Framework for Unity

Big edit after feedback:

This is nothing more than a passing thought at the moment but it's something I keep thinking about doing. I've made quite a few tools over the years, so I thought about putting them up on the asset store.

After feedback, this is what I have so far:

-Voxel based GI solution
-Terrain system?!?
-Physx extensions (potentially for animation)?

Keep suggestions coming.

 

There are already popular assets on the store for some of those points. I would concentrate on things that are not available in the store yet. Otherwise you'll put a lot of effort in things where people are likely to pick a different solution anyway.

If you put stuff on the asset store, especially more complex systems, you'll have to provide customer support a lot, unless you want to get bad reviews. Just look at some popular scripting assets to understand what kind of support you have to deal with.

[EDIT]
Unity development is moving rather fast in recent years. It can be quite a challenge to keep compatibility and support the newest release in a timely manner, especially if you provide a lot of different complex systems. Always supporting the latest version and having backwards compatibility seems to be a very important for many asset customers.
[/EDIT]

If you want to reach many customers, $250 might be too expensive. I think it's too much for many hobbyists at least, which I believe most asset store customers are. Companies on the other hand might come up with their own solutions anyway, rather than buying things off the asset store.

Here is what I would do as first steps...

1. Look at the asset store to figure out which of your features are not available in the store yet
2. Look at forum threads of popular assets to figure out if you want to provide such support
3. Provide separate assets rather than an all-in-one solution, e.g. "New input manager" and "Deferred decals" (which is a bad example, because there are already quite established packages for these things)

Splitting it to several smaller packages allows to reduce the price and let customers pick what they really need. It feels not good to buy an all-in-one solution where you own half of the stuff as different assets already or is available for free.

 

You can't do all of that and do it all well, not as well as the assets that specialize. No single person could. So what you are in reality proposing is a bundle of sub par stuff compared to the assets that specialize.

That could work in some cases but not this one, because you don't have a focus. Like if you bundled a bunch of tools for making a specific type of game that inherently carries more value. But a bundle of rather random stuff for a discount, versus purchasing individual assets that specialize in one thing. Not really interesting or a great value considering the fact that it won't be competitive with the specialized stuff. And that's not a knock on you, it's just literally there is no way a single person could do all of that well.

 

Performance is the primary concern, a lot of them I probably won't get around to although anything to do with a shader would have to be done (again due to performance). I've had thoughts on how to approach it but not sure yet.. Might require a new rendering solution..!

I've looked at a lot of asset store stuff and own quite a bit but I've had another look, some of the timline editor (cutscene) manager looks better than mine.. The input manager is open source so that's a maybe if and when I ever get to it.

UI system would be a port from another framework I have, not sure if I want to do it but I suggested it because it's done already..

I need to get to the crux of general performance issues, see if there's anything I can do about it.

I can't really sell a lot of them seperatley, they just wouldn't work right without each other.

 

Who says? You're assuming I'm starting from scratch here, which on a lot of stuff I'll be porting it from a framework I made a while back before Unity 5.X got released.. When UE4 came out I dropped it, not to mention the ample amounts of resources already there to be used.

Also you're assuming it will be a single person working on this.

The first to come would be terrain / GI and shaders.. After that see what happens and I'm not denying it could take me a long time. Still, it gives me something to do..

If there's no specific interest, I'm more than happy just to port it all into Godot.. At least then I know there will be no restrictions and as GD3 has a lot of stuff I already need to get this working quicker, hey.. It could be a better idea.

 

Well the strong implication was it was just you. If you have further specific information to the contrary then fine that changes things. But I don't see you saying you have a team of people ready to jump on this.

In 20 years as a lead developer, you would be the first person I have seen to take on that much and do it all well, regardless of how long you have been working on it.

 

But what about your exosuit game? :-/

 

Cool story bro.!

@Martin_H

It's just a passing thought Martin, I've accumulated a lot of tools and stuff over the years and wondering if it's worth getting back to roots in the more development side of things (maybe earning a little bit out of it?). It seems it's what I'm best at..

I doubt I'd actually get through half of it, just start with porting over the terrain system / GI and shaders I have.. In all fairness I need it for my game anyway.!

 

Well if you plan on working on such a huge bundle why not bundle it in your own engine with a base framework like Monogame and maybe have something like a patreon for funding. Seems like a better fit to me for this kind of project. So i vote for your second thought!

