Lumen | Inside Unreal

[VIDEO PLAYBACK] [MUSIC PLAYING] - Hummingbirds. Fighter Jets. Giant hornets. Can they fly? Yes, of course they can. Eggs. Eggs. Eggs. Can they fly? No, absolutely not. [HEAVY BREATHING] - Hey, Kevin. - Yeah, Daryl. - What do you reckon about flying? - Uh-oh. - Oye, Tess, stop bloody bouncing your sister into the coral tree. Bloody caviar. Sorry. What were you saying? - I was, um-- I was just talking about flying. - Oh, right. Right. Why? - I don't know. I just think it might be really interesting to see the world from that angle. - Wait. Are you saying you want to fly? - Well, maybe. - But you're an egg. - Yeah, I know. But I could wear a helmet or something. - Oye, Tess, stop putting the bloody decorative pebbles up your nostrils. Look, Dars, I just don't think eggs are meant to fly, mate. Ow. Tess, what did I just say? - Oopsie. - Flight Simulator, where you can fly anywhere you can dream of, as long as you are not an egg. Please verify that you are not an egg. - Melanie, do you reckon I can have a go at flying your kite? - I don't know, Daryl. Isn't it illegal for eggs to fly or something? - Well, yeah, maybe. I don't know, but I wouldn't actually be flying myself. I would just be holding-- - Oye, babes. Is this egg bothering you because I'll end him if he is. - No, sorry. I was just-- - Oh, my God. That's like so romantic. - Melanie, your kite. [MUSIC PLAYING] - All right, Humpty Dumpty. Do you know why you're here? - Um, uh, I tried to fly. - You tried to fly. And do you know what happens when eggs try to fly? - Um, they have fun? - This is the third time this year you've tried something like this. What's going on, Daryl? - I-- [MUSIC PLAYING] - Look, you're a good guy, Daryl. I'm going to let you off with a warning. But this is the last time, OK? You should really get some help, mate. [MUSIC PLAYING] - Egg, a vessel for housing tiny embryonic lifeforms that have evolved over millions of years to live in complete symbiotic harmony with one another. But can eggs fly? Yes, I think they can. AMANDA: Hey, everyone! Haven't had enough new features and updates to try out in Unreal Engine? Well, good news! Unreal Engine 4.27 Preview is now available, with new virtual production and in-camera VFX workflows, updates to architectural pipelines, and production-ready Pixel Streaming. Visit the forums for a full list of updates in 4.27 Preview 1, then download it from the Epic Games launcher or GitHub! Also ready for you in 4.27 is RAD Game Tools' cross-platform performance-oriented video codec Bink Video and their performance-oriented audio codec, Bink Audio. Both of these are now freely available via the launcher and GitHub--learn more about them on the feed. If you're ready to get hands-on with UE5 Early Access, we're hosting a series of livestreams on Inside Unreal where our dev teams go in-depth with UE5's latest features. Check out the past streams on Nanite and Lumen via the Unreal Engine YouTube Channel and join us in the coming weeks on Twitch or YouTube live, Thursdays at 2PM Eastern, for more UE5 content. Don't forget, Unreal Build: Broadcast & Live Events is next Wednesday, June 16! The free virtual event will feature sessions with deadmau5, FOX Sports CEO Eric Shanks, The Weather Channel, and more incredible innovators. Head over to the Unreal Build event page to see the full schedule and register! With digital fashion opening up new doors for designers, strut over to the feed to find out how Epic MegaGrant recipients Gary James McQueen and Moyosa Media created Guiding Light, their first all-digital fashion show, in Unreal Engine. Developer Sumo Newcastle asked themselves, ""What if Game of Thrones did Payday in the Robin Hood universe?"" Hood: Outlaws & Legends, a class-based PVPVE multiplayer heist experience, is their answer. Venture over to the feed to learn how Sumo Newcastle built and are continually balancing Hood: Outlaws & Legends and its novel mechanics. And now over to this week's top karma earners. Thank you to: ClockworkOcean, Mahoukyou, Everynone, Micky, Tuerer, Shadowriver, Luos, Andrew Bindraw, mindframes, and Bojan Novakovic. Get whisked away into an enchanting world with our first community spotlight, Palia, a community simulation MMO. Build a home, a life, make friends, and explore the mysteries of humanity's past. Learn more about Palia and sign up for their pre-alpha at palia.com. Next up, set in a post-apocalyptic future, the short-film AWAY, follows ""The Traveler"" as she scavenges through a burnt forest and the ""ghost city"" she lives in. Made to explore storytelling with LED volumes, the short was created in a single production day and garnered the team at ICVR 5 Telly Awards! See more incredible work from the team at ICVR.io. And last up is a stunning seaside setting, called Oblivion, by Thilo Seifert. Inspired by their camping holiday memories, the scene started as R&D for building an ocean shore shader. View more technical details on Thilo's ArtStation page! Thanks for watching this week's News and Community Spotlight. VICTOR: And we're live. Hi, everyone. And welcome to Inside Unreal, a weekly show where we learn, explore, and celebrate everything Unreal. I'm your host, Victor Brodin, and today, we're here to talk about Lumen. To help me in this endeavor, we have co-host and Senior Technical Producer, Chance Ivey. CHANCE: Hey, how's it going, everyone? VICTOR: And the star of the show, if I may, Daniel Wright, Engineering Fellow for graphics. Welcome. DANIEL: Hey, guys. VICTOR: And last but not least, Galen Davis, evangelist for Quixel. Welcome back. GALEN: Thanks for having me on. VICTOR: As always, I'm glad you're all here. I'm not going to do much of the talking. And so I would like to hand it over to Daniel. DANIEL: I thought you had stuff you were going to show at the beginning. VICTOR: No, that's it. [INTERPOSING VOICES] DANIEL: You said there was like seven things to show. VICTOR: We already showed all of that before-- CHANCE: We showed the fun stuff. VICTOR: Yeah. DANIEL: OK, we were busy chatting. VICTOR: We did that. It's your turn now. Lost track here. DANIEL: OK, it's my turn. CHANCE: Let's dive right in. DANIEL: OK, so I put together a few slides to talk about Lumen. And early access has been out for two weeks, so I had a little bit of time to write up some best practices with the way I see people using Lumen and maybe how you can get better results. To start out with, Lumen is-- we have a team. You might see the guys on the forums answering questions, Krzysztof Narkowicz and Patrick Kelly work on Lumen along with me. So let me just give you a quick overview of our motivation for Lumen. When we set out to make this thing, we wanted to solve the problem of dynamic global illumination. And instead of making something that was targeted at current gen consoles and trying to scale it up, we just kind of bit the bullet and targeted next gen consoles so that we could make something really high quality. But we did also set out to scale to high-end PC, especially enterprise use cases where quality needs to be top notch. That's also part of Lumen's goals. And we really wanted to provide performant reflections together with dynamic GI because it's really difficult to do dynamic global illumination together with reflections. That's not something that existed before in the engine. CHANCE: That's great. DANIEL: Lumen also needs to work with large open worlds, which is a big focus of Unreal Engine 5, particularly with Nanite. Nanite makes crazy levels possible, millions of instances. Just ridiculous content, really. And Lumen needs to work with all of that. It can't be the thing keeping people from making awesome levels. While all running in real time. And that means all of Lumen's algorithms had to stream and they all need to be GPU based. But indirect lighting can't just work outdoors. It's much more important-- indirect lighting shows up much more prominently indoors. And this also is by far the hardest problem for dynamic global illumination because the entire room can be lit by a very small area. And we have to find that area reliably in order to give good image quality. And it's not just enough to solve these individually, but we have to solve them seamlessly so that you can walk into a room that looks great and walk back out into a complex open world. So these have been the problems that we've been focusing on and trying to solve with Lumen. CHANCE: Daniel, since I'm curious as a not artist here, for indoor quality, by far, is the hardest problem in real-time GI. Why is that? Is it because of all of the different geometry that's kind of in an enclosed space and trying to get the bounce right off of things? Or lights affect more things that are closer to the camera? Is there a reason that that is, say, harder than an outdoor space? DANIEL: Yeah, it's that there are a lot more shadowed areas where only GI is on screen. Like in this screenshot here, the direct lighting is just in the bottom left. Everything else is 100% Lumen. Lumen skylight, Lumen global illumination. So every pixel, Lumen is providing all of the lighting for it. Whereas outdoor, this is not a middle of the day scene. But in what we shipped for Valley of the Ancient, because the sun is what you see the most. So we can get away with a lot lower quality there. And that's true of most outdoor scenes. CHANCE: Got you. DANIEL: So that's half of it, basically indirect lighting's more prominent. The other half of it is that to solve indirect lighting with good quality, we need to have very low noise. And that means finding where the incoming lighting is. And the incoming lighting is in a very small area for indoors. Like behind me, the whole room is lit up just from this one window. Whereas outdoors, you can kind of look at any direction and you'll find some