A different perspective on Unity development.!

@Billy4184

That's the kind of thing I'm thinking ..

 

Regarding trees, have you considered doing what was done in old titles (particularly, racing games such as Colin McRae 2.0) and just make textured wall of trees in a distance in a place player isn't likely to get to (because terrain obstacles would block him before jig was up)? This would give you an illusion of huge forest without not too many trees.

 

@GarBenjamin Speaking of which.., you should definitely check out game called sacrifice. There is a design-related article explaining their design process somewhere on the web too. The game had very bizzare looking enemies, however they made sense within the game world.

 

Plants also follow pattern based on moisture, temperature and altitudes (maybe undergroundness too), you can use that to justify their design OR to pick real life plants coherent with their environment.



Define rules per biomes and you are good. You can create your own biome using your own parameter.

Also in art the Rules of thumb is to classify plants in grass, bush and tree size, that help a lot organizing landscape, with always, for each, the dull background model (ex plain grass), the occasional foreground model that have a bit more details (generically flowers) and use them to break the pattern then the landmark "unique" (or rare) object that give identity to the place. Landmark can also be just a place or plain void (ie in a forest). Look at the new zelda wii U for a strict used of these rules.


Look at how there is small tree repeated, larger tree as break up (not really seen here) and a different landmark tree in a rock that steal attention. The bush and flower add accent without really calling attention to themselves and break the grass repetition.


Same rules applied to a depressed clearing (there is a hidden sub boss fight there), tree are using the same mesh but attention is rarely drawn toward them, they frame important part. Instead of using the same mesh tree you can use similar trees to add variation, but not so far apart they look distinct.

People tend to remember landmark more with a general feel of the place, background element frequency are used to have that feel.

Now this is just the basic, you can safely try to break the rules and add details when needed.

 

That's a really useful graphic, I'll have to keep a copy of this around.

 

What about randomly associate word ideas with the rpg game ? or mix different gameplay mechanics ? Some times it can bring new concepts. There is many RPG mechanics and we could find lot of new ones if we try random new associations.

 

Talking specifically about game mechanics, then yes by all means learn from history and work with what works. Mix up different mechanics to come up with something new. Some things work, some don't, and experimentation is always good.

As far as my post, I was speaking mostly of the thematic elements of a game rather than the mechanical.

 

There was some action or RPG games using cards, or puzzle games using RPG elements , another rpg with combat is some mini shoot them up game.Or turne based RPG mixed with real time actions and triggers.
There is lot of space for innovation when people try new things that stays interesting.

 

Funnily enough one of the biggest things I enjoyed about Witcher 3 was gwent , I might outsource a 2D game that integrates for funzies.!

 

This post is very very useful, if you have more thoughts and examples on this, please share.

 

You could even make successfull mobile games, but perhaps you won't like making mobile games because the complexity wiil be too low for you

 

@Billy4184

This is what Unity looks like fully custom, new renderer / shadows / shaders / post (bar their new tonemapper (actually like that). Well this is what it looks like to the best of my abilities and to prove I'm not "cheating" (like I just did it in UE4 ).. I'll post the in-editor screeny as well.!



Edit: I got the Time of Day system working YAAY!..




In Editor:

 

That's a lot of ping pong ball on the floor lol

 

Totally reminds me of the gallimimus flocking scene in Jurassic Park (from the "Making Of" DVD supplement) with the little tracking balls in the grass before they digitally removed them.

TANGENT.

 

I wonder why Unity doesn't just hire you @ShadowK to work on their rendering systems? It seems like if you can do this kind of thing "from the outside" using only the limited things you have available to work with... if you actually could work on the inside right in the source... who knows what you could do and be better performance probably. In this case just maybe next year's Unity would equal or surpass UE visually.

 

They have far better people than me chief, like good ol' Sebastian. There's plenty of people they should hire, like Aieth / lars / sonic ether although I'm not sure how much "say" they'd get in actually implementing stuff.