Edit: Something like Nez but for 3d

 

The framework I made is over 5 years old now (it's not quite good enough to be called an engine), it's outdated and majorly lacking compared to even one man engines like S2 HD.. If I was going to implement it I'd be crazy not to pick up something open source like Godot.

There is nothing that causes more vitamin deficiencies than making a game engine..! Especially with engines like LY and UE in the loop now it's physically impossible without a team of a hundred for it to be worth it.

I guess it's a bit of a shame, wouldn't of minded getting something out there for Unity.. Although @Peter77 is probably right, even as a base package (Terrain, GI, Shaders) it'll be too expensive and one's with bigger teams who can or would pay for it would probably DIY anyway.

I'm sure I'll find other ways to contribute.

 

I think Unity has the fastest pace of development than any engine out there now. It would be really hard for you to update your framework to Unity. In 2018 there will be new lightweight entity component system, C# job system, new custom compiler technology, visual material editor, visual scripting etc. If I were you I would at least wait a little bit to see what Unity is going to introduce in 2018 version.

 

Extremely valid points, again all of this is a passing thought.. I've not a clue how much work it would be to integrate said creations in the first place, then to have to re-do the integration a few times over would be interesting / frustrating?

So another question, if there's one thing on that list that would be most useful (general question to everyone here) what would it be?

 

@ShadowK : Don't let the negativity put ya down man.... Granted it may be quite a big project - I've found I've learnt the most knowledge I have from doing massive projects, whether it works out financially or not, the experience alone is worth the endeavor.

I say go for it, what do you loose even if it don't work out? The time wouldn't be wasted because you gained knowledge.
So you don't loose anything, so go for it man.

 

I don't start threads off with a personal anecdote, but I am loosing my skills at this point.. I've spent the last couple of years tweaking vertex's / sculpting / texturing etc. etc. and just generally faffing about. So when it comes to what I used to be alright at I don't know my head from my behind.

So thanks, I agree.. If nothing comes of it no biggie, y'know at least it gets my brain working for a bit and at this point that's what I need.

 

I have no doubt that there's plenty of framework stuff that could be made for Unity, but quite frankly I think that this would be a question of developing huge systems from scratch just to make a relatively small iteration. With a lot of the stuff you've listed, I think there are highly polished tools out there that do 90% of what you want. Even if you improved a little on them, at best you'd be entering a crowded and competitive market, and especially at the price point you're describing it seems like it might be very hard to sell.

If I were you I would aim at the 'next generation' of tools - stuff that is either sorely lacking or hasn't really been done yet. Things like

• SVOGI (or whatever it is that lumberyard has);
  
• A semi-procedural character creator (not sure if Unity is going to do something in this area).
• A better cloth physics/creation framework;
  
• Procedural anything - animation (I think the physics based animation is a good idea), sound generation, level generation, building generation, asteroid generation, planet generation etc etc

I guarantee this will get the wheels spinning, and I think not only would it be a lot of fun but it would do more for advancing game dev IMO than iterating on existing tools.

Finally I think Unity have found some coffee because lately everything is on the map, they are even jumping into making starter kits and stuff like that. So a lot of the core stuff that you're talking about might be taken up in the next few years by them, making it even harder to compete. So I would try something more radical, the sort of thing that would be too difficult for Unity to be able to sell to their investors for at least half a decade.

 

I agree and would sum it up as "focus on something that has no use on mobile", because that's likely something you won't see from UT and the Assetstore devs seem to avoid stuff like that as well. SEGI is open source now, if you're interested in realtime GI you could get involved there.
Another approach worth thinking about imho is "what would be the thing you need most, to drastically cut down the time to make your next game?".

 

Any idea when this terrain system is coming? That was the reason I was looking into it, but sure I agree.. It's going to take me a long time to do it anyway so there's no point if they're sorting most of it out in the next couple of years.

@Martin_H

Funny you should mention SEGI, I'm currently looking into it as viable option.. As said, not starting from scratch as I'd never get finished (ever)..

 

No idea, but if nested prefabs are on the way then anything is possible ..

 

I'd just dump it all to github under permissive license.

The wondering about getting back to the roots reminds me a scifi story about the baker.

Basically, there was a young boy training to be a baker. He watched space ships (scifi, future) fly from nearby spaceports and wanted to become a ship captain.... and one day a miracle happened - some visitor in their bakery offered him opportunity and he took it.

Years passed, the boy grew old and was working as a captain on a big passenger cruise ship. The dream came true, but he was mostly dealing with mundane bureacratic stuff all day through. So when he went to sleep he dreamt about baking bread....

