Just sounds like a worse version of the LPVs the UE4 already has. Not to discredit his work of course, making a solution like that by himself is of course a very impressive task.
Also, SVOGI in the CryENGINE (SVOTI as they call it) runs at a 10 FPS difference from a normal scene as well. Unlike your solution though, it has nothing to do with the materials and it functions as a standard, full GI solution. This is something that -can- be done if only the effort is put into it.
My solution can be used with normal lights or materials, I can tell you sure the different is not just 10FPS in a scene with only directlight, if you put that in UE4 sure give more FPS
Right, but it still requires the materials to be changed individually to support it. SVOTI is literally just a checkbox to turn it on and suddenly the scene is using it.
It’s nothing against your project, I’m honestly going to probably end up buying it at some point.
However, no matter how good a plugin may be, I’m always going to be hesitant to make a long term project that relies on it. There’s no guarantee that 1-2 years down the line it’ll still be updated and maintained for those future engine versions, and if for any reason it’s not updated and the project relies on it then suddenly that’s a ton of work that I either need to redo completely or I’m stuck on an old engine version and all the bugs/issues that come along with that. A plugin will never be a true, complete replacement for official support.
Yeah always is better a full managed system, but as said I made 2 systems, A. you can change the materials but for that you have the material instances, change one create childs.
Second part is you can edit the engine and use the normal lights without change any material https://github.com/Hevedy/UnrealEngine/commits/Hev_RASL_4.11
But yeah will be always good to have a full managed system, anyway the full managed system won’t give to the artist the freedom of edit parts or give different .
wow that sounds like the most incorrect statement ever made.considering his solution is 100% fully dynamic with UNLIMITED bounces and includes real time reflection and occlusion
I have tested numerous GI solutions and i agree with Hevedy, that is by far the best GI solution i have ever seen or handled.
I just quoted his website, those are his own admissions of the technique’s flaws.
That describes almost every dynamic GI system. LPVs and VXGI are both the same in that regard. As far as I’m aware, there’s only one non-100% dynamic GI technique, the one Enlighten uses.
VXGI has this covered as well.
The last update LPVs had added this, and VXGI had this from the beginning.
In all honesty, the technique isn’t necessarily a bad one, it’s just not really an improvement over what we have already, and looking at his own cited drawbacks it’s more limited than VXGI by far. Things like reflections and extra bounces are just extras on top of the technique, what’s more important is that we get a solution that runs fast and actually works with every type of light, which that solution doesn’t offer. Those two points are the most important things to work out. Also, if you want the technique to work indoors as well as outdoors, you can’t have major issues with light leaking, which is why LPVs and SEGI aren’t really what people are looking for. VXGI gets around the light leaking problem by just throwing voxel density at it until it goes away, generally they’re small enough to the point where there isn’t overlap through walls, which is why it’s so performance heavy.
no bro i have tested both vxgi, lpv, dfgi, hfgi, ahr.
your statement that this is a worse version of lpv is so wrong its not even funny. that’s why i mentioned it had unlimited bounce.
Plus his solution just got released two weeks ago and its in early alpha.
second of all the polish is better than every gi solution i have tested by far.
I have actually taken a week and tested these solutions using the same scene unlike you.
Now i can’t say the same about HFGI because i have actually seen its magic at work and still work with it trying to make things around it.
Its simply amazing.
and that is why i think this is some sort of conspiracy. with their use of distance field and height field technique for other things. there is no way they can’t finish up HFGI in no time. Especially when himself said that HFGI can be optimized so well that it can run on console.
Look, I would rather avoid getting into an argument of all things over this topic and keep things to a normal discussion if possible. There’s no cause for personal attacks or assumptions, this isn’t a topic to get so heated over.
Even taking you at your word that you have tested all six techniques mentioned here, it doesn’t change the fact that the author of SEGI himself listed these flaws on his own website. Go there yourself and have a look, I’m not making this up just to badmouth the solution: http://www.sonicether.com/segi/
The problems he’s running into are ones that a lot of these solutions tend to have, it doesn’t matter how many bounces or extra features you have if your light shines into the wrong room entirely. That was why DFGI at the time seemed like it could be a promising middleground.
HFGI has its own limitations, mainly that it’s based off the terrain color, therefore only useful for scenes that make heavy use of it. It’s perfect for things like the Kite demo, but not really outside of that. That was why it was being developed in conjunction with DFGI, the two were meant to be done together.
There is no conspiracy, Epic just had other priorities at the time. The point of this thread is to try and make dynamic GI a priority again.
Why are you using a faux argument and personal attacks to escape the discussion? When did this discussion ever get heated or when did i ever personally attack you?
cause i said you are ‘so wrong its not even funny’.
huh? smh…
When is an obvious truth personal attack?
The fact is you are wrong and its bad to portray people’s work in that light. I mean have you seen LPV? Go download the sponza scene from cryteks site and run LPV on it and then lets do a comparison (only because i deleted most of those things from my drive already.)
But let me forwarn you, the results in comparison will be laughable. That’s why i said your statement is ‘so wrong its not even funny’. You take offense to that because you know its truth.
Now back on topic. Clearly i know the limitations of his tech since i have been following it for over a year now. If you actually tested the demos you will know that putting up a transparent road map and time table doesn’t mean the solution is full of light leaks everywhere as you are trying to portray it to be.
Secondly, the solution was just realeased in early alpha weeks ago.