It does make me wonder why Unity isn't a cutting edge, performant graphical powerhouse right out the box some times (in the same sort of way I wonder what I'd like to eat for lunch )? I know the guys from Alloy have been giving them hints for a long time, although again Unity already have some very talented people.!

 

@ShadowK what was the performance on that last scene? I am drooling here with this grass density

 

Around 35 FPS on a 390, although it's all fairly unoptimised as it was just a sling together to show Billy. I recon I can get 50 - 60 out of it..

 

Hmm i really wished that the fps counter would be a bit higher

I will need some exteriors for my current project, and to tell you the truth i am kind of lost lately with how someone should approach it with unity at the moment. The forum is filled with complaints about speed trees performance and billboards issues, speedtrees not getting batched if you have wind on, i haven't figured out yet if gpu instancing is something that will help with vegetation in the short term (though i have seen that the devs are working on a new vegetation system and shaders), baking lighting on terrain with trees being almost impossible, issues with mixed lighting and many other things that i cant remember.
Also comes to no surprise to me that both of the last two demo/movies from unity use neither unity's terrain nor the terrains vegetation system...
It seems that at the moment an established workflow for making a terrain with vegetation that looks and runs decently doesn't really exist. Am i right or have i missed something fundamental here?

 

Thing is, the most densely foliage populated game I know of is Witcher 3 and that only runs @ 45FPS on my 390 w/ max settings. Those screenies have a far longer foliage draw distance and they also have a lot more heavy post (including real time GI which W3 doesn't use).. Hell CryEngine will only run a small scene at 50 FPS when densely populated..!

If I could get it to run at 50 FPS then I'm already ahead, games just aren't made like those screenies.. There will have to be a lot of compromise and trickery to get it running on it's lowest common factor. Using meshes to bridge gaps between foliage and terrain helps a lot also occluding as much as possible, using culling distances etc.

GPU instancing will help greatly, but you still have to be smart when designing your terrain to maximise it's efficiency.

I've not had massive issues with SpeedTree performance TBH.!

 

Looks pretty first class to me! Very much appreciated and I'm sure a lot of people here feel the same way. There's nothing quite like seeing a glimpse of what's possible from someone who knows what they're doing.

Can you describe in general terms what you did with the custom rendering, and what do you think made the most difference to the quality out of all the stuff you did?

ps the thing I most like about it is how crisp the grass looks - it's not washed out and flat the way it usually seems to be.

 

Star Wars Battlefront has some pretty amazing and dense trees/vegetation too, and from what i have gathered the performance is great and quite scalable to lesser hardware.



Btw what i was referring to was mostly for unity's stock capabilities. It would be interesting to me if you have the time to port the same scene in stock unity to compare visuals and performance

 

@Billy4184

I'll do a breakdown once I'm done pratting about..

@moure

I'm not sure if you can notice too much, but I cut down the quality settings / baked umbra and made a bunch of tweaks here and there.. This is how it looks @ around 55 - 60 FPS.! Biggest issues I'm coming across with the dense amount of foliage is the G-Buffer and amount of shadow casters so I've had to really cut back on it.!

 

Because what he's developing is edge case solution suitable for one scenario.

Unity wants their rendering pipeline to be as general-purpose as possible. Take a look into their cginc files, and you'll find a LOT of ifdefs/workarounds for specific platforms/hardware there.

for example, AutoLight.cginc has this entertaining comment:

 

Did you implement that using surface shader + custom light models or do you have custom rendering routine (meaning you gather visible light and run passes by yourself)?

 

@neginfinity they could always hire him to put in a few more ifdefs for the PC devs! Looks like they're working on it though, what with the Adam demo and all that ..

 

Ideally it would make much more sense to aim at REMOVING ifdefs (and not adding more of them) and making code cleaner.

