Unreal Engine 4.18 Released!

[spacegojira]

I completely agree, it hurts to see how Epic is still working on improving static baked lighting. Epic just needs more devs working on lighting and shadowing. Every dev-rendering merge to master I hope to find improvements to dynamic lighting and shadowing, but recently, it was mostly all just static lighting stuff (apart from the awesome vulkan improvements, thanks Rolando!).

It feels like Daniel is the only one working at lighting and shadows, so whenever he works on static lighting, 0 people are left for the dynamic side. If Epic would have at least 2 devs for lighting with one constantly working on dynamic lighting, that would already be a huge improvement. But I get that finding people with the skills of Daniel is very, very hard. I just hope Epic is really trying to.

I second that, while I think static lighting has it's place, it is indeed really painful to see how neglected dynamic lighting is. And I see nothing that gives me hope of it changing anytime in the near future sadly.
At this point I am already thinking of switching to CryEngine (or the Amazon version) for the next project, even though it would be a pain to relearn everything.

Can 4.19 be focussed on dynamic lighting please?

[motorsep]

Can 4.19 be focussed on dynamic lighting please?

Do you mean dynamic shadows (as they are super slow, especially in VR) ?

[Zeblote]

Do you mean dynamic shadows (as they are super slow, especially in VR) ?

So dynamic shaders are just slow in general in unreal? And here I thought I was doing something wrong.

[DanielW]

I completely agree, it hurts to see how Epic is still working on improving static baked lighting.

I second that, while I think static lighting has it's place, it is indeed really painful to see how neglected dynamic lighting is.

Lighting development does tend to cycle between improving static lighting and dynamic lighting. There's so much to do and so few people.

Earlier this year we made a major push to get dynamic lighting features used in Fortnite running well on consoles and mid spec PC. We made a huge number of optimizations and cross platform fixes. Those dynamic lighting features are: Cascaded Shadow Maps, Cached Whole Scene Shadows (4.13), Distance field AO optimizations (4.16), Ray Traced DF Shadow optimizations (4.16), Volumetric Fog (4.16), LOD for many dynamic lights (r.LightMaxDrawDistanceScale), Capsule Indirect Shadows. That's a lot! Any UE4 project using dynamic lighting and shadowing benefits hugely from these optimizations.

Right now we're filling in some holes in the Static lighting and making sure that's world class. Focus on dynamic lighting will come back around again.

[Maximum-Dev]

Lighting development does tend to cycle between improving static lighting and dynamic lighting. There's so much to do and so few people.

Earlier this year we made a major push to get dynamic lighting features used in Fortnite running well on consoles and mid spec PC. We made a huge number of optimizations and cross platform fixes. Those dynamic lighting features are: Cascaded Shadow Maps, Cached Whole Scene Shadows (4.13), Distance field AO optimizations (4.16), Ray Traced DF Shadow optimizations (4.16), Volumetric Fog (4.16), LOD for many dynamic lights (r.LightMaxDrawDistanceScale), Capsule Indirect Shadows. That's a lot! Any UE4 project using dynamic lighting and shadowing benefits hugely from these optimizations.

Right now we're filling in some holes in the Static lighting and making sure that's world class. Focus on dynamic lighting will come back around again.

All that's appreciated.

I think current situation of baked lighting in UE4 already suffice vast majority of projects utilizing baked lighting. But when it comes to those utilizing dynamic lighting, there's really a big disappointment.

1. Sun (directional light) still isn't physically based. That's a major block in way of achieving realism and staying "correct".
2. There is no Time of Day editor.
3. CSM blocky shadowing still isn't fixed. See here and here.
4. There's no proper localized Image Based Lighting support.
5. LPVs are neither production ready nor receiving any major improvements.
6. For some reason ShadowDepths cost a ton as soon as a dynamic light interacts with a tessellated material.
Etc.

I get that nor Fortnite neither Paragon would've benefited from any of this given one is heavily stylized and the other is an arena shooter within an static environment. But majority of projects these days are leaning towards fully dynamic environments and thus while UE4's baked lighting is pushing to be world class, it's dynamic lighting features are still way behind 2007.

Not blaming you for anything. Just think there could've been more balance between the two.

[Syed]

But majority of projects these days are leaning towards fully dynamic environments and thus while UE4's baked lighting is pushing to be world class, it's dynamic lighting features are still way behind 2007.

Well I am quite skeptical with this.. because not many projects are actually needing this although it sounds cool. I definitely welcome Epic integrating properly fully dynamic environment lighting but most of the projects can do with static lighting, and therefore make themselves more available to lower end machine..

[Maximum-Dev]

Well I am quite skeptical with this.. because not many projects are actually needing this although it sounds cool. I definitely welcome Epic integrating properly fully dynamic environment lighting but most of the projects can do with static lighting, and therefore make themselves more available to lower end machine..

We can continue using baked lighting and create static environments for another decade however, it's not a matter of whether we can or cannot. It's a matter of the direction the entire industry is taking. Most engines these days are going towards fully dynamic lighting regardless of it being possible to use baked lighting.
Just saw this: https://80.lv/articles/ssrtgi-toughe...-real-time-3d/ Food for thought.

You can't just "not have the dynamic lighting features" and keep claiming most projects wouldn't need them. Nobody wants to spend an eternity on baking lighting for open world games so they're forced to not use it, but then dynamic lighting in UE4 even lacks the basics, as mentioned in my previous post. Today UE4 is ~10 years behind regarding dynamic lighting compared to other engines it's because baked lighting took all the focus and dynamic features weren't valued much.

