Realtime Dynamic GI + Reflections + AO + Emissive plugin - AHR

[Enter Reality]

Dude, sweet!!!

Best luck for this project, I'm eager to see the progress and how it'll work when it'll be released

Thanks in advance for releasing this plugin for free, this is a huge effort!!!

PS: Consider to add paypal donations

[iniside]

Well, that escalated quickly!
To answer the questions:

• Still haven't started to code the plugin. That screen was taken modifing the source directly

AFAIK you won't be able to make plugin for it. I bet you touch quite a bit of rendering code, and rendering plugins are not possible beyond some simple changes, like new shaders.

Well, that escalated quickly!
To answer the questions:

• Raytracing complexity is independet from scene complexity, but voxelization is not. The voxel resolution suffers on large scenes. I have a few ideas to solve that though

I'm by no means, expert but there are few optimizations that come to my mind:
1. Cascading voxels, around camera in set radius (so entire scene won't be voxelized, and further voxels will be bigger).
2. Prevoxelize static geometry. Static geometry would be voxelized only once. Do not voxelize characaters and small moveable geometry. It would be possible to extend it by for example revoxelize static geometry when it change state (for example, moved, destroyed), and then again do not update voxelization until next state change. This would probabaly save you a lot of time.
3. If you really needed to voxelize scene every frame, you might consider updating further cascades over several frames.
3. Tweak able voxel resolution.
4. If you would use OpenGL you could make use of ARB_Sparse_Texture, to access voxels saved in 3D texture and save a lot of memory this way, As you wouldn't have to store everything, only parts which are voxelized.



In any case great work! If you are considering giving it for free I would consider talking with Epic about integrating it into main branch. When it's ready or usable.

[max.pareschi]

So if you are coding in DX11 chances are no Mac/Mobiles etc? (various OpenGL's)?

Well, since even a PS4 has difficulties running any kind of GI scheme I think we could just forget about mobile GI for thjis generation.
Seriously guys, forget about it please.
Lightmass is more than fine even for the big games, it's just that people don't want to hustle over UVs and baking time.

@RyanTorant
This is looking good btw!
If i were you i'd take a free scene from the marketplace and get it to work with your GI system, so you can better show the benefits and the quality.
I mean, Sponza sucks really bad uahaha.
Try the new Interior Day and Night scene by Koola, if you get the same quality even 10fps should be ok for now.

Cheers!

[DylserX]

If a plugin is not possible, I would love for there to be a Github branch to take a look at how this would look in Caffeine, Which uses baked lighting for the static interior lights (space station) but there is a rotating sun which illuminates some dark rooms.
I may be a bit too excited about this thread...

[Daedalus51]

Well, since even a PS4 has difficulties running any kind of GI scheme I think we could just forget about mobile GI for thjis generation.
Seriously guys, forget about it please.
Lightmass is more than fine even for the big games, it's just that people don't want to hustle over UVs and baking time
Cheers!

Sorry if I am being disrespectful or rude, but if you say something like this you have obviously no clue what you are talking about. Lightmass is not only the workflow suicide for a huge open world project, it also just crashes because of the unbelievably huge amount of data and is completely unable to handle really large areas. Epic didnt even realize that the Landscape demo from the market place crashes the Engine while baking for most people because they dont have enough RAM. Epic uses 32GB as a standard, so everyone with less than that has super huge issues in building this map.

There are a lot of reasons why its not working and I could talk about that for hours (based on my own experience) but thats not the intention of this thread^^

And yeah...very awesome work. Love the idea of seeing this in action in one of the maps from the market place. Sponza sucks indeed^^

Cheers

[iniside]

Well, since even a PS4 has difficulties running any kind of GI scheme I think we could just forget about mobile GI for thjis generation.
Seriously guys, forget about it please.
Lightmass is more than fine even for the big games, it's just that people don't want to hustle over UVs and baking time.
Cheers!

I'm sure all those people making games with changing environments and dynamic ToD, find lightmass and lightmaps in particular, very useful.