lighting. But indoors, most of the directions are just black, except for that window. You'd better find it. CHANCE: I see. Makes sense. DANIEL: Some of Lumen's features. Lumen provides dynamic global illumination, which is lighting bouncing off every surface onto every other surface. And in indoor scenes with really bright base color, really reflective materials like white paint, it needs to bounce multiple times. And Lumen provides multiple bounces. If it doesn't bounce multiple times, you get to this kind of like unnatural dark look. And artists have to compensate for it with other techniques. This is showing how high quality Lumen global illumination can be if you crank everything up. This is not the default settings. This is a scene from Rafael Reis. Did I say that right? CHANCE: I believe so. DANIEL: You guys corrected me last time. This is a project that he's still working on. And I saw his screenshots on Facebook. And I reached out to him and I said, can I show these for the Lumen Livestream? Because it's just unbelievably good quality that he's created here. CHANCE: I'd say that he's still working, too. It's something we discussed yesterday. To my untrained eye, this looks amazing. Like really, really spectacular. Detail in here, everything looks like you would expect a real room to look like, right? DANIEL: And he did not bake lighting for this. CHANCE: Yeah, 100% dynamic. Wow. DANIEL: Lumen is solving this in a handful of milliseconds, where as to bake lighting, you got to wait for minutes and can't move anything. So thank you to Rafael for letting me show your scene. It's really cool. Lumen also provides shadowed skylight, which allows indoors to be much darker than outdoors. And when you place a movable skylight in your scene, it just works. Lumen picks it up, shadows it automatically and removes it from the interior areas. And Lumen does propagate emissive with a little star. CHANCE: Yeah, we've seen so many photos online of emissive only scenes and what. I'm sure we can dig into detail there later. DANIEL: Yeah, I have a few slides on like best practices with emissive. You can't just replace your light sources with emissive right now. CHANCE: Yeah. DANIEL: It has its limits. Basically, you need to keep the emissive subtle. But it does propagate emissive. And Lumen provides reflections, it traces extra rays to solve reflections if the material's smooth enough. And they integrate together with the dynamic global illumination. So in the reflection, you see GI, which is very important because otherwise, it would just be black everywhere except where the sunlight is. CHANCE: And a question on that. On the reflections, are those also casting the same color light back out and influencing other parts of the scene, too? DANIEL: No, the reflections are only influencing what's on your screen. CHANCE: OK, got you. DANIEL: There's no caustics, which is the sun comes in the window, bounces specularly off the ground, makes a pattern on the ceiling. Maybe that will be Unreal Engine 6. CHANCE: Right on. DANIEL: Lumen supports clear coat properly. So there's actually two layers of reflections, not just one, which is really important for car paint and some other materials that have a lacquer over the top. Some of the text is hanging off the screen. VICTOR: Yeah, I see that, actually. DANIEL: I think that's just this one slide. I think it just messed up. VICTOR: Everything else seems to look good. DANIEL: OK, good. So Lumen does provide dynamic GI and sky shadowing on translucency and volumetric fog. But it is much lower quality, which is something we hope to improve. But it is a much harder problem than solving GI on opaque pixels. Because there's arbitrary number of layers for every pixel. So now, I'm just going to go over some of the Lumen settings that you'll see in early access. If you make a new project, Lumen's on by default. It's default dynamic GI method for the engine. But if you bring over an existing project, then you just need to go into Project Settings and set this one thing, which you can find really quickly if you do a search. Just set Dynamic Global Illumination Method to Lumen and it will automatically set the other dependencies, which are these three guys. These are the things that we change to be on by default in Unreal Engine 5, which is Lumen GI, Lumen Reflections and Generate Mesh Distance Fields for Software Ray Tracing. CHANCE: So that first one would be more about if you're moving a project over, and then if you want to take Lumen and turn it off or toggle it back, this would be where you would check? DANIEL: Yeah, you can do it here for the whole project. You can toggle between Lumen and not Lumen. CHANCE: Got it. And project wide settings, too? DANIEL: Yeah. CHANCE: That being in projects. DANIEL: Yeah. Yeah, all the project settings are for your whole project. So that you can also change the method in post-process volume for just one level. I'll show that in a couple slides. CHANCE: Super neat. DANIEL: Here's some other important project settings, which I'll talk more about later. But these are the ones that you need in order to get top ray tracing quality from Lumen. You've got to turn on Support Hardware Ray Tracing, which makes it so that we compile all the shaders and build all the structures to support hardware ray tracing. And then you have to tell Lumen to use hardware ray tracing, which you can toggle this one on and off and see the influence of it. CHANCE: Got it. DANIEL: Lumen also has a couple more settings in the post-process volume. You can change the method per volume. So you can have one-- we use this in our QA stuff all the time, where this level uses Lumen and this other level uses something else. And then you can crank up the quality of Lumen. This is mostly meant to be like, in this one hallway I need higher quality. It's not meant to target different platforms. That happens somewhere else. That happens in the Scalability.ini file. CHANCE: So this would be more about like, you're building your scene, your building maybe a game or something where you're going to move through a space and you've got the budget in this space. So things are closer to the camera, you want to just right now spend those extra milliseconds on bumping these numbers up using a post-process volume? DANIEL: Yeah? CHANCE: Got it. DANIEL: And like Arches projects that generally don't have a ton of different maps. Project Settings are really powerful because you're automatically setting across all the maps. If you're in a big production, people are making test maps right and left and you want to make sure they all have consistent lighting settings. That's why you get to have the project setting. But if you're not in a huge project and you can just control it all yourself, you can just do everything through the post-process volume. So Lumen doesn't have that many controls, you notice. And not that many things to fiddle with. Actually, because Lumen is a lighting simulation, most of the inputs to Lumen are actually coming from your light settings-- your existing light settings, your material values and the exposure you setting on the camera. And then Lumen is taking those and propagating them through the simulation and bouncing the lighting. CHANCE: That's great. DANIEL: So let's look behind the scenes at how Lumen actually works. So just a high level peak. From the high level, Lumen uses software ray tracing by default, which is something we've developed for Unreal Engine 5. And it can also support hardware ray tracing, which I'll talk about a little bit later. First, Lumen traces against the screen-- the depth buffer. And if that misses-- sometimes those rays go behind something or they go off the screen-- then Lumen traces against an alternate representation of the scene that we maintain, which has signed distance fields for meshes. And then once the ray hits the signed distance field, then we get the lighting for that hit point with a surface cache. You guys have any questions about that? Does that make sense? CHANCE: As a non-rendering, non-artist person, I think that it's a little bit over my head for this one, but I get it. DANIEL: I have some nice screenshots to show to hopefully show it a little better. So Lumen doesn't just use one type of ray tracing, it's a hybrid ray tracing pipeline. So screen traces, like I was saying, when they go off the screen, they can't find anything. So then it switches to the next method, which is tracing the individual objects distance fields, which is more reliable than screen traces because they work off screen. But there is a cost to those traces. So once the ray gets even further, then we only trace against this global representation where we've merged everything down. And those traces are extremely fast. And that's what's called the Global Distance Field Trace here. So there's no way to see this visualization. I just hacked the code to show this-- to let you see behind the scenes. But this is what ends up happening for a single scene. Each pixel's GI is solved through a combination of different types of traces. CHANCE: Are some of these some of the dynamic settings you've built over the years for UE4? You're just able to take some of those-- some of the same processes and apply them all together to get the best rounded out result? DANIEL: No, we developed a lot of new software tracing technology. Our screen traces are all new. They're hierarchical z-buffer traces, didn't exist before. Our mesh distance field traces-- we had distance fields in UE4, but they were not sparse. This is my next slide. And our global distance field is way better, too. It's also sparse now. So we've basically rewritten global distance field and mesh distance fields to be way better in UE5 for Lumen. Those were used in UE4 for distance field shadows and distance