 

We can't compete with Unity regarding how things integrate into their tech. They can alter C++ core if that is the thing to make C# tick faster. I'm not sure we can make a very good terrain system without some rework of their C++ core...

 

@ShadowK
There is Gaia to generate terrains and CTS to get great visuals with a very optimised one pass shader, or MegaSplat.
They are among the best options for terrain, i don't think it worths it trying to make the same.

You must propose plugins with new features or with advantages like better performance or better tooling.
Anyone can put plugins in the store, if you want to release something just go for it; don't wait some years that Unity gets new lighting or terrain system making your plugins obsolete.

 

This would be along the lines of UE's system, not just a bolt on or shader.

There's no reasons I can't use .DLL's for plugins and then obviously abstract certain components, there's a ton of issues I forsee before I start. One of the issues was frustrum culling with the original terrain, per piece had too much overheard.. Chances are I'd need access to the graphics API's..

For point of origin reset I need to grab the transforms of all objects and move it around the player, again issue with that lies on optimised multi-threaded transforms and the ability to manipulate it.

Also I'd have to optimise the rendering path and a lot of issues I've suffered with from DR is fill rate from foliage, I had to make many tweaks to UE before they got an FR path.. There's plenty of examples of free un-restricted implementations of an FR+ renderer which would solve a large array of issues, problem being it's a lot more work.

Not to mention the chance of it breaking with every new release, if I can submit to a master branch of an opensource engine I wouldn't suffer these issues.

The likes of Godot already use an FR+ renderer and the system is wide open. It's only a matter of porting it and then commiting it to the Git.. It would be easier on several levels.

The question isn't really should I do it, because even after Unity implements a new feature it takes a while for it to be the best it can be, it's should I do it in Unity?

Honestly the last time I made an attempt to do anything worthwile in Unity was in 4.6, I wanted TAA and was lacking a velocity buffer which was a pain. Things have probably changed since and I know Unity has made a lot of effort to open things up, as to how much at this stage I'm not sure?

Anywho, my first pet project is optimising the hell out of SEGI.. I need a proper real-time GI solution for my game anyway and my LPV implementation doesn't look anywhere near as good as these new fangled voxel cone tracing algo's. That'll set me up for the next two or three months just to start with.

 

Scriptable Renderloops will be coming to Unity. Maybe that'll give you a boost in productivity if you want to change the rendering pipeline anyway.
If it's mainly about skill maintainance and improvement I'd choose based on what you want to make your next game in. The way I see it even if in Unity half your time is spent wrestling with Unity-specific perculiarities, that time should benefit you when you work on your next game in Unity too. If you do your big framework projekt in Godot and it turns out you spent a lot of time on Godot-specific issues there, that time might be mostly wasted if your next game isn't gonna be made in Godot as well.

 

You can succeed and make a great plugin people want, but you'll have to be daily available at launch and still available lot later (until you quit and your plugin becomes deprecated like many plugins).

If you put lot of time and energy making some plugin, it must worth it , you must make money or get some other befenifs like leraning for example. Don't do it if you think it will be a waste of time and energy and a pain to maintain as Unity releases are quite fast.

Why needing real time GI ? Witcher 3 is not real time GI but it's a awesome looking game.
I think the content and game interest matters thousand times more than to get the best lighting possible
Question yourself what is your goal ? digging on tech or making a game first.

Anyway there is CryEngine or Lumberyard about real time GI already, or UE4 LPV looks good for outdoors.
Even with that tech available i don't seen some impressive and interesting game made by indies lol

 

The Witcher 3 is also a two and a half year old game. According to an interview their reason for not using real time global illumination was solely because they felt it was too expensive at the time. If we assume they made this decision early in development the latest graphics hardware could very well have only been the GeForce 700 series.

They might have been able to predict the state graphics hardware would have been by the time of their game's release, but since the game's release we've had one entire generation of hardware and we're on the cusp of a second generation. What may not have made sense then may make sense now.

https://www.dualshockers.com/the-wi...her-lighting-consoles-graphics-and-much-more/

 

They used GI probes, and as they said it delivered good results. So yes it's not full real time GI advanced tech like Cry Engine.
But i was meaning : It's not top tech lighting that will transform a poor game in a good good game. If you don't have skills and a team to make a great game, does it matters if you have a 3D engine with real time GI lol ?

 

Well I'm not going to release something and then not bother to support it, doesn't matter if you release a game, general piece of software or a plugin / addon it will need supporting..