Not to mention big environments, where where lightmap memory (used on player machine), will out weight any possible benefits.

[Chosker]

I'm sure all those people making games with changing environments and dynamic ToD, find lightmass and lightmaps in particular, very useful.

Not to mention big environments, where where lightmap memory (used on player machine), will out weight any possible benefits.

^^ this.
even Epic themselves aren't using lightmass on Fortnite because of dynamic time of the day

[max.pareschi]

Sorry if I am being disrespectful or rude, but if you say something like this you have obviously no clue what you are talking about. Lightmass is not only the workflow suicide for a huge open world project, it also just crashes because of the unbelievably huge amount of data and is completely unable to handle really large areas. Epic didnt even realize that the Landscape demo from the market place crashes the Engine while baking for most people because they dont have enough RAM. Epic uses 32GB as a standard, so everyone with less than that has super huge issues in building this map.

There are a lot of reasons why its not working and I could talk about that for hours (based on my own experience) but thats not the intention of this thread^^

And yeah...very awesome work. Love the idea of seeing this in action in one of the maps from the market place. Sponza sucks indeed^^

Cheers

Yeah, you could probably be right.

At the same time (and that's from my own experience, so take this with a grain of salt) from a business standpoint you would have to be a very solid and financially backed-up dev to develop a huge open world game with dynamic lighting on mobile.
I don't want to sound disrespectful to anyone here, since the community here is much nicer than the unity one, but you all know what this means.
Whoever wants to do big time stuff on mobile either develop their own tech to be used inside of any engine or they just release their games without complicated gfx features, focusing instead on gameplay and novelty. For me at least that's the smart way to go.

Not to mention that you could just split your open world and manage it in tiles using streaming and lods.
You have less than 32Gb on you dev machine, honestly? just buy it, ram is dirt cheap these days. The roi on an investment like that is a day max, and human time is much more precious than cpu cycles.
The only real problem I could see for lightmass on mobile is memory management on the device (as Inside said), and with careful prep work that can be done right.
Also destructibles could be a pain, but those are edge cases on mobile really.

In any case really, I'm one of the first that started a thread about alternative GI methods so it's not that I don't want that to happen.
I also have come to grips with Lightmass and its convoluted workflow and final gathering nonsense: really UE devs, this seems like working in mental ray, unexposed properties, no precise documentation with use cases, leaking if you don't snap on the grid?? What about DCCs with no UV grid (like cinema 4d?) and again, Final effin' Gathering? in the meantime Arnold, Octane and many others came and blew MR and similar tech out of the atmosphere with a bazooka.

Anyways, I just like to be realistic, and work with the tools I have now.
Besides, I'm so much accustomed to waiting times on vfx that lightmass baking seems like a joke to me! I usually wait 20min to 2 hrs (depending on complexity) for just a frame

Seems to me that a lot of people want "GI now!" just because it's looming on the horizon and it's way cooler than other stuff. I'd settle with better reflections and penumbra shadows with geometry area lights anytime, gi can be faked in so many ways.
Ofcourse the workflow would be soooo much nicer with that, but you're not gonna use it on mobile now. 2-3 yrs maybe, not now.
For now I'd worry about realtime shadows for instance. Or performance for those who can't use metal.

[max.pareschi]

@RyanTorant
Aw man, sorry for hijacking the thread. Of course I'm looking forward to your stuff! Thanks a lot.

[iniside]

Ok. This entire discussion spanned from unfortunate mention of mobile hardware.

Not everyone is working on mobile games. I'm of mind that mobile gaming end on chess, or other simple 2D game. In that case you don't even need lighting beyond emissive material . For tha matter It's beyond my mind why would anyone make 3D mobile game. When I see that game is 3D in PlayStore, I just skip it and look for another 0o.

There is a reason why I have beefy PC. To develop and to play beefy, fancy games .

It's beyond my head why would play shooter or complex 3D RPG (which would really use fully dynamic lighting) on mobile device ;p.