field ambient occlusion. So our mesh distance fields are way higher quality in Unreal Engine 5, which we rely on heavily in Lumen. We now have a sparse representation. So we're only storing data where we need to, near the surface. We generate mipmaps of those distance fields so that we only need to load the ray version depending on how far away you are. And as a result, we were able to double the default resolution, which since it's 3D, actually means eight times more data. Because it's doubled in x, doubled in y, doubled in z. And four times higher max resolution per mesh, which is really important to supporting those large meshes that were failing with distance is fields in Unreal Engine 4. CHANCE: I'm sure that helps handle the hyper high-res Nanite assets as well a little bit better. DANIEL: Yes, it makes the system a lot more robust. If you just go and place a mountain, it works much better now. In Unreal Engine 4, there were a lot of artifacts if you placed a large object with distance field ambient occlusion. Since we increased the resolution so much, you'd expect that it got at least 10 times more memory. But actually, it uses about half the memory of Unreal Engine 4 because of streaming and the sparse representation and a bunch of optimizations I did. CHANCE: Interesting. DANIEL: We also made the distance field building 10 times faster, which is kind of necessary. Because once we increase the resolution, it became quite a pain point. But it's 10 times faster now so that it's kind of like hidden in the background for the most part. And we improved this mesh distance field visualization so you can actually get a better sense of what the distance field scene looks like. It does some simple lighting now using the distance field normal and the hit position. CHANCE: Got it. On those improvements there, are those Lumen specific? Or would they affect the DFAO as well? Or are they just mostly for Lumen? DANIEL: They're not Lumen specific. Distance fields are usable by other rendering features. CHANCE: Great. Cool. DANIEL: Like I said, they were very much developed for Lumen so the benefit isn't quite seen in other features as much as in Lumen. So distance fields only give us, basically, what's in this visualization. They give us a hit point and a normal. But how do we actually know what was the material color, what was the incoming lighting there so we can bounce it to your eye? Well, distance fields can't give us that so we had to come up with a different solution. And what we came up with is what we call the Surface Cache. And it's basically, capturing meshes from different orientations and storing them off into an atlas with all of the material properties sorted out. And we maintain this as you fly around the scene. We're recapturing all the time. I got to say something about the Surface Cache, it has one big content limitation. And this is the biggest content limitation of Lumen, which is that the Surface Cache can only work if your meshes have simple interiors. The walls need to be separate meshes, they can't all be in one mesh. And it's not something we're happy with, obviously. But it is a current limitation. And I can't remember what else I was going to say about that. CHANCE: It's all good. How big is that big sheet there in memory? DANIEL: It is a 4096. CHANCE: 4096, OK. DANIEL: These are all cached so we're not recapturing them all every frame. We're going to be recapturing a very small portion. Basically, what we can afford to cache-- to recapture. And they're kind of recaptured-- as you move around, we recapture them at a higher resolution. Or if you move away, we recapture at a lower resolution. And these captures, these really small, low-resolution captures are actually stupidly slow without Nanite. Because if the mesh doesn't have a really good LOD setup, it's going to process way too many triangles to render at like 8x8. And also just based on the limitations of the hardware rasterizer, you have to render into one view at a time. But Nanite, since it's a custom rasterizer, it has a multi view feature. So we can just render it-- we just blast into all of the different views for each different projection of the mesh at the same time. And we measured, it's between 10 and 100 times faster, depending on how inefficient the original content was to the point where you basically need Nanite. If you're going to use Lumen with any high-poly assets, and especially instanced static meshes, they need to be Nanite. [INTERPOSING VOICES] GALEN: Oh, sorry. Go ahead, Chance. CHANCE: No, no, no. It's good. GALEN: I was just going to say maybe it's worth lingering here for a second, because-- I mean, we just had our livestream about Nanite last week. And some of the limitations there with the tech as it currently exists for early access. So what if I'm making a scene just hypothetically with lots of things that are moving that are not Nanite? What would you kind of prescribe, I guess, for a scene that maybe has a handful of Nanite assets, but there's also a handful of assets that you know for sure are definitely going to be moving and therefore, not Nanite. DANIEL: The main thing we need-- that Lumen needs to be efficient is to have LODs set up. If you have a million poly mesh with no LODs, it's going to be slow. But if you use instanced static meshes, it has to be Nanite. There's no way around it. The performance increase for Nanite is the only way to make it fast enough to use. GALEN: Makes sense. CHANCE: And you can use Nanite on lower-poly meshes, too, as we discussed last week, too. It doesn't have to be a super high-- [INTERPOSING VOICES] DANIEL: You should definitely use it. Because of the GPU driven rendering that Nanite uses, it can blast through-- there's very little overhead per instance. So after we've captured the material properties for the surface cache, we still need to light them. So we go through and apply each light to all these little cards in the surface cache. And we also do a GI gather in the surface cache. This is how we provide multibounce in Lumen. And that GI gather is not particularly high quality right now. And I'll show you some artifacts from that later, which we hope to improve. So you can see the combination of the mesh distance fields and the surface cache. You can see what they look like by going into the editor: Show -> Visualize -> LumenScene. And this is basically what Lumen sees while it's doing a global illumination or reflection trace. If the screen trace doesn't have anything or it goes off screen, then Lumen picks up the ray and traces in this representation. So this is like the most important view mode for Lumen. Because if your scene in the LumenScene doesn't match what's on your screen, there's going to be view-dependent artifacts in GI. So this is your number one tool to seeing what is Lumen doing? Why is there an artifact? CHANCE: Yeah, and you can A/B this with what your-- you can toggle this back and forth to make sure that it looks roughly the same color-wise and size-wise and whatnot, right? DANIEL: Yeah, it's really convenient because if you use the G key in the Editor, you could just flip back and forth between them. Because you can have different show flags for like the G-mode and non G-mode. I'll show you some of that later. So these mesh distance fields, they're pretty accurate. Not as accurate as the raw Nanite triangles, but they're pretty accurate. But they do cost something to trace through, because the cost of tracing is dependent on how many meshes the artist placed. If they placed a ton of messages, especially overlapping meshes, each trace has to go through all of them. And so we have an optimization for software ray tracing, which is to merge them all together into a global data structure. We merge all the mesh distance fields into a global distance field. And in Unreal Engine 5, the global distance field is double the resolution as Unreal Engine 4 and very efficient cache implementation. Not something you need to know but behind the scenes, it's better than what was used in UE4 for distance field ambient occlusion. CHANCE: Right on. So when I asked earlier about using these things for DFAO, it doesn't really matter as much if you're just going to use Lumen going forward? DANIEL: Yes. Yeah. I do hope we find other ways to leverage distance fields, but Lumen is the primary user. So on top of merging distance fields, we also merge the 2D surface lighting into a voxel representation. And that's shown here on the right. The left is actually tracing against the mesh SDFs, getting a really accurate hit, looking up into the 2D atlas. And on the right is the merged Voxel version, which is so much faster to ray trace against but more approximate. So that optimization is always used by default after two meters. We trace against the individual meshes for the first two meters, and then we switch and we trace against only the global merged version, which makes software ray tracing very efficient. But in some crazy scenes where you just have a ton of meshes overlapping, with this project setting, you can tell Lumen software tracing, only trace against the global version. And that's very fast. CHANCE: Got you. And that's also just there in Project Settings? DANIEL: Yeah. Software tracing has some limitations. It's kind of like to make this stuff real time, we had to limit our scope of the problems we were trying to solve. We only support static meshes and instanced static mesh. Screen traces work against anything, but the distance fields only exist for static mesh and instanced static mesh. And we don't have Landscape. We know we have to add that. That's not intentional, just didn't get to it in early access timeframe. In 5.0 release, if you're building an Unreal Engine 5 project, you can use Landscape with Lumen. You don't need to