LPV can actually be relatively decent if tuned correctly. However, since 4.18 was released, it appears as though the engine doesn’t support the feature at any level as I’ve had no luck in getting it to work again.
Honestly, LPV has its issues, but they should at least allow it to be used in the engine and make sure it keeps working to some degree. I rather have LPV and decide not to use it, than no dynamic GI solution at all. Especially since the only way to get VXGI is to compile it from source, which really limits the number of indie devs that can use because of third party incompatibities. (For example, both the indie licenses for the Ikinema and Simplygon integrations don’t allow for custom engine builds.) And unless most devs have tens of thousands of dollars to spend on the enterprise licenses its a complete double-edge sword.
It’s a very frustrating spot to be in for small indie devs who do everything they can to try to ensure their product is balanced between good graphics and good performance.
And as a result I’ve been working on a less automatic solution. But at this point, it requires the dev to manually add every material in the scene to a data table along with an accompanying vector variable containing the reflected color values of the respective material. Which takes an obnoxious amount of tedious number entering. I mean, one of the levels has about 700 different materials.
It would be nice to just have SOMETHING to use, even if it isn’t perfect.
haven’t tried it myself but at least I’ve seen that LPV is now shown as a plugin. if you go to Edit -> Plugins, is the LPV plugin active?
there was a nice guide about LPV which showed how to get relatively good results. IIRC the biggest issue with LPV is that it just stops with a hard cut after the voxel distance so there’s some noticeable popping at the edge of the LPV range. Also I think there were some indoor-outdoor light leaking issues which made it not feasable for a number of game types, but I can’t remember if there’s any other dealbreaker with LPV
Thats not really true, SVOGI which is fully dynamic GI is runing on consoles and even not high end pc’s very well. Beside of SVOGI a ton of games are using different “maybe not directly dynamic” GI techniques even on older PS3 games. Probe Based GI is probably the most common one, used in FarCry games or even in The Diviosn on PC and Consoles.
You mean to tell me that the only GI solution they had left was Light Propagation volumes and they got rid of that too?
I’m so ready to abandon this engine at this point, especially since custom shaders are a nightmare for those of us that are used to normal engines where adding shaders and rendering elements is simple and easy and doesn’t need any “blueprint”.
There is simply NO excuse for not having dynamic GI at this point. This engine has been around for quite some time now.
They could do something as simple as some fake GI like use the ao to piggy-back some fake bounce light, similar to what SSDO does.At least that’s something.
To answer about light propagation volumes in terms of their performance: it performs beautifully.
I’ve used VXGI for some time now and it is garbage as far as I’m concerned (it might look better, but in a standalone demo I tried it enabled and disabled with very reasonable settings on a GTX 1070, and I lost 100 FPS or so when enabling it. This is on their own demo!
Nvidia, if you’re going to promote Voxel-based GI, and then claim that a GTX 970 can use it without issues, then you might want to make sure the hardware you sell is even capable of running it reasonably well, seriously, it’s obvious that no modern GPU is good at handling lots of voxels).
Last rant: Epic Games, I can guarantee that you are losing customers/business due your refusal to focus even a single release on GI.
You had LPV’s partly implemented for ages, something that runs pretty well, then never bothered to finish it, and now apparently got rid of it completely? In a way it’s kind of funny that whoever is in charge doesn’t get this somehow.
I have found that VXGI 2.01 which was introduced in 4.19 is actually quite usable for production. The framerate has been significantly improved with the single pass solution that VXGI 2 offers. Even VXGI 1.0 can be tweaked to usable levels with a few parameter changes to consolevariables.ini if you are on a lower engine version than 4.19.2.
On a 6GB 1060 these settings look decent enough and can maintain 90+ fps Epic settings in most situations. Stacklevels has the biggest performance impact by far, 3 is the highest I would go(the default is 5). The specular tracing in VXGI 2.01 has almost no performance impact with above settings and removes the need for reflection captures.
Most solutions don’t work in the way that Epic wants them to work, for example VXGI is too much for many machines and doesn’t work on consoles. Enlighten doesn’t work with dynamic objects and still requires baking.
I don’t think there’s going to be a dynamic GI system added until the next consoles get released and they drop support for things like Xbox One.
100 FPS don’t tell anything. It’s relative measurement. From 1000fps to 900fps cost would be only 0.11ms. From 200fps to 100fps cost would be 5ms. So what was the cost? What was the baseline. FPS isn’t good way to measure cost. Always use absolute time.
Hi guys, what about this solution? Tutorial UE4 Dynamic GI - YouTube Right now it’s working fine. Don’t know about big projects but for little scenes, it’s good to use
LPV isn’t great in UE4, it’s slower than it should be, only works in a set space which isn’t very detailed, and it only works with certain types of lights, and it still has issues with light bleeding (not good for indoor lighting). It’s mostly useless due to those things.
Sorry for late reply but its a bit MUCH trying to keep up with best lighting method on less than stellar hardware etlal, so yes why would xbox one support be removed , given On June 11, 2017, Microsoft lowered the prices of the 500 GB Battlefield 1 and 1 TB Forza Horizon 3 Xbox One S console bundles by US$**50 , **hardly archaic ( can’t turn off bold). Anyway , dynamic lighting for world comp open worlds is far too costly nor are there any lighting methods that are working or ideal for dynamic situations( building static takes forever on detailed terrain ) anyway short of very fast & current hardware will DGI coming in 4.24? work for such setups ?
I wouldn’t expect any new dynamic GI systems to be targeting something like an Xbox One, we’ll see what they can do for the new generation of consoles which have been talking about raytracing features but I kind of doubt that they will be able to do that all that well. At the very least, some of the dynamic GI options that had previously been explored (like SVOGI) might be reasonable to use on the new consoles and can be added in to the engine.
" Anyway , dynamic lighting for world comp open worlds is far too costly nor are there any lighting methods that are working or ideal for dynamic situations( building static takes forever on detailed terrain ) anyway short of very fast & current hardware will DGI coming in 4.24? work for such setups ? "<sorry if not clear, but while ty for console verify, this was for my PC-appreciate any help on that front .Ya, high end lighting would indeed be a lift for consoles so the GI sound solid later.
ASking because adding too many directional lights far too costly right now , and fog & skylight aren’t scattering the lights to 3 of the tiles,they are dark on 2/3 furthest areas on all those tiles.
Theoretically they’re still working on raytraced GI, with final gather, 1spp, and ambient occlusion it should work ok.
Still, without heightfield GI being brought back, and included in reflections, it won’t work for open world games. Not too mention some form of volumetric fog interaction, what’s the point of the upcoming volumetric updates without it? Also how well does RTGI work with Niagara atm? Not to mention lack of spline meshes, etc. etc.
Unless Epic actually makes a “triple a” game again, let alone an open world game, I suspect GI and any derivation of will still remain half functional for most such titles, just like almost everything else that isn’t close enough to Fortnite. But hey Fable 4 is in development, if they’re using UE4 again maybe the team will push commits back to UE4 main branch like they were doing with Fable… whatever that title was that gave almost working LPV, I forgot already.
OK that is ODD,'if true, can you back that up somehow ? How is the world comp ‘module’, part of UE4 suppose to be usable if lighting is insufficient atm?
How was the kite demo done, if accurate lighting isn’t available for world comp ?
I surely hope, UE4 hasn’t abandoned world comp in favor of actions games, I hope your inference is out of date .