There's many pro's / con's to real-time GI, one of the pro's is workflow.. You don't need to worry as much about uv mapping / lightmap channels, it requires far less setup and it can save you a lot of time.

Negatives are you have to be more budget conscious before starting your game, with modern solutions the footprint is relatively low like 1 - 2 msec on a GTX 780. Although I'm not really convinced it looks better than Enlighten, so for small games I don't see much of a benefit.

For larger games Enlighten becomes somewhat useless, GI cache is always too large or the amount of texel density is too low for it to be worth using, not to mention the amount of time it takes to pre-compute. Lightmaps are useless for large games and whilst the likes of W3 was artistically well made, that's it.. Doesn't look any more impressive to me than Crysis 3 which is far older.

UE's LPV solution is nothing more than a half finished attempt by Lionhead, as impressive and exciting as it initially was it really does have a lot of problems and now Lionhead is no more and Epic aren't interested in finishing a GI solution, so your choices are limited.

When it comes to LY / CE, ever tried releasing a game in one of those? I'm not entirely convinced it would be harder to make a GI solution (no matter how basic) for Unity than try and fight constantly with a AAA beta engine or whatever it is CryTek do. LY does seem like it will eventually get there as they are ripping out most of the original CryTek core, but that's not an easy task and it's not going to happen for many years to come.

 

https://github.com/sonicether/SEGI/releases

 

Could be. I don't know if it's the same approach as Lumberyard's method, but I could hardly get the built basic demo to run on a laptop with a GT540M so I think there's a lot of room for improvement.

Insta-lighting that's extremely performant is still a very big opportunity. Anything that doesn't require any fiddling around and doesn't have important caveats is going to be a winner. I think this area is still very much up for grabs although SEGI has been a massive step up.

 

I'll be updating my progress here: https://forum.unity.com/threads/segi-fully-dynamic-global-illumination.410310/page-34

 

Lightmapping is not great for open world indeed.
Ubisoft games uses AnvilNext engine that is an evolved version of AC 4 game tech, it's not full real time GI but the baking is faster than a full lightmapping.
https://bartwronski.files.wordpress.com/2014/05/assassin_s-creed-4-digital-dragons-2014-no_notes.pdf

For some outdoor game it looks good enought if you need directionnal light only and static meshes only, it's really faster than real time GI. It depends on features and quality you need.
www.youtube.com/watch?v=IGO7VNcq6uE

What do yout think about Wicked Engine ? It is not what you are trying to do ?
www.youtube.com/watch?v=siui7gLxcIQ

CryEngine has SVOGI and it works well

Many games or Mmo have been released with CryEngine, it's a work in progress engine like Unity or UE4; it has new realeases with changes or new features, if you got the ability to learn it and use it you can indeed make and sell a game.
www.youtube.com/watch?v=yj4NJhZiMAs



Anyway, if you can achieve something twekable and that works for open world, i think it will be very interesting for lot of people

 

That's all very nice in the land of high-end gaming PCs. Their game's also on consoles, though, where the rate of change is much slower.

 

The Xbox One X is a pretty respectable improvement over the original Xbox One. It's definitely not as impressive as a high-end gaming PC but the hardware may still be improved enough that realtime GI may make sense now.

http://www.techradar.com/reviews/xbox-one-x
http://www.pcgamer.com/heres-how-microsofts-xbox-one-x-compares-to-a-pc/

 

It remains to be seen if the Xbox One X will actually sell. It doesn't really have any compelling reason to buy one.

 

Sure, but unless you want to cripple your sales the game also has to target the original Xbox One. So that'll be your base target, with extra bells and whistles or higher settings for the X, rather than fundamentally different tech targeting it specifically.

 

Looks rather old hat comapared to what you can do with Unity and Enlighten already.

Yep, their voxel cone tracing has been released under MIT.. I'm looking into it as a possible grounds for a solution as well, as much as I think SEGI is cool it has a lot of issues. LPV is real time, that's the whole point of it.. It's generally restricted (like a single bounce, you can do more but meh it's a long subject), it's quick though if nothing else.

I was using CryEngine years back and I know a fair amount about it, not many games have been released with it and even some of the one's listed never got finished: https://en.wikipedia.org/wiki/List_of_CryEngine_games

On top of that even Star Citizen moved to LY..

I tried CE5 and it's buggy as they come and most of the original issues from the SDK are still there.. Also if you can make an MMO you've generally got a lot of $$$'s which means you can shape the engine to fit your needs as opposed to when you're a small team or indie where you rely in the engine to get by.