I don't want to continue the debate here because it simply gets us nothing. But as a person who used to work with a 2007 engine and then switched to UE4, I instantly felt the ongoing lack of dynamic lighting features which is still going on year after year.

[SirToxe]

Thank you for using the native resolution for rendering the UI on Retina display Macs! It's just such a small detail but the editor is now a lot more enjoyable to work with without that ugly low resolution font rendering. And it feels kinda snappier. Well done!

[motorsep]

Just saw this: https://80.lv/articles/ssrtgi-toughe...-real-time-3d/ Food for thought.

That's Unigine Engine, made by a Russian company who is heavily into all kind of simulators. Thus, they have to have support for massive worlds and photorealistic real-time lighting. Those kind of projects pay their bills and so that's where their dev focus is. If you want to make a game with that engine, well, good luck with that (I had a license for v1).

While I am in no way oppose fast and good looking real-time lighting (GI in particular), UE4 is a cross platform engine made for games/entertainment first and foremost, and I don't think consoles can pull real time lighting efficiently. I'd rather have more features added to forward rendering and have high-performance dynamic shadows than real-time GI.

[motorsep]

I wonder when hotfix #1 is coming to 4.18..

[Maximum-Dev]

That's Unigine Engine, made by a Russian company who is heavily into all kind of simulators. Thus, they have to have support for massive worlds and photorealistic real-time lighting. Those kind of projects pay their bills and so that's where their dev focus is. If you want to make a game with that engine, well, good luck with that (I had a license for v1).

While I am in no way oppose fast and good looking real-time lighting (GI in particular), UE4 is a cross platform engine made for games/entertainment first and foremost, and I don't think consoles can pull real time lighting efficiently. I'd rather have more features added to forward rendering and have high-performance dynamic shadows than real-time GI.

I think you missed the point. I linked that article to show where advances in real-time lighting are getting at. I didn't imply Epic should give us dynamic GI right now. -high-performance dynamic shadows- isn't anything apart from what I said a bit earlier. Let me quote myself:

"6. For some reason ShadowDepths cost a ton as soon as a dynamic light interacts with a tessellated material."

So I said exactly that. And that's not going to receive any improvements as long as baked lighting is Epic's main focus due to their internal projects needing no advances in real-time lighting. Sadly, For some reason Epic doesn't want to try and get their feet wet on open-world / large-outdoors and that's the #1 factor why everything is lacking on that end. (be it open world tools or outdoor lighting).

My whole point is that UE4's baked lighting has got way ahead of it's dynamic lighting to a point where those who use baked lighting really achieve what they want but those who use dynamic lighting are all tied up by performance issues or lack of features.

[Zeblote]

Lighting development does tend to cycle between improving static lighting and dynamic lighting. There's so much to do and so few people.

Earlier this year we made a major push to get dynamic lighting features used in Fortnite running well on consoles and mid spec PC. We made a huge number of optimizations and cross platform fixes. Those dynamic lighting features are: Cascaded Shadow Maps, Cached Whole Scene Shadows (4.13), Distance field AO optimizations (4.16), Ray Traced DF Shadow optimizations (4.16), Volumetric Fog (4.16), LOD for many dynamic lights (r.LightMaxDrawDistanceScale), Capsule Indirect Shadows. That's a lot! Any UE4 project using dynamic lighting and shadowing benefits hugely from these optimizations.

Right now we're filling in some holes in the Static lighting and making sure that's world class. Focus on dynamic lighting will come back around again.

So, now that static lighting has achieved world class, you'll switch back to improving dynamic lighting for the next releases?

[PixelbyPixel.TV]

Has anyone had any issues with Steam not being able to create sessions? We ungraded today and it appears that sessions are failing to be created, although the create session callback returns true and loads the level.

Note: We found the issue. We were hard coding the session name with a specified FName, the patch notes mentioned there were changes with a new NAME_GameSession. Changing the Game Session to the #define GameSessionName has solved this. The blueprint Node destroys the game session by NAME_GameSession.

[John Alcatraz]

Well I am quite skeptical with this.. because not many projects are actually needing this although it sounds cool. I definitely welcome Epic integrating properly fully dynamic environment lighting but most of the projects can do with static lighting, and therefore make themselves more available to lower end machine..

I think the most successful games made with UE4 (Ark and PUBG I would guess, Fortnite BR regarding the amount of players too) use dynamic lighting. People want to have games with huge open worlds, and those have to use dynamic lighting. Also, almost every single game can benefit from a dynamic time of day, even games that are fully indoor with only light coming through the windows. So even games like Paragon or Unreal Tournament would immediately look more dynamic and interesting.

Baked lighting is severely limiting the available gameplay a game can have. For example can't have dynamic destruction of things like buildings, since its baked. So baked lighting should really be used for almost no games any more. I do see that it makes a lot of sense for architectural visualization and that's what these world class features that UE4 recently got are mostly used for I think. UE4 is definitely the best engine for any arch viz, because it has this really perfect baked lighting. But almost any existing and future game would be better with dynamic lighting.

[iniside]

Back to lighting.

I personally would like to get some middle-ground. In most games, most of environment is static (even if parts of it are destructible). But lights tends to be dynamic (moving player with light, time of day, etc).

What I'd like to see is some precomputed lighting solution that works with dynamic lights, but with mostly static geometry. Recently there were some interesting papers released about baking light transport in sparse way (so It doesn't take **** load of memory and space).

Fully dynamic GI solution is probably out of question (not enough man power, to maintain it), but precomputed one + Scree Space, cloud be nice. (and it cloud eventually replace lightmass).