Anyway. I have a question. Serious one. How do you want to fake GI with:
1. Fully dynamic lighting, and more or less dynamic environment ?
2. PBR material setup, where faking lighting on it, will usually break it for different lighting condition ?

[Daedalus51]

Yeah, you could probably be right.

At the same time (and that's from my own experience, so take this with a grain of salt) from a business standpoint you would have to be a very solid and financially backed-up dev to develop a huge open world game with dynamic lighting on mobile.
I don't want to sound disrespectful to anyone here, since the community here is much nicer than the unity one, but you all know what this means.
Whoever wants to do big time stuff on mobile either develop their own tech to be used inside of any engine or they just release their games without complicated gfx features, focusing instead on gameplay and novelty. For me at least that's the smart way to go.

Not to mention that you could just split your open world and manage it in tiles using streaming and lods.
You have less than 32Gb on you dev machine, honestly? just buy it, ram is dirt cheap these days. The roi on an investment like that is a day max, and human time is much more precious than cpu cycles.
The only real problem I could see for lightmass on mobile is memory management on the device (as Inside said), and with careful prep work that can be done right.
Also destructibles could be a pain, but those are edge cases on mobile really.

In any case really, I'm one of the first that started a thread about alternative GI methods so it's not that I don't want that to happen.
I also have come to grips with Lightmass and its convoluted workflow and final gathering nonsense: really UE devs, this seems like working in mental ray, unexposed properties, no precise documentation with use cases, leaking if you don't snap on the grid?? What about DCCs with no UV grid (like cinema 4d?) and again, Final effin' Gathering? in the meantime Arnold, Octane and many others came and blew MR and similar tech out of the atmosphere with a bazooka.

Anyways, I just like to be realistic, and work with the tools I have now.
Besides, I'm so much accustomed to waiting times on vfx that lightmass baking seems like a joke to me! I usually wait 20min to 2 hrs (depending on complexity) for just a frame

Seems to me that a lot of people want "GI now!" just because it's looming on the horizon and it's way cooler than other stuff. I'd settle with better reflections and penumbra shadows with geometry area lights anytime, gi can be faked in so many ways.
Ofcourse the workflow would be soooo much nicer with that, but you're not gonna use it on mobile now. 2-3 yrs maybe, not now.
For now I'd worry about realtime shadows for instance. Or performance for those who can't use metal.

Sorry, I might missunderstood you then^^ I was not talking about mobile I was talking about next gen games with large worlds (like what we are working on with Dead Island 2) And there, Lightmass just kills everything.

Also, I was talking about you guys when I mentioned the power of the PC. Of course I have enough RAM at the office^^ But I would say that not everybody who uses Unreal 4 on the subscription right now has a machine as strong as the Devs at Epic. So thats why I said it. Also, our world is split up into multiple tiles etc...but nope, Lightmass still cant handle it. So I dont know what your example is based on (have you tried building a 10x10km landscape filled with tons interactive objects, breakable physicalized objects, houses that can fully be explored, millions of grass patches and thousands of trees and rocks and what not?). Mine is based on the fact we tried to build our game with Lightmass which was a complete failure. So of course I want something that makes this whole workflow just way more manageable and less unstable.

And just to say something to what Inside mentioned...he is totally right! PBR breaks extremely fast if the lighting is not holding up to it, so that is something very crucial as well (and I would say that lighting and reflections belong so closely together that you cant split them up...at least not with PBR).

Cheers!

[RyanTorant]

People people, I know how fun it is to discuss about this topic, but
lets try to keep this thread focused.
Hijackers will be prosecuted!

[max.pareschi]

Sorry man I opened up Pandora's box here.

I'll copypaste some of the more relevant replies and open a thread in rendering, because it's still and interesting topic.
Have a nice one!

[Daedalus51]

Yeah...sorry^^ I just love talking about these things But you are completely right with what you were saying and my mouth shall be closed on this from now on Lets just talk about this cool AHR!

Cheers

[RyanTorant]

Great then! I'm in for a thread discussing all this. Where are you going to post it?