worry that like we won't add Landscape support. We definitely will. I don't usually make promises about future releases. But that's one-- Lumen has to have that to be successful. World position artifacts-- world position offset causes some artifacts with Lumen because we cannot replicate per-vertex deformation in the distance fields, which were built off of the non-deformed version. And we don't have a good solution for that right now. What we hope to do is the same thing distance field ambient occlusion has where there's just like a bias you can provide and say, don't self-shadow within this distance. That allows you to do some sway without having it self-intersect and get all black. CHANCE: I see. DANIEL: And there are a bunch more limitations. By the way, the Lumen documentation is really good. Check it out. We spent a lot of time on it. We were very thorough with all the limitations and stuff. And you can find the full list of limitations in the docs. CHANCE: Well, thank you for doing that. I'm glad we have docs at early access. That's super great. DANIEL: Yeah. So that was all software ray tracing. Didn't require any special video card. It ran on any DirectX 11 hardware or greater. And has options to run super fast. But supports limited geometry types. So let's talk about hardware ray tracing in Lumen. So here's how you turn on hardware ray tracing for Lumen. You have to tell the project to support hardware ray tracing, and then you have to tell Lumen to actually use it. And you can toggle this one on and off and see what it's doing on the fly. So hardware ray tracing, high level, it can provide the highest quality, but it also costs the most. And you need a special video card to run it. And on PC, it only will work if you're running under D3D12. And the video cards that you need that support hardware ray tracing are either Nvidia RTX 2000 series or higher or the AMD RX 6000 series, which are both very recent video cards. And hardware ray tracing in early access, Lumen's use of it is still partial support. We support it for reflections and for part of the final gather, but we're still working on supporting it in all of Lumen's facets. So you actually need distance fields and hardware ray tracing in early access for Lumen to work. Nanite enables an order of magnitude-- several orders of magnitude higher detailed geometry. But it accomplishes this through some rasterizer specific techniques and decompressing the vertex format on the fly, which are not compatible with hardware ray tracing's limited API and representation. So we cannot hardware ray trace against the raw Nanite triangles. Instead, Nanite provides a proxy geometry version, which is a simplified version of the geometry. And hardware ray tracing operates on that. And most of the time, we can cover over this mismatch with screen traces because what's on your screen is the actual Nanite rasterized geometry at full resolution. So the screen traces are tracing against the full resolution version. And then once it goes behind something, it gets blocked, then it switches to the proxy geometry. But in some cases, where you have something really important like a car, the proxy geometry will cause some leaking. And then you can raise the Proxy Triangle Percent 200 to get rid of that leaking and have the hardware ray tracing use the full res version. CHANCE: And to be clear, this would be Nanite settings to get rid of some of these artifacts, correct? DANIEL: This is on the static mesh under the Nanite category. CHANCE: Got it. OK. And this would be probably like really case by case, something that you might encounter at some points if you go asset by asset. Not something that you should consider project wide as part of a workflow. But just kind of as you find things, just kind of-- OK. DANIEL: In fact, there is no project wide setting. So you can't-- CHANCE: Well, I'm just saying-- DANIEL: You can't just set it in project like-- CHANCE: Yeah, as people build their pipelines, start importing assets and say, oh, we heard this thing or we found this one asset, there's a hole and we changed this thing, so now every asset we want to change this to be-- up the [INAUDIBLE] proxies. It's not necessarily something we would encourage. It'd really be more about, hey, just using it like you normally would. And then as you find issues, this is a good place to look to try to overcome those. DANIEL: Yep. So you'll see in the LumenScene visualization, you'll see what those holes are. And then as you increase the proxy triangle percent, you'll see that it gets fixed. In this Lake House level from Rafael, which I've been testing with hardware ray tracing, I didn't have any issues with the Nanite proxy geometry at all. So I didn't have to do anything to fix any artifacts there, even though what we're ray tracing against is a simplified approximate version of the triangles. You'd expect on a curved surface, some of the rays would self-intersect, but actually, the screen tracing is doing a pretty good job of escaping the proxy geometry. GALEN: A quick question actually. So for the last two projects that we've done internally here at Epic, I'm just curious if you felt like you've seen diminishing returns as far as increasing the percentage of the proxy geometry for Nanite assets with regard to Lumen? Is there some-- I know that we must have been using kind of Megascans assets for those last few projects, so I was just curious, do you feel like there's kind of a sweet spot or some area where you feel like we're seeing diminishing returns once we get past a certain percentage for those setups? DANIEL: For the most part we just use the default Nanite proxy triangle percent, which gives you like 2,000 triangles, about, per mesh. And it's been good enough, except on cars. That's the main pain point. Because cars have very thin surfaces, and especially if you start from a very high-poly mesh, the Nanite proxy geometry will be a heavily simplified version of that. And once it gets simplified to that degree, some cracks will show up. And we may improve these in the future, such that this is not even an issue anymore. But currently, we generally don't have to touch this, except on cars. GALEN: OK, yeah. Because I'm just curious if we're saying, most like 2,000 triangles per asset, increasing it to say like 20,000 per Nanite proxy, would you expect to see huge return from that being 10x or no? DANIEL: Do you mean in terms of quality? GALEN: Yeah, in terms of quality for Lumen, yeah. DANIEL: Generally, if you just-- I think the default is like 1%. The proxy geometry is 1% of the original. If you set it to 2% or 4%, it will fix all the bugs. So you don't have to go all the way up to 100, you just need to go up a little bit. GALEN: Cool. CHANCE: So that's a very little bit. I guess it's double. But I mean, still, 1 to 2 is not that big of a jump. DANIEL: But it's just on an individual mesh, yeah. CHANCE: Yeah. DANIEL: I mean, people are not really going to have to touch this setting. I just mentioned it for full disclosure. And I think if you're making a scene that has to be perfect, ArchVis, then this is a setting you need to know about. But otherwise, you pretty much-- it just works. GALEN: Cool. Awesome. DANIEL: Some other things to know about hardware ray tracing is that it does not scale to Nanite-size scenes. And in early access, you need to keep your scene kind of small at less than 10,000 instances. But this is something we're actively improving already. And actually, in the main branch, we can already afford 100,000-- 10 times more. And hardware ray tracing does support skeletal meshes, which is one of the big draws of it over software ray tracing. However, those skeletal meshes actually cost a lot. Because it's dynamic geometry, we have to rebuild the ray trace acceleration structure for every single frame. So it's very important to keep what ray tracing is building for them, very low poly. And we're working on a ray tracing LOD bias so that what's on your screen can be very high quality, but what we're ray tracing against can be a lower quality. That does not exist in early access yet. CHANCE: And just to clarify, because as not an artist, you say that it's more expensive on skeletal meshes or skinned meshes. That's not because they move, because they deform some, right? Because the amount of surface area that a color might actually show in a scene is more or less, depending on its state. Or is it more about movability? DANIEL: It's entirely about deformation. CHANCE: Yeah, OK. DANIEL: So the ray tracing is accelerated by a bounding volume hierarchy. And for static geometry, just build it once at load and then you're done. But for dynamic geometry and deforming stuff, you've got to rebuild every single frame. And that is very expensive. The last thing about hardware ray tracing performance is that if you want to use it, if you want to build your scenes so that you can use it, you cannot overlap your meshes to a crazy degree. You cannot copy/paste meshes and have them all overlapping. Because hardware ray tracing has to go through every single one of those that's overlapping every single ray. And it becomes incredibly slow. So hardware ray tracing overall, we see is about 50% slower than software ray tracing, while being more accurate. But once you start overlapping meshes, then it can become significantly slower than software ray tracing. So this is something to keep in mind as you're building your scenes. If you plan on using hardware ray tracing, this is something you need to keep in mind. CHANCE: Kitbashing has huge costs. Don't tell Daniel how we built Valley of the Ancient. GALEN: Yeah, I've never seen this screenshot in my life. DANIEL: Hey, I didn't want to out you. CHANCE: No, it's all good. DANIEL: You guys mad about that screenshot? CHANCE: No. It's fine. We talked yesterday with-- or last week with Brian about how much we learned on the project. Kind of taking a lot of these new technologies and throwing them in and seeing what we could actually do with them for early access, right? And so just like with Nanite, there's a lot we learned about what we can and cannot or should and shouldn't do with Lumen to stay within performance budgets. DANIEL: I think it's pretty cool that even though there's 15 layers underneath her foot, we have a software tracing solution that gives good performance in this situation. I have a pretty crazy thunderstorm going on. VICTOR: There's a thunderstorm going on, Yeah. I was going to say, if I drop or me and Daniel drop, that's why. CHANCE: Vic, Daniel, and Chance... DANIEL: My plants are getting a good drink right now. OK, so I talked about how Lumen traces rays to find light. But I didn't talk about how we actually provide global illumination and reflections. So the Final Gather is the name for the process of propagating the light from the LumenScene to your final pixels on the screen. And figuring out that light with a very low amount of noise, because we can only afford to trace about half a ray per pixel, which is a very small amount of rays that we can afford. Where, to actually provide noise-free GI, you would need about 200 rays per pixel. Because ray tracing is so expensive, each individual ray costs a lot. You have to make the best use of it. CHANCE: And the difference there, too, just to understand the half, we can only afford this, but you need 200 for these things, again, is it indoor because you have all that bounce from a single kind of location? It's filling up all that space and you're really close to it generally? DANIEL: When you go to trace the ray, you genuinely don't know where the lighting is coming from. Could be from anywhere. It's very different than direct shadowing where the engine has a list of all the lights and it just can go through them and send a ray to every single one. With global illumination, because it's the entire scene that's bouncing light into you, you don't have a list of them. You have to just send rays out and see what you find. So the smaller and the brighter those areas are, the harder it is to solve, the more noise there will be in the final image. And that's why indoors need a lot more rays to find that lighting than outdoors. So we have this problem where we need 200, we can only afford a half. How do we make this 400 times faster? And some existing-- [LAUGHTER] CHANCE: This is only, right? DANIEL: This is why dynamic GI is such a hard problem. And existing approaches mostly fall into two buckets. Either people use irradiance fields where they do ray traces from a very small set of directions that are spread out far apart. And then they have to do a whole bunch of stuff to hide the fact that they're not actually tracing from the pixels on the screen. They're tracing from over there somewhere, which might even be on the other side of a wall. They have to store occlusion in the probes. They have to interpolate it. They usually require the artist to set up volumes around the level manually. And if you have a little complex area, the artist has to go and put more probes there. They also usually have a very slow lighting update speed. If you change the time of day, they need to update very slowly because, otherwise, dynamic-- because they're not tracing for very many positions, a dynamic object going through the probe would cause a lot of artifacts that would just get spread all over the place. So irradiance fields are great for performance. But they are very limited in quality. Not what we were looking for for Lumen. And the other dominant approach to solving the final gather is a screen space denoiser. This is where you just trace one-- which is all you can afford-- one ray per pixel distributed randomly. And then you have an extremely noisy image. And then you try to go and look at the neighbors on the screen and reuse them to get something that's more stable. But the high level problem with this approach is that when you're denoising after the fact, you've already traced the rays, it can only do so much. I'll show you a comparison in a little bit that's just showing how noisy the input is with screen space denoisers, which makes it very difficult for them to overcome to actually produce good indoor quality. So for Lumen, we went in a different direction. And we used screen space radiance caching, which allows us to trace from very small set of positions on the screen. We generally downsampled by about 1/16 and do more to ray traces at each of those positions. So each probe in the radiance cache has much more stable lighting. And then we interpolate that to neighboring pixels. And that allows us to do way more rays per point to solve to get stable lighting. And then we do the integration with the Nanite super-detailed normal at full resolution so that the final lighting is super detailed. CHANCE: Another reason these work so well together every time. DANIEL: Yes. And it informed us on our Lumen algorithm design. We knew from-- some lightning. We knew from day one with Lumen that we had to support extremely detailed geometry. And that's another thing that screen space denoisers fall down at. You can only reuse other rays on the screen that came from similarly facing geometry. You have to have a normal weight in your filter. And that basically rejects everything with Nanite geometry where every single pixel has a different normal. Lumen also has-- to get really stable distant lighting, we also have a world space radiance cache where we placed a much smaller number of probes and trace from those and reuse those for many pixels. And you can actually visualize that world space radiance cache, see those probes in the world, and what Lumen is gathering for them. So this is just for distant lighting. A lot of people are used to seeing a probe grid like this. But other techniques use this as their final lighting. Lumen does not. This is just distant lighting. And we're still ray tracing from the pixels on the screen, which is how we're able to get such detailed and direct shadows. Just to summarize everything, why Lumen gets such great results indoor quality versus a screen space denoiser is that basically the input-- the raw traces that come out of-- the result of the raw traces with Lumen is already so much less noisy that once we go to do our filtering, we don't have to filter very much. So here on the right, Lumen is doing a quarter of the traces of a screen space denoiser and already has a much more stable input. This is not the final image. The final image will be after filtering. But you can see that Lumen already has-- the Lumen filtering has a lot less work to do. And I'm going to be presenting on the Lumen Final Gather techniques this year at Advances in Real-Time Rendering at SIGGRAPH. This is just super high level. VICTOR: And we'll make sure to let everyone know when that's happening so they can tune in if you're excited. DANIEL: Yeah. Another slide deck to make. CHANCE: It's OK. You do lots of them. DANIEL: I'll use some of these. VICTOR: I'll make sure to send you the updated template with the new logo for next time. DANIEL: Hey you didn't say anything last night. VICTOR: I didn't. DANIEL: I don't know if you can complain. VICTOR: That was not a complaint. Like I said, I will make sure that I send you the updated template next time. CHANCE: Yeah. We just love that Daniel's here, going legacy on us. DANIEL: They said it's like every quarter, there's like a new template-- a new Epic template. And I hate 75% of them. So I keep coming back to this old one because it's just like it hides in the background well. It doesn't distract from what you're trying to show. Some of the latest ones, one of them was bright yellow text on a black background. It's like offensive to your eyes. VICTOR: Coming from marketing here, I'll make sure to bring your feedback back to the team. Because, honestly, I do find it valid since you are one of the programmers who are producing our tools here. CHANCE: Alan Noon in the chat from our team, I feel attacked. DANIEL: Did he make it? CHANCE: I don't know. VICTOR: He's here. CHANCE: I do find it a bit awesome that it's one of our rendering engineers about lighting that's talking about how bright the text is on a dark background. I think that that is-- DANIEL: And past a certain contrast level, your eyes-- it just distracts you from the content of the presentation basically. Anyway, I'll get back on track now that I've finished ranting about that. So Lumen Reflections. So I talked about the final gather. I talked about how we do ray tracing, I talked about the final gather, how it uses that ray tracing to produce very low amounts of noise in our GI. And now, I'm going to talk about how Lumen solves reflections. So Lumen Reflections, we reuse the traces from the final gather if the material is very rough, because we already did that work to figure out what's the incoming lighting. We already did a bunch of work to make it low noise. And if you think about it, as the material gets rough, the specular lobe gets wider and wider. And eventually, it just converges on diffuse, which is a full hemisphere. So that's one of the reasons why Lumen Reflections and Lumen GI go so well together is that half the reflections on the screen are actually being provided by Lumen's final gather. For the rest of those pixels that actually have a very smooth material, we trace extra rays to solve the reflections that are distributed just in that reflection cone. And that does increase the cost of Lumen. It shows up in Lumen Reflections. CHANCE: Got it. DANIEL: And then Lumen reflections has a spatial and temporal denoiser that runs in. And