It's all a matter of circumstance.

 

Good suggestion Billy. This is something that seems to be a non-competitive realm. IDK why standard Nvidia physics cloth is not built into Unity at this point.

@ShadowK ever considered something along the lines of the issues you identified with dialogue and animations, from our discussion in the other thread?

Personally I think you have much too much knowledge in lighting/shadows - not to consider that as the area you should focus on.

 

You're going to have to remind me which thread that was, but yeah with todays technology there's little to no reason why dynamic lighting solutions shouldn't be implemented. Even the likes of SVOGI only have a 2 - 3 ms overhead on an old GTX 780 which by modern standards wouldn't stand up to a $200.00 GPU like a 1060, the 960 isn't far off and you can pick one of those up for a $100.00..

Even on PS4 / Xbox 360 the overhead is about 4-5 msec and this is all dependant on you going with a voxel cone tracing / cascaded voxel ray tracing solution as opposed to something lighter like light propogation volumes.

I can understand why Unity went Enlighten, it makes total sense.. It's about the only semi-realtime solution that has a chance of working on mobile.

The only issue is I haven't kept up on modern techniques, I stopped after an implementation of LPV's about 3 - 4 years ago.. I don't even fully understand the technical implementations of physically based rendering, so it'll take a lot of study and longer than someone with pre-requisite knowledge but I'm sure there's something I can do.

I can't imagine it being totally different to cascaded light propogation volumes, it all bounces rays (or beams) around in a sparse / (non-sparse) voxel octree for culling etc. It's just about finding the lesser of evils, I'm not convinced cone tracing is the answer though from research..

It all seems a little "random", anywho point is if there was ever a time to do it.. It's now.!

 

How about a OnGUI framework that makes sense to use for developers to rapidly develop their tools? Something I once considered with the node stuff, but oh man... what an undertaking. (I honestly wouldn't recommend this though lol, Unity really needs to overhaul their editor framework)

 

This is how it works for Ubisoft open worlds, this was presentation for AC 4 that is an old game, but their last games uses some evolved version of that tech.
Open world game does not need Archi Viz quality lighting.

Kingdom Come, Wolcen are examples of indies able to do it (or Prey about AAA).

Yep, and about your team size, skill and needs.

Perhaps some evolved, modernized LPV ? or a mix with some voxel or probes.

 

Yes Kingdom Come the indie's with nearly 80 staff members, I'm like a moth to a flame ..

 

Maybe I'm misreading that... what's wrong with an indie growing to be a large studio?

 

Are they truly an indie studio once they've grown to that size? He's saying that they're no longer an indie developer because they've reached a size equivalent to many AAA developers. Bethesda, for example, had a team of about 90 for Skyrim yet no one refers to them as an indie studio.

http://en.uesp.net/wiki/Skyrim:Development_Team

 

If they're independent, then yes.

If "indie" means "independent" then their size is irrelevant. On the other hand, if "indie" is a catch-all term for whatever the person saying it wants to mean at the time...

I suspect that Bethesda is never referred to as indie/independent because they aren't and never were. Bethesda Softworks was founded as a division of another company.

 

B-but Minecraft in 4k..

 

I'm going to buy the X, which hopefully allows me to (re)play some of my Xbox One games with an acceptable frame-rate.

I'm just worried that the additional horse-power will cause developers to be even lazier in terms of proving a stutter-free experience. Optimization seems to be an dying art.

 

If you've got $20Mil+ to spend in staff costs alone over the course of three years I'll class them as AAA, if the sole purpose of "independant" means without a publisher then it really has no meaning if used in Zengardens context of the little dev who can..

 

But that... is a meaning? And it doesn't rule out "the little dev who can", either. And it doesn't rule out "AAA". These things aren't mutually exclusive.

I'll class them as "successful". I've got my fingers crossed that you'll be saying my stuff isn't indie.

 

In context chief, in context.. Not split apart, not changed to suit a specific point .. AAA is ultimately defined by a dev team with a lot of money.. The point I originally made was CE is a bit much for an indie dev and Zen used a massive budget title as a reference to what an "indie" can do..

Anyway semantics, I always define indies as solo or small team.. Anything else that's large budget is either an A, AA or AAA dev..

 

We could also just stick to the Wikipedia definition, so we're all on the same page (literally)

Independent video game development
https://en.wikipedia.org/wiki/Independent_video_game_development

AAA
https://en.wikipedia.org/wiki/AAA_(video_game_industry)