 

The renderer is nothing really bar an attempt to try and increase performance, I switched back between Unity's standard / mine and I really couldn't tell that much difference. Yes it's just a custom light system using inverse square decay etc. (not all the shaders are surface though, some are vert / frag) (it's still using GGX BRDF). Unity's "standard" shader doesn't look quite right to me, so I just replaced it.

It's nothing particularily fancy.!

What really helps is a decent AO (trying SSDO) http://kayru.org/articles/dssdo/ / Bloom / CA (Fringe) / IBL (Unity standard (gradient based) w/ SSS on foliage and raymarched volumetric scattering (All the grass is in the tree section w/ mat applied). I'm not saying my implementations are the best (bits need fixing). I also want to try out Sonic Ether's Bloom and the TOD system is Tenkoku.. I'm going shopping to see if I can make it better.!

Oh and I actually founds some tips on Enlighten, not half bad.. I'm only using 1 Texel res (real-time) and the results are actually pretty good.!

Really need a better DOF solution!.

Then finally a lot of tweaking.!

 

Yeah I guess that is what I was thinking. Even if they don't hire you or specifically use your work... if you are able to knock out a shader that looks better than the standard shader... I don't quite understand why not include that kind of thing into stock Unity. Even if they don't want to replace their standard shader. Just offer an option Standard Shader 2 and it'd be basically your replacement shader.

EDIT: Or at least you can put it up on the asset store charge $3 for it "ShadowK's Standard Shader Replacement" with a couple screenshots. A scene using the standard shader. Same scene same exact viewpoint using your shader. That way it is available for other people to use and at the same time you get something back for the effort & time you spent both creating it and setting up the asset store listing to share it.

 

I'll probably just dump the framework as a package somewhere when I'm done with the game, they are very "select" screenshots and there's a platitude of issues at the moment. Even if I charged $100.00 it wouldn't be worth the time to support it.!

Anyway, I've gotten what I need out of it. I'll pop in from time to time, but it's about time I got off these forums and started tackling the mammoth task in front of me all in.

Wish you all the best of luck, this time I am going to resist ..!

 

Tenkoku is pretty insane. The fog is stunning and the overall effects are top notch. It absolutely destroys performance on low end hardware. I just posted a bunch of notes on performance improvement in his forum thread.

It's absolutely beautiful, but man does it need a serious optimization pass. I drop it in an empty scene with terrain and i'm hovering around 50 fps already.

The sacrifices we make for the things we love.

 

3$ , you are so greedy

 

Once upon a time I wanted to integrate order-independent transparency into unity. Then I hit a big issue with directional lights, outdoor scenes, and another issue that prevented the framework from working on non-DirectX renderer.

In general, I have impression that asset store is the most suitable for non-coding assets while coding and shader ones often aren't worth the hassle.

 

The vast majority of the asset store code I've seen is quite scary, this very much includes some very popular, well reviewed and widely used assets. Like... woah crazy bad.

 

Yeah I understand that. The aftercare for things can get very expensive! And being free people shouldn't have any expectation from you to support it (although there are always some who seem to lol).

So does this mean you are actually using Unity now?

Good luck man!

 

Really?! $3 for a shader that can end some of the frustration people have with graphics quality in Unity is being greedy? lol

Or are you being sarcastic? I guess maybe.

 

Between pure content (graphics, sounds, music) and coding-based assets... yeah I'd expect the coding-based to definitely be the assets requiring the most ongoing support. I think as long as you priced it accordingly it would work. Either super cheap like $3 and say "as is... don't expect support at this ridiculously low price... that's why I priced it so low in the first place" or else charge a premium price to cover all of the help desk time over the coming months and years.

I guess with the super low price you could always create a FAQ over time handling the most common questions. So you'd be providing some kind of support just one with minimal time and effort.

 

With a license that allows it to be integrated into open source engines like Banshee? Off the top of my head that's one of the big advantages to giving away the framework with a permissive license.

