4.17-4.18 LPV is completely broken. 4-19 - what's next ?

Dear Epic,

You are making an impressive and amazing product. I have been using it almost a year, i love it and i am close to release my own project.

My product is critically depends on Dynamic GI - Light Propagation Volume. I see this feature as one of the most important in UE.

I tried to upgrade my project to 4.17.2 and i’ve got following:

  1. LPV Intencity is not working as intended UE-51644 Unreal Engine Issues and Bug Tracker (UE-51644)

  2. Emissive material does not contribute to LPV. Can’t find a ticket in the bug tracker - but saw it before. There are a couple of topics around. It seems like related to (1) https://answers.unrealengine.com/que…usion-bug.html

  3. Finally got stable crash in the shipping build with LPV enabled(probably race condition) - UE-49205. This one is marked as fixed in 4.19 - but never checked. Unreal Engine Issues and Bug Tracker (UE-49205)

So i rolled back to 4,16.3 and still there. I can’t go further.

I do programming over 20 years and i completely understand software development, therefore i tried to move on the previous major version with hope to get stable engine. But i see that i can’t do it either. You add a lot of new features - but these critical bugs exist within 2 major versions and probably more - lots of projects, depend on LPV got stuck in 4.16. How can we develop our projects with such mess ? I can fix simple issues in engine by myself - but not these above.

Dear Epic, could you please pay more attention to fix critical issues for existing functionality.

Sincerely, Alex V.

Unreal Engine is unfortunately not a good choice if you want stability. Its development is completely guided by Epic’s internal projects. If Epic needs a bug fixed for something they are developing, then it might get fixed. If not, its going to stay broken.

Emissive materials contributing to the LPV or the Lightmass would be really cool, since they can be used to create static lighting, isn’t the lightmass considered static lighting ?

The 3rd has been solved, coming to 4.19.

LPV in UE4 is like Paper2D and alikes (isn’t really an official tool, someone submitted and its added, but nobody at Epic really support these tools).
In case of LPV, an external studio developed it then Epic integrated into the engine from a pull request; but Epic staff themselves never used it so Epic will most likely never touch this “feature”…

And guess what: LPV was developed by (if I remember this right) Lionhead studio. Lionhead is dead.
The rule of thumb is: If Epic Games isn’t using a feature you rely on then don’t expect support or even bug fixes.

Is there any real reasons why you are upgrading to 4.18? If everything is working as it is in 4.16, then might as well keep it. Stable is sometimes subjective, as ultimately it is you that make the call.

As for me, I am still using 4.15 and does not plan to upgrade as it already gives what I want. I may upgrade to 4.16 though if the time permits.

That’s OK. I guess Epic should mark such features as “experimental” and “not supported” then…But I see that this one has official documentation and contained in official release as supported feature.

Anyway, I think it is great and awesome for Epic to have such feature as dynamic lightning.

Of course - sticking on the same version of engine minimizes bugs - at least makes it known. But it increases the cost of your future upgrade on the head version since one day you must do it anyway if you want to use new cool features and develop long term project. So i usually use “one major version past from head” politic - it smooths upgrade and gives me opportunity to use cutting edge features.

Emissive materials contribute to the LPV in 4.16 - I don’t actually know about lightmass and static lightning since i use fully dynamic.

LPV was an Epic-feature I believe, it was in UDK. The stuff from Lionhead was just a few tweaks and fixes to make it work better when they were working on Fable Legends.

when was LPV ever in UDK? (pretty sure it wasnt)

It’s such a shame that LPV wasn’t brought to a usable state (cascades etc.) since it’s initial integration in 2015. Then we would at least have something now for bounce light in a fully dynamic lighting scenario.

These sound like statements made out of frustration, not based on facts. We’re working with one of (if not the) most powerful open-source engines available, for free. Epic contributes massive amounts of energy to their engine, integrating hundreds of pull-requests, implementing feature requests, fixing bugs, so-on-and-so-forth. Let’s not forget how much knowledge and training they make available via documentation and live-stream training sessions. Members from the Epic staff release content goodies regularly, as well.