Thirdly, I believe its a conspiracy because its very clear that Epic stopped the development of GI so that in 2 years they can push the release of unreal engine 5 with it.
LPVs were also someone’s work, the guys over at Lionhead. And I’m not portraying SEGI any differently than the author of the technique himself, I’m just saying exactly what his own website says. I don’t know how I can be any more fair to the guy than that.
Yes, I have. It’s honestly not as abysmally bad as you make it out to be, it’s just not where things need to be in the engine. SEGI has the same problems that LPVs have, it just has more features added to it.
Again, I don’t claim to portray it in any way other than what the author himself cited on his own personal website. My only exposure to this technique is what the author personally wrote about it, and I’m inclined to believe him since I know he’s done good work in the past on other projects.
Unreal 5 won’t be coming out for a very long time, much more than only 2 years. I highly doubt Epic will wait that long to do something about this.
Just to chime in a bit, i think it’s not really fair to judge something with just reading the cons on a website.
LPV uses at its base just volume textures and reflective shadow maps. While its results tend to be good outside, it breaks in interiors because of a lot of light leaking combined with volume resolution.
SEGI, which uses a voxel cone tracing algorithm (so this means that it’s a SVOGI algorithm), also suffers from the same problems, but since it has a different structure behind (which is the voxel scene representation) it can also attach other properties to the data structure to identify blockers and occlusions.
This means that it can be progressively refined to handle way more scenarios. Of course one could hack LPV into this too, but its basic structure would have to be rethought to achieve the level of precision that SEGI (or SVOGI, or SVOTI, or VXGI) handles from the get go.
SVOGI can also encode rough to semi glossy reflections aside from specular, and a carefully coded system could very well stop using reflection probes or just use them for very far fallback.
So no, LPV and SVOGI class solutions are not the same, even if they are suffering from the same visual problems (just for now though, author lists a lot of improvements for occlusion and leaking on the page), SVOGI algorithms start off way above where LPV left in terms of precision.
AHR, which is kind of a SVOGI class thing, is the most promising solution for unreal now, but the guy is alone and seems to have lost interest in coding it fast.
Anyways, the SEGI demos are very polished, only minor glitches for a 0.8 beta done by a single guy in unity.
I would so much like for a cryengine solution in UE, but i don’t really see that happening now. They seem focused on other things.
It’s a pity that Unreal’s plugin architecture can’t interface well with the rendering backend, otherwise more people would code stuff like this.
Interesting, wasn’t aware it was a cone tracing system, the summary only mentions “voxels” which is of course best known for LPVs in that context.
Of course, but unfortunately I’m not a Unity user as well so I don’t really have the means to test it myself and find out how well it works, I can only go off of what his website says and the test demos he has listed. The low preset seems to run pretty well all things considered, but the scenes are fairly static with extremely low poly objects, almost like the scene was built targeting mobile. Do you know if he has one where there are more moving objects, or higher scene complexity than procedural cubes/sponza? Also, it seems to be working without something like a skylight, is it just capturing a cubemap of the sky for that or is it actually an emissive sky?
Sad part is the CryENGINE V’s solution is just SVOGI, they even call it that in the changelogs, they don’t even bother calling SVOTI there. And yeah, Epic is focused on other things, but that’s just all the more reason to keep bringing it up. It’s the most commonly requested the engine has ever had, we need to show that interest hasn’t died off over the last two years. Realtime GI is a much more useful for most projects even just as a previz tool before baking than something like VR, which seems to have completely taken over their in-house focus.
Yeah, the guy behind it is very well known for his Minecraft shader mods actually, he did the most popular one there is for it. Very advanced stuff, even volumetric clouds and actually putting LPVs in that engine. His prior work with LPVs in that plus the vague description of ‘voxels’ is what lead me to believe SEGI was based on LPVs.
Yeah. There’s also the question of just how well it would run. When Nvidia tried to use more of Unreal’s normal interfaces for VXGI, there was a 4ms performance hit on the technique. If this is ever going to happen, it’s going to need to be an actual standard part of the engine. Even if someone were to come out with a really, really good system as a plugin, I would still hesitate to use it for a long term project because of what’s happening with AHR. If it’s not actively maintained, you could get stuck on an older engine version pretty quickly.
The whole argument about whether SEGI is good or not is pointless. We want something native in UE4, with great support, with receiving constant updates until it runs very well with really good performance.
Agreed, apologies for letting things get so far off topic.
Well, while I agree to a certain extent to this, the very existence of SEGI (which is just a plugin running on UNITY, most probably done in C#) is a very good indicator that the technology and the speed of current hardware is enough to handle that kind of GI.
So while SEGI is uninteresting to us it serves to legitimate our UE GI claims 
Hey, I recently responded to another thread about Dynamic GI over here: UE4 Users Wishlist - Feedback & Requests - Epic Developer Community Forums
I’m documenting all of the discussion going on about Dynamic GI right now, so please keep up the conversation and I’ll keep an eye out.
I know this thread is only about GI but I’d like to point out that in the current state there’s no Sky, Water/Ocean, Road/River, GI (exuding LPV).
I’m blown away by SVOTI, again.
Having these light bounces without waiting even 1 second, and still running so smooth to develop any games with, is truly amazing…
even more amazing is this:
Hmm, I wonder why they put resources into a technique that is so old and clearly inferior to what the competition is doing.