 

Good to know. Been thinking about using it for my new game.

 

Do you mind roughly describing what you didn't like in the standard shader and what is different (roughly) in your implementation? Would be interesting to read.

 

Ugh! I need to stop reading these threads. My wish list is starting to grow out of control.

 

To be clear - i'd still recommend it. I went through a good amount of the code and pointed out some fairly easy to fix problems (most were just legacy things that never got cleaned up), and I'm sure he's going to include a good chunk in the next update. I hacked my own copy and got very good performance out of it (although I had to cut some of the low clouds until proper fix).

But man, that sky... totally worth it.

 

When you're eyeballing stuff this whole thing get's very tricky, it could be nothing more than how light decay is calculated which adds a lot to "physical correctness". I've not spent practically any time looking at Unity's shaders, by now I just by default replace anything and everything. Although I did do some testing, using constant parameters in a UE mat and properties on Unity's standard shader.

What I do know is every implementation of models like GGX is different in every engine, from UE4 to Unity all the way over to frostbite. Each with different approximations and trade off's, it would simply take me too long to describe all the differences.. Although, I'd check out the specular D modification in this doc if you get my drift ..

https://cdn2.unrealengine.com/Resources/files/2013SiggraphPresentationsNotes-26915738.pdf

Personally, I'd forget deep diving into this.. I made the shader code for that scene in Unity 4.X even before Lux came out.. If you're aiming for graphical "powerhouse" just use Alloy, I've had a quick look and it's far in front of what I've done so I'll probably just use that.

@Ryiah

Probably under MIT or something..

@frosted

Yeah, there's quite a few things.. Stuff left in late update, reflection probes don't need to be tied to a framerate and constantly cycled (that's why you keep getting CPU spikes). Unity automatically updates it's own reflection probes anyway and when I tested it at least in editor it didn't make any difference (play mode on a X1000 cycle)..

But apart from the "red" cycle, the gradient curves are bang on for IBL. The fog is nice even though I might change the way it falls off.

Bit rough around the edges but it's cool, not had much luck with suimono yet but I think I need to RTFM.

 

Ah, so that's what Youtube kids in comments section mean when they go around spamming 'Unity has fake PBR!' They are just referring to a badly adjusted/calibrated light falloff parameter.

But I wouldn't say it is exclusive to Unity. I still hate Unreal point lights and emissive materials compared to CryEngine but can't exactly point out why. It is as if CryEngine ones were sharper and Unreal 4 had some excessive sheen to them making them feel a little fake. Same for the specular reflections feeling off at times.

 

Maybe it's related to the fact they avoid analytic light ...

 

Not necessarily, as said and as you're saying there are many tradeoff's that has to be made for variable scenario's it's just a matter of choosing the right one's and even then it can be subjective in some circumstances. By default in UE4 unless you plug a constant in and change it, specularity defaults to 0.5 which in most cases "does the job" although in others it might need changing.

Thing is, Unity has to care about stuff like low end mobile so it's just generic performant stuff that caters to many sub markets. It seems great at what it does, although I'm looking for accuracy and doing things like sub surface scattering for non-metallic objects is going to have a performance tradeoff but also look different.

It's like many are hoping for a one fit's all sort of answer when in reality it's a massive accumulation of little things that equate to a final picture. It's a hell of a lot of work.! For someone anyway, plenty of good stuff on the asset store.

 

This might come handy - voxel based global illumination in Unity:
https://www.assetstore.unity3d.com/en/#!/content/64763
http://www.sonicether.com/segi/

I love how much difference it makes.

 

Performance and stability in a real game scenario ?
It will be a long road before you can compare it to super fast Cryteck SVOGI

 

Probably, but this is still work in progress. Hopefully, they can match 90% of Crytek's performance.

 

My pathetic machine gets 7 fps in his sponza demo

Really though, looking at svogi, the effect is incredible. It's just waaay better than traditional ssao and the like.