given only one ray per pixel, it tries to provide you with like very smooth glossy reflections. By default, Lumen Reflections-- even if you turn on hardware ray tracing, by default, when the ray hits something, we get all of our lighting from the surface cache, which is very fast. And makes Lumen Reflections about twice as fast as if we had actually evaluated lighting at the hit point. So it's a very good scalability option. But it is on out of the box. And I saw some people reacting to that. If you test Lumen Reflections by default, it's not as high quality as ray traced reflections. Well, if you set Lumen Reflections > Quality to 4-- magic value-- you get-- on the hit point, Lumen will go and actually evaluate the lighting at that hit point, which gives a much more accurate lighting and reflections. So if you are-- just to compare against ray traced reflections, if you want to compare between the two, make sure you set the Lumen Reflections > Quality to 4. And the reason that's a hidden setting is because it's early access and it came in hot. GALEN: Did I remember you saying something about reflections related to HLODs also, or did I completely invent that in my brain? DANIEL: It's a work in progress. GALEN: OK. I just wanted to check. DANIEL: It's not in early access. GALEN: OK. DANIEL: Some other things about Lumen Reflections versus ray traced reflections, right now, Lumen does not have all the same features as ray traced reflections, but we're working toward that. And we're working toward Lumen being the unified reflection method of Unreal Engine 5 that works with Lumen GI that works with Lightmap GI. Right now, Lumen has a bunch of things that ray traced reflections don't, especially in dynamic scenes particularly. If you use movable lights, ray traced reflections don't really work, especially movable skylight because the skylight will not be shadowed. Lumen has screen traces integrated so things that don't exist in the ray tracing scene will still-- Lumen Reflections will hide that mismatch quite well, specifically Nanite meshes. Since we're not ray tracing against the actual Nanite geometry, ray traced reflections can have a lot of self-shadowing artifacts with those. Lumen Reflections also support software ray tracing. So it's not like you have to have a special video card on PC. Any DirectX 11 hardware. And we have dynamic GI and shadowed movable skylight in reflections through the use of the surface cache. And Lumen Reflections support clear coat properly with actually two layers of reflections. The sharp one on the top with the lacquer, and then the glossier one underneath, which is not something that ray traced reflections supports. But ray traced reflections supports lightmap GI in the reflections. So you can use lightmaps for your GI together with ray traced reflections. You cannot do that yet with Lumen Reflections. That's something that we want to support for 5.0 release. So long term, looking to the future, ray traced reflections are deprecated. And we're going to merge all of their remaining functionality over to Lumen Reflections, which is our scalable, unified reflection pipeline for Unreal Engine 5. CHANCE: Right on. DANIEL: I wrote up some best practices from observing how people use Lumen and from our own projects developed using Lumen before early access release. First thing is, basically, the way that you can use emissive with Lumen and expect good results. If you try to actually use emissive meshes to replace light sources, you're going to be disappointed with the results with Lumen. Or another way to look at it, this is just a feature that hasn't been implemented yet. Lumen does pick up emissive, but the smaller and the brighter-- because Lumen is accepting lighting from any direction, it doesn't know where the emissive meshes are, the smaller and the brighter the mesh, the noisier it will get. So you can use emissive with Lumen, but you need to kind of keep that caster large. And you need to keep it dim. [INTERPOSING VOICES] DANIEL: Go ahead. CHANCE: You would still want to go through and maybe put a point light or two in those spaces to really get the lighting you want in that space? DANIEL: That's the other way you can use Lumen emissive is you actually place a spotlight or point light and you have a little emissive light bulb, but you just keep it dim. You don't crank it up so that it's luminous trying to propagate it through the scene. Those are basically the two uses of emissive that we see that are successful. CHANCE: Is there any problem with blowing it over the value of 1 to get some bloom on that, too? DANIEL: No, but as you crank it up, you'll start to see noise appear in the scene. Like this, where-- I mean, this isn't even the worst case. You'll start to see a lot of noise appear. CHANCE: And even a red X will show up. It's crazy. DANIEL: Right on. Yeah, I programmed that in there. CHANCE: Yeah, cool. DANIEL: This is a screenshot from the engine. Yeah, once you get to a certain setting, it's like, no, no. Don't do that. CHANCE: That's great. DANIEL: BaseColor is a huge impact on global illumination. So every time the light bounces off a surface, it is attenuated by that base color. So for the light that's directly lit-- for surfaces that are directly lit by the sun, it's only applying the base color once. So if your base color is like 0.1, you're getting 10% of the sun's energy. But for indirect lighting, it's applied every bounce. So if your base color's 0.1, you're getting 100th of the sun's energy in the GI, which you basically are not going to see that. So your base color needs to be bright if you want GI to actually show up. And here's a comparison from one of our internal projects. The base color's on the left. The top is grayer and darker base color. And the bottom is more saturated base color. And you can see on the bottom right, that's basically how much brighter and more noticeable the global illumination is than the top right with the dark base color. If you look at the base color by itself, it doesn't look that dark, right? It doesn't look like it's going to have that big of an impact on the GI, but it does. If you turn on Lumen in your scene and you don't really see anything, this is almost certainly why. And the single biggest impact on having good lighting in your scenes is keeping your base clear bright enough to actually bounce the lighting where you'd expect it to. GALEN: Yeah, I think it's worth maybe lingering here for a second, too. So some of the discoveries that we made specifically working on Valley with this. One of the things that we noticed in compensating for this problem early on was that our guys were kind of cranking the exposure to kind of compensate for just general darkness in the scene. And that was something that Daniel flagged to us immediately. And was this kind of like, hey, look, this is something that you all should be aware of as far as how to sort of make this content look the best that it possibly could-- or it can. And so I mean, we did go back and brighten up albedos across the board for the assets that are in the light world specifically. And as a result, I mean, the bounces were so much better. We also increased the value of the skylight. So I think the intensity was set to 1 before Daniel took a look at it. And then we raised it to 1.3. There were some cases, certain shots where we actually raised it to 1.5. And in some cases, even a little bit higher than not to accommodate some kind of caverness type shots and that type of thing. Yeah, I mean, it was something that we learned through the process. And on a product level for Megascans, we're definitely hoping to make it so that Lumen is as compatible with the Megascans calibration as we possibly can out of the box. So something we're working towards actively. DANIEL: And both of our projects using Megascans ended up having to boost the base color. So don't be afraid to do that. Don't be like, oh, it's scanned. It's got to be perfect. It's actually really powerful to put a multiplier in your materials that's driven from a material parameter collection so all your rocks can have the same multiplier. And then you can go in the material parameter collection and just scrub it for the whole scene and achieve your artistic vision. And just see what the impact of that is. And it can be really surprising how much that one setting can add to the realism of your scene. GALEN: Yeah, and so we're trying to address that on a product level, like I said. We definitely want to make it so that out of the box, you wouldn't have to worry about something like that. But it's something that we have to actively kind of comb through the entire library to kind of get the result that we're looking for. So yeah, our processing team is going to be communicating a lot more with engineers going forward. And hope to have a much better result out of the box soon. CHANCE: Awesome. DANIEL: So that's all about BaseColor. Next best practices is about Indirect Lighting Intensity. It's such a powerful setting. And it's so fun to set it just to see bounce and then remove bounce. But we actually have a bug with our implementation of it, which is that screen traces do not know about this setting. When they hit something, they have all of the lighting from all the lights baked down into one value. And they don't know, oh, that light was supposed to be five times brighter. Which is something we hope to improve in the future. But what this will cause right now, if you crank up Indirect Lighting Intensity to a large value, you'll get basically view-dependent GI from Lumen, which looks like artifacts. As you look around, the GI will be shifting. And then, as I was saying before, the single biggest thing to troubleshoot with Lumen is the surface cache, which can see Lumen's version of the