Now do the big, paying studios get a priority in the grand scheme of things? Of course- Epic still has bills to pay and employees to feed, and a profit to be made. Welcome to the real world. Still… we the community reap plenty of rewards and free stuff. If someone doesn’t like something, they aren’t being handcuffed here. Plenty of other options, like GameMaker for those types.

Back on the topic of LPV/GI in general… We have a few options available now, and none of them are a magic bullet. I’ve worked with all of the options, obscure to “official”. One can definitely make well performing and amazing looking scenes using any of them, with enough research and tinkering. Are any of them a drop-in, one button solution to complete sandbox, open-world and indoor GI lighting? Nope- get to work and do what you can with you have, and pray to the gods that the magic bullet is being engineered.

Epic isn’t stupid. They’ve been around for at least a few days in this crazy industry. Chances are they know how powerful and desired GI is. While the whole debacle with poor Lionhead getting put on the chopping block sucks and surely threw a monkey in the LPV wrench, I’d bet Epic is hard at work or at least has their eye on a badass solution for then engine in the GI department.

I hope too that they invent best GI that works well for inside and for outside. If it still takes some time then so be it, perfection and stability needs time. Unreal easy learning curve even for non-programmer, openness, stability, future proof is why i want really commit myself only to UE4

Oh really? Okay… LMGTFY:

This is the current list of fixes that is coming to 4.19. Please review. Unreal Engine Issues and Bug Tracker (Search Results)


I believe this would be a full list of bugs in 4.19 :slight_smile:

First of all LPV is a product original created and made by Crytek and used in Crysis 2 and Crysis 3 this means Cryengine 3 back since PS3 gen.

Second point wasn’t in UDK, and no wasn’t made by Epic Games and as @**BrUnO XaVIeR **correctly said the UE4 version was made by Lionhead studio back in 2014.

As the last point @aue idea why you guys request this feature, the LPV need negative lights as the leaks are everywhere, as Crytek done and the way is mean to be done, and UE4 don’t support it <negative lights>. As sub part, the cost in Cryengine vs UE4 is way WAY WAY bigger like x3 times the cost of the original Cryengine 3 version, that without count the fact that is a tech from 2010 or 2011 really updated that even Crytek dropped.

You guys would do better in request a different tech

Well, there you have it. It’s something they want to do, but can’t currently be justified unless it aligns with internal needs. Sounds legit. On the community side, they are working on more highly requested stuff and the easier, less problematic things to handle. It’s on their radar though. Let’s not forget that we can still implement stuff, if we so please. Having access to the source code and all.

Epic might not be waving a huge flag that says “WE ARE WORKING ON THE BESTEST GI!!”, but I can’t shake the feeling that they are working towards it. Look at the work Ryan Brucks has been doing with ray marching, for example. Slowly but surely the engine is getting the legs it needs to provide awesomeness, in my opinion.

To quote a nerd, Inigo Quilez: “Writing a global illumination renderer takes one hour. Starting from scratch. Writing an efficient and general global illumination renderer, takes ten years.”

An efficient, robust, cross-platform GI solution isn’t something you whip up quickly. It isn’t something that is going to play well with an already established rendering pipeline necessarily, either. Look at how many people freaked out when the decision to switch over to the ACES filmic tonemapper was made. Guaranteed the same will happen when a GI solution is implemented. Previous projects won’t look the same, people are gonna freak, bla bla bla. That’s certain to be a factor in this situation as well: easing into it.

So time and resource constraints, legacy compatibility, and many other factors here. It’s not Epic dangling a carrot in front of us.

The documentation already labels LPV as “not ready for production”. If it’s mentioned anywhere as supported, those sources almost certainly predate the closure of Lionhead Studios and the sentencing of LPV to limbo.

You mean when they remove all the previous lighting in the entire engine and replace it with something that uses a completely different standard that is incompatible with all previous art styles?
Raymarching did not need any enhancement or modification to the engine, but it is only possible with custom nodes in the material editor.
While any GI implementation may take time, that it is explicitly not being worked on is what makes that hypethetical time period even more foreboding.

While I’m sure some optimism (or at least less pessimism) is warranted, somewhere, your justifications for optimism are so thin that it only encourages more pessimism.