety of ways. And personally, like the Adam demo, there's something that just doesn't feel right to me for games in the way it looks - something that feels more like a low-budget movie renderer than the kind of crisp, vibrant and well-defined visual aesthetic of something like Battlefront.

For me, I will be making a proof of concept of the game I'm making in Unity, and maybe doing one in the Unnameable Engine as well, and depending on the results I'll decide what to do. I think this is what anyone should do when trying to decide (and not just visuals because although another engine might look better, likely there are other things about it that aren't so good).

I downloaded the latest Post Processing Stack v2.0 post effects last week and spent a bit of time having fun with the graphics. So far it looks good enough for me to warrant just sticking with Unity based on all the other things I like about it, but being a very visual-focused dev, if there's something better I don't know if I can resist chasing it.

 

Rome is a really nice Asset, my problem with high quality Unity scenes like Rome is the performance. I quickly recorded this clip with the demo scene. Its recorded on a Ryzen 2700x OC to 4.15 all cores and a 1080 TI FE@2.2 / 6 ghz, 3200 CL 14 memories. Shadowplay takes a few fps but its basically the same result without it. Its a relative high framerate though it should be much higher on this hardware and this simple scene with no open spaces. But even with these relative high frame rates the game still feels pretty "laggy". It microstutters. I have never experienced this on other engines.

 

I would love to see the game running that on a PS4 Pro at 4k res at 30fps. Can I see it please?

The point I'm making is... it is a leap and you're wrong.

 

Each engine has it's own architecture.

Don't ask a response, take some days to try UE4 and C++, and make your choice.
(C++ won't be as fluent and easy as C# and you'll have a learning curve about code and toolset.)

 

Btw, it seems it uses many cores, I wonder if Ghraphics jobs are turned on

 

Fair point, in context it might be quite a magnificent thing. It's pretty magical how Unity get their stuff to work so well on a huge range of platforms.

However in absolute terms, I don't think those graphics are at the point where I would say "Unity is right up there with the best of them". Something about it still comes across as quite lossy to me.

Don't get me wrong, I like Unity. I wish I could just move on to something more pretty but it's so damn convenient in a lot of ways. If only there was a way to have the best of both worlds ...

 

Curious - was occlusion culling properly set up for this scene? Your mention of no open spaces made me wonder, since occlusion culling should have an immense effect on this.

 

I agree, I have vastly more complex levels and it thunders along well above 90 at 2440p with 980 card. I think there is much everyone can do to optimise scenes, and I think we're entering a territory where it may be important to have staff to optimise, because it is a full time job to get something running very efficiently the way unity engine responds best to.

Things like automatic lod on the blog gave me hope Unity would consider this being just as important as a core engine optimisation but I don't think they quite see it like that yet. But as fidelity grows, so must the need to optimise. An engine is only as fast as it's least efficient part.

 

I might be misinterpreting - but is book of the dead running on a PS4 Pro at 4k 30fps? I tend to ignore tech demos that can't be run locally and/or don't list hardware up front and I never saw a note on what hardware book of the dead was running on.

 

I'm sure Unity will be more forthcoming in days to come, It's not my place, sorry. But when you do see information publically on blog or otherwise make note if it's PS4 or PS4 Pro, regardless.

Generally, Unity will be aiming for demonstrating things at native resolution as I understand it, otherwise it is illogical.

 

This is the build from their homepage I don't own the asset. Pretty sure Peter knows about occlusion culling though I must say umbra is really bad at ocluding alot of times

 

I've noticed that occlusion culling can also be really heavy, although I have mostly tested in outdoor scenes and this is often grass culling overhead related to the kinda crazy inefficient native grass implementation.

 

A lot of people use umbra wrong. It's not really for occluding a billion little pieces, it's for occluding chunks. Lots-of-things should be grouped into culling groups and this will work with umbra as well.

Having said that, I think there's problems preparing assets for Unity to digest. Automatic LOD on blog was a nice start but it didn't seem to go anywhere. In addition to that, there should be much better combining workflows built in and I honestly am not seeing why it can't cover common open world use cases practically everyone wants.

 

Unity is not geared towards big worlds as most indies won't make those type of games, i guess that must be why such features are not priority while the new job system is on higher priority because it's usable for any game type.