scene with the Show -> Visualize -> LumenScene. And a lot of times, when it's black, it's because the mesh did not have a simple interior. It was like a whole room in one mesh. And if you want good Lumen quality, keep an eye on this view mode and make sure it stays in sync with your main scene. We know all of these like content requirements suck. And it's not just works. But it's better than building lightmaps, right? It's better than authoring lightmap UVs and waiting 10 minutes for the lighting build to finish. These are kind of the things that allow us to make it real-time. CHANCE: Over night, you mean for light-- DANIEL: Overnight. Yeah, depending on what you have the quality set to. These are kind of like the corners that we had to cut in order to make it run in a handful of milliseconds. GALEN: Yeah, I have lots of memories of working on projects where we'd kick off light bakes, it takes several hours, you come back and everything's broken. So I think it's a good tradeoff. CHANCE: That's right. Yeah. DANIEL: There's bugs that only happen when you crank it to the highest settings. So everything was fine on preview-- everything was fine. And then you put it on the highest settings and you go to bed. Wake up the next morning right before your deadline, and it's like-- CHANCE: Yeah, designer editor door right there that opens and casts shadow. DANIEL: Or like a door that moves or anything dynamic. [INTERPOSING VOICES] DANIEL: OK, so Lumen platforms. Since we targeted next-generation console and high-end PC, that's all it runs on. And it is a bit of a forward-looking technology that just like-- Unreal Engine 5 is not actually released yet for shipping, so we hope to scale it down better to more platforms that are out there today. But this is what we have right now in early access. We do not support Lumen on mobile. And we're not even working on that because we don't think that we can provide a version of dynamic GI on mobile that will be Epic-quality basically, that will be good enough to use. CHANCE: I think on that note, is that a mobile hardware kind of bottleneck or is it the rendering tech like Vulcan or OpenGL? Or is it a combination of all things? DANIEL: It's 100% about the performance. Because we use software ray tracing, we can run on a lot of different hardware. But if you use dynamic GI on a phone, it's going to get hot and your battery's dead. So it's just not a good fit for the platform at this point in time. I mean, there are certainly like probe-based approaches-- like irradiance field based approaches that can work. But we're not currently working on that. And similarly in VR, VR has crazy high resolution, two eyes, and usually needs to run at 60 or 90 frames per second. It's very difficult to give acceptable dynamic GI quality under these constraints. So we're not developing Lumen for VR. I know some people are disappointed about that. I'm sorry. There are limits to what can be done. CHANCE: On that note, a question came across earlier and it had to do with deferred versus forward rendering. This is a deferred rendering technology, correct? DANIEL: Yeah. Yeah, so the performance is just the first reason it can't work in VR. The second one is that Lumen does require the deferred shading pipeline. And most projects are relying on-- most VR projects rely on forward shading for lower overhead and MSAA. And Lumen does require the deferred shading pipeline. And then if you want to use hardware ray tracing, that has some extra requirements, which you can find fully spelled out in the docs. CHANCE: Something I've seen come across chat here. A lot of these things you said-- I know there's a lot of caveats in here. If these things aren't in the documentation, we can definitely get them in there as well to just to make sure that people can kind of use Lumen to the best experience they can on the hardware that they've got. And that they don't paint themselves into strange corners like we have a few times. DANIEL: Yeah, the only one that's not in the doc is VR. I forgot to put that in the doc. But everything else is in there. So let's talk about Lumen performance briefly. Lumen relies very heavily on temporal super resolution to allow Lumen to render at a much lower internal resolution than what's output finally on the screen. And temporal super resolution does an incredible job of getting close to the quality of that final resolution while paying a tiny fraction of that cost. And we feel like this is really the best way to get the best final image quality is not to actually render at 4k native, which is a crazy amount of pixels. But to actually render at a much lower internal resolution and then use temporal super resolution to get an image that looks much better than 1080. It looks close to 4k. However, so consoles has it set up by default very well-- the upsampling. But PC does not. So if you load Unreal Engine 5 on a 4k monitor on PC, it's going to render at 4k internally and it's going to cost a lot. So in editor, you can go and lower the screen percentage and get much better performance. CHANCE: And it really comes at very little visual cost. I mean, I think when we were working on Valley of the Ancient we did that. A lot of our recordings were done on 4k monitors. And we had to do a lot of A/B testing to kind of see specific areas that we needed to use TSR or let TSR do its thing and bring it up and-- there's very unnoticeable difference in so many different things there. So I encourage you to play with that setting, too, community, to find kind of what you actually need and what you can get away with. It's a really powerful feature. DANIEL: The other thing about Lumen Performance is that by default, you get Epic scalability settings. And Lumen is, by default, tweaked for a 30 frames per second on next gen consoles. If you want your game to target 60, it's useful to look at the High settings set up for Lumen. And those are the settings we're expecting to use for 60 frames-per-second games on next gen consoles. However, the whole Lumen scaling down thing is super early at this point. We didn't get that much time to work on it for early access. So we hope to have much better results there for 5.0 release and later. CHANCE: Oh, right on. DANIEL: Yeah, which segues into things we want to have for 5.0 release. I don't want to promise very many things. And none of these are promises. But we do expect to have full hardware ray tracing in pipeline in Lumen so that you don't need to have the memory cost of distance fields and ray tracing acceleration structures. And we expect to have landscape supported. And we're looking for your guys' feedback, which we've already gotten a lot of on the forums. Thanks. Thanks for that. We read all that and helps inform what we're going to work on next. CHANCE: Yeah, I'd definitely encourage everybody to go hop in there and take a look at that. Or add their thoughts, I guess I should say, to that thread. Let us know what you think. That's why we do early access, right? DANIEL: Yep. CHANCE: Before we hop over into content, I just saw this [? item ?] come up. Daniel, how long have you been kind of working on Lumen? You and the team. I mean, you said there was three or four people. I mean, how big is the team size and long have you all been at this? DANIEL: There's three of us. We've been working on it for two years. We started in March 2019. And we've made a lot of improvements in the last 12 months. Hardware ray tracing didn't even exist 12 months ago. Patrick Kelly has been working really hard on that hardware ray tracing path in Lumen. And having Lumen work in these Nanite scale levels has been an incredible amount of work. Krzysztof has absolutely incredible work to make that actually happen along with a ton of other improvements. Lumen, 12 months ago-- in that land of-- Lumen in the Land of Nanite demo, we've greatly increased greatly improved Lumen since then, which is actually what I'm going to show you. CHANCE: That's awesome. DANIEL: So this is internally called Reverb. Can't hide that because it's up here. But Lumen in the Land of Nanite demo, in-editor-- so the performance is lower in-editor than it would have been in-game. I'm going to lock FPS to 30 so that the streaming will be smoother. And so in here, Lumen is providing everything that's not this sunlit area. So if I go ahead and turn off Lumen contribution, it switches to distance field ambient occlusion, which because it's just ambient occlusion, leaks skylighting. It's not sky shadowing, it's just ambient occlusion. And it has some self-shadowing artifacts. This was our Unreal Engine 4 dynamic lighting that can actually ship on consoles. And so this is a difference between our Unreal Engine 4 best technology versus Unreal Engine 5. And you can also toggle Lumen Reflections. I'll show that a little later. So in this demo, which is entirely Nanite meshes for all of the environment geometry, Lumen is bouncing the lighting from the directional light. Well, this is actually a spotlight. But it's bouncing the lighting and it's shadowing the skylight. So you can see really detailed sky shadowing here where individual meshes block the sky lighting. CHANCE: Right. So everything in here casts a shadow also, right? Yeah. It's got a directional-- or spotlight and a bunch of-- just a ton of super high-poly meshes that all cast shadows. DANIEL: It's really fun to just start hiding stuff. Start moving stuff around. CHANCE: That's the first thing the first thing I noticed in working in Valley of the Ancient was deleting a mesh or moving a mesh and everything just kind of like readjusting really quickly. And I'm like, what? I'm just not used to seeing results like that. DANIEL: Not going to get this with lightmaps. CHANCE: Yeah, not at all. DANIEL: That never happens. CHANCE: Yeah, you'll get the