Download

What stops Epic from finishing LPV so it works on PC and consoles ?

Just saw a community tutorial on LPVs and it looks awesome. Googled about LPVs on consoles and apparently Cryengine’s implementation worked on PS3 and XB360. So I don’t quite understand why Epic doesn’t just get LPV implementation finalized ? (or outsource it for someone else to finalize it and put it into master)

Thanks

I think LPV was abandoned because they want to do something better, though I guess by now they could have gotten LPV to a usable state until they could get a good dynamic GI system in.

It was developed by Lionhead studio for Fable Legends and we had some sort of generalized version IIRC which they maintained. Now there is no Lionhead studio and as result no improvements on LPV in UE4 anymore

Tbh LPV as good as it gets if one wants to have good performance and dynamic GI. UE4 games aren’t known for hyper realistic looks and IMO LPV provides great GI for non-photorealistic visuals (I mean, Crysis 3 looked amazing and it used LPV as far as I know, and they strove to be as photo realistic as possible). Just sayin’… :o

The resolution is poor, the performance isn’t great, and it needs stuff like cascades so it can do larger environments, it’s not usable in a lot of projects because of that.

“Finishing” LPV’s or any other introduced GI solution definitely got my vote.
However

I have spent at least 50 hours through multiple attempts with LPV’s and the result ALWAYS has been worse then without LPV’s - the artifacts and leakage is just awful. Yes, you can “tweak” some areas, aka sweep the dirt under the carpet, but the Issues are so large that once you finished cleaning up one side of the floor, a giant pile of garbage sticks out on the other side.

Try this https://forums.unrealengine.com/showthread.php?130970-Dynamic-GI-Getting-the-Most-out-of-LPV-(-Light-Propagation-Volume-

Most of the time GI is overkill. If you have a static daytime, static GI/Lightmass is fine and you can get pretty results with it.

Having a static daytime is quite boring.

Sure, but I am not asking about LPV for static lighting :wink:

I don’t want to put bad critique on that Tutorial, let’s just keep it to that: I’m well aware of that thread, it’s just scratching the Surface of LPV’s and his values for a few screws don’t change the result.

Epic was never working on LPVs, that was Lionhead. It was abandoned because Lionhead shut down. Epic does want to do a better system at some point, but it seems to be fairly low on the priority list after how long it’s been on the backlog. It’ll probably be another year or two is my guess before they get serious about it, after they fully exhaust what you can do with baked lighting.

Probably the same reason they don’t finish DFGI and HFGI resource constraints.

Looking at the backlog/missing features I’d say their rendering team is understaffed then.

The LPV “technology” is outdated as hell… Crytek dropped it years ago and replaced it by better solutions “voxelization”.
Sadly we talked here on forum like 100000 times about it… But it seems like Dynamic GI is not an important feature at all for epic.

The development of LPV had nothing to do with Lionhead shutting down, that was fairly recent. I’m pretty sure that LPV wasn’t just a Lionhead development either, I know they did some work on it but I don’t think it was all them.

This and probabaly lack of company push in this direction (noone is interested and they don’t need dynamic GI for their games).

This topic will be probabaly back at Epic when Fortnite will be under active development again (if ever). Since it’s the only game that requires fully dynamic lighting.

And this too. There is how many people working at rendering exclusively ? 3-4 ?
For engine of this scale is drop in the water. Crytek at their peak had entire R&D section dedicated for real time rendering (and say what you will, but it shows).

There are some new interesting GI techniques like Point Based Global Illumination (which could partially precomputed for static objects).
Distance Based GI is probabaly dead end (simply because it have limited range, which is even lower than Voxels). Although reusnig existing data structe instead of introducing new one incompatibile with rest, is always nice.

3 years now. I’m guess you guys are looking at otoy and vxgi?

Yes I agree with this lol… I suspect that there is no real need for many projects to use GI… people want GI because they are cool, but it is not really needed. They can just complete the games with static lighting baked. Maybe there are projects that cannot be completed with GI, but the number is very small.

The guy working on GI was relocated to work on something else;
GI will always cause bad performance, while they are focusing on increasing engine performance so there’s your dilemma…