HLOD for example combines models for you and generated the atlas texture.
https://docs.unrealengine.com/en-US/Engine/HLOD/Overview
I wish Unity would have the same built in.

 

HD is geared toward big worlds, sorry to correct you. The default shadow range is 1000 meters and the geometry is transformed to the camera space at 0,0,0 to improve precision for huge distances.

So yes, Unity is designed toward big worlds. The idea is big is the only problem. Solve that and small takes care of itself. In addition, big does not equal lots of content, many indies make use of the space with financial success.

It's true that in the past, Unity struggled but we've moved on from that now, and new problems are emerging - how to feed an advanced and hungry engine without choking it.

 

Quite a lot of old perceptions and presumptions need deprecating now for sure. Much progress has been ongoing for some years now but the 2018 cycle really should ram the points home in various ways.

 

Yeah I do what I can to dispel old misconceptions without coming across as a fanboy but also maybe moan a little as well so Unity can hear what it's like in this grungy sector of game dev.

 

You didn't get the point.
Big world is about streaming system, origin shift, terrain tiled, open world manager tool etc ...
For example Zelda Breath of the Wild.
HLOD is one feature usefull for open worlds and performance. Unity is not pushing in that direction, perhaps later.

 

Unity isn't pushing in that direction, not because they're focusing on indies, but because these are extremely specific tools that people can implement themselves if they need them. Even most major games don't need origin shifting and the like.

 

I doubt that game uses any kind of origin shifting. You can go 10 miles before needing origin shifting with the right renderer, HD so happens to be able to handle that with a stable image. 10 miles is beyond big. You can bet AAA fudge the world unit numbers big time. Origin shifting is a last resort for any game. I don't doubt some do it, but I bet it's considerably less than you're thinking.

In any case, if you can go 10 miles, that's an asset and authoring problem, which backs up my point. We hit this problem before the limits of a game engine without origin shifting.

 

Zelda BOT is very huge, i'm not sure you would cover it without origin shifting.

There is many plugins trying to improve Unity , for example Vegetation Studio, because Unity lacks optimization out of the box.

Anyway we are away from the topic main discussion

 

Well, origin shifting is only one of the things that was mentioned.

Stuff like streaming and tools which are actually designed to support open-world-friendly workflows are still a thing Unity lacks. Agreed, you can roll your own tools / workflows / systems for those things (as I'm doing), and smoke and mirrors get you a heck of a long way in terms of presentation, but I do agree that there's a lot Unity could be doing here to make this style of project easier.

That said, this style of project is indeed getting easier. For starters, multi-scene editing is fantastic and pretty much enabled the approach we're taking now.

Back on the topic of origin shifting, 10 miles isn't necessarily "an asset and authoring problem" depending on the style of project. For a modern day "AAA" game, sure, you need to densely pack your environment and 10 miles of that is a budget of millions. But, say, a driving sim on realistic roads you could automatically generate from map data? With the right tools you could blow 10 miles out of the water, and it's not rendering problems you'll have first - it's physics.

None of that is to knock Unity. I expect to make custom tools and workflows for any non-trivial project, and it's more than fair to say that this project wouldn't be where it is without Unity. But this is a popular style of game, there are tools that let smaller teams achieve more of it, and I think it's valid feedback to tell Unity where they could help us more for particular types of projects.

 

TL;DR;

Stay with unity

all other engines are lame

 

So your are telling me static props are not meant to be culled with umbra but that we need to create our own system with the Culling group api? Never seen a official post about this from UT

 

No, he was indirectly responding to me saying culling had a lot of overhead and referencing grass. He's partially right, I misspoke when I said "occlusion culling" has a lot of overhead, I should have said "the culling process has a lot of overhead".

Native unity grass can have a lot of overhead rebuilding meshes (once density passes certain thresholds) on camera rotation, this isn't occlusion culling - it's frustum culling I believe, and is not connected to Umbra.

It's not possible to use umbra to cull grass or anything equally tiny, so he's kinda right that umbra would be bad for this.

 

Ah yeah, I would stay away from vanilla foliage all together, better to use vegetation studio.

 

Agreed. VS is quality work.