lighting needs to be rebuilt. DANIEL: That's right, actually. CHANCE: Then you get your art team yelling at you because your design broke the lighting. DANIEL: Yeah, and in a production environment where there's a lot of people working on the level, the lighting is basically never up to date. The baked lighting is just never up to date. It's always unbuilt. And that's just gone. The problem is just gone. CHANCE: That's fantastic. And then in gameplay, you could destroy that ceiling, right? Say that's a destructible mesh, right? You just blow it out and then you can walk through there. You don't have to have an art team come and make sure that that visually, from a light perspective, looks realistic. It's just going to do what you expect it to do. DANIEL: Yeah. Yeah, usually, you would like hide that mesh from casting a shadow in Lightmass so that the bouncing-- so you have to pick between two evils, basically. Do I want the sun to bounce and it's still there when the rock is there when it looks wrong? Or do I want there to be no bounce? You can only have one of those scenarios look right. OK, so let's see. What else was I going to show? Let's look at the LumenScene in here. Some black stuff, which are artifacts. Oh, let me first show what the distance fields look like. So this is the first-- this is the representation that Lumen is ray tracing against. And it's not as detailed as the Nanite geometry, but this is showing it with simple lighting. So it's not supposed to look good. It's just trying to show you the detail of the geometry. You can see that everything is like in a blobby form represented. So that's where we get our software ray traces from. And then you can also look at the global distance field, which shows our merged version. This is the super fast software ray tracing, because there's only-- for any given point in space, there's only one distance field that you have to check. You don't have to go over all of them. And then we put that together with our surface cache and get the LumenScene. This is what we're using to give hit lighting when the screen trace didn't work. So the screen trace-- so it doesn't match the triangles in a lot of different ways. But the screen traces do a pretty good job of covering over those mismatches. There is one thing I want to mention is that, you see all these little bright salt and pepper spots? That's noise in the surface cache multi-bounce. And that's something that we definitely want to improve. That's not the quality that we hope to ship out. But didn't get to that in time for early access. CHANCE: That's fascinating. DANIEL: See what else I'll show you here. This is kind of a cool scene up here. Let me turn on the directional light. And as soon as that comes on, you can see all kinds of bounce lighting. This is skylight only. And then turn on the directional light and can move the directional light around, the GI updates. Can get some pretty dramatic lighting with that, especially if you make it later in the day. CHANCE: And so this is just the skylight and the directional light doing all of this work in this space? DANIEL: Yes. Where did my directional light go? It got lost. This widget is a little bit of a challenge to work with. But you can make some really dramatic sunset lighting. All bounce lighting, this is all Lumen. It would just have been this with distance field ambient occlusion on Unreal Engine 4. CHANCE: That's wild. GALEN: We didn't even have the sky atmosphere stuff for this project, which is kind of interesting. I'd be really curious to see hooking that in seeing the difference there, too. DANIEL: Oh, man. And I love these stalactite shadows here. These are so good. CHANCE: Yeah, really rich, really detailed. DANIEL: Yeah, we're doing our traces at such a low resolution, you can't even tell. And in this case, the geometry is so detailed but you still get really accurate shadowing. I mean, I didn't find this time of day before when I was testing. This is really cool. I got to look around. CHANCE: Chat's asking for a blue, like a blue sun. DANIEL: A blue sun? All right, let's go to an alien planet. That's cool. It looks like a horror movie. CHANCE: Yes. DANIEL: Where they just have a super bright blue spotlight on the character's face. CHANCE: Star Trek or something. DANIEL: Green planet. Red planet. Blue planet. CHANCE: The room right now is very-- the art director of this project is probably like, what are you doing? DANIEL: Yeah. What are you doing to my baby. It was never meant to be seen like this. VICTOR: On the topic of playing around with light values, Dylan Burrell had a question. They're asking if Lumen is assigned to play nicely with lighting and physically correct units like 5,000 lux for sunlight and using real world camera values? DANIEL: OK, so this project did not use it. It had a sun value of 100. We do have some internal projects that have a sun value of 65,000. And Lumen is working well with that. It's not the most tested path, though. It's something we want to support. And I think is working, but I'm not 100% sure. DANIEL: The big gotcha that we found with setting your sun to physically accurate values is that your emissives basically don't show up. It appears like they're not working at all. Because the exposure is adapted to the sun values, your emissives have to also be 10,000 times brighter to show up. That can be a little unintuitive, especially if you have some test levels in your project that don't have that 65,000 sun light. CHANCE: Then you have to change all your material. And since it's just for those spaces, so it doesn't look so out of place. DANIEL: Yeah, it's better to just make sure all your test levels use the exact same sun units as your main level. And then just bypass this problem. But it's super intuitive. You drop an emissive material in and you're like, the render's broken. Where's my emissive? All right, that's enough of this one. Now I want to show you guys, this is the level that Rafael Reis sent me from-- his company is ue4arch.com. He sent me a early version of his next project and gave me permission to show it, which I really appreciate. Thanks, Rafael. CHANCE: Thank you, Rafael. That's great. DANIEL: This is a level that we did not have before early access release. So there are some Lumen bugs that we would have jumped on. But I think it's a good example for me to show because it's basically your guys' experience as you use Unreal Engine 5. So this level that Rafael's made looks amazing. This is all using Lumen lighting for all indirect. No lightmaps. So if I turn off Lumen, only direct lighting's visible. Now, obviously, you would never have shipped it like this. He would have shipped it with lightmaps, which I can't show right now. CHANCE: You don't want to bake lightmaps over the next-- DANIEL: No. Yeah, actually, I was trying to make the comparison and I was just too annoyed by the hassle that I just, never mind. I don't have time for this. CHANCE: Go in the right direction then, right? VICTOR: Daniel, did you clamp the frame rate? DANIEL: Nope. CHANCE: Yeah, and to clarify again, for folks that might have just joined, we're doing that not because we can't go over 30, but we're dropping it to 30 just because the streaming rate that's going out with some of the compression looks way smoother. VICTOR: The frames interpolate much better, yeah. CHANCE: Yeah, so much easier. DANIEL: So this is Lumen cranked up. This is not the default quality. This is basically like the best Lumen can look in the current settings. So Final Gather Quality is set to 4, and reflections quality is set to 4. And this is the quality that we can do for really clean indoor scenes. And let me just go through the different parts of the house and show you. Show you what's going on. So the bounce lighting's working nicely here. I really like the skylighting coming in this door from the side. You can't see the door, but the skylighting's there and it's stable. We got the emissive television, which has kept dim and subtle so it's not blasting artifacts. The curtains here are actually emissive, too. That's how he's achieving the subsurface look without it actually being subsurface. CHANCE: Nice. DANIEL: And if we go back here, this room looks really nice. Lit entirely by this emissive curtain. CHANCE: Wild. Oh, yeah, because I guess the directionals not getting back to that space, right? DANIEL: Yeah, it's shadowed by that curtain. CHANCE: Yeah, right on. DANIEL: And the bathroom here, you can look in the mirror. And we can look at the difference between the default reflections-- just ignore that flashing arrow. Got to fix that. Default reflections and the higher quality hit lighting. CHANCE: Yeah, that's awesome. DANIEL: Which is more comparable with ray traced reflection quality. VICTOR: And you can see it there in the tile, as well, not only in the mirror. CHANCE: Yep, right next to it. DANIEL: That one is actually the transparent glass here blocks more accurately when you set it to higher quality. So go into the next room. This room looks really nice. I forgot there was a bathroom over here. There's another mirror. CHANCE: That's a nice house, in general. DANIEL: It is extremely nice. It looks so good. I couldn't believe it when I saw his screenshots. That's why I messaged him on Twitter and asked him if I could show this. Because I mean, we graphics programmers work on Lumen all day. But we mostly just see artifacts because that's our job is to make it as high quality as possible and fix every problem with it. And then it gets out there in the community and we're just cringing over here like, ah, I got to find all the bugs. And then we see results like Rafael's results and we're just amazed. CHANCE: Yeah, it's fantastic. DANIEL: It's one of the most fun things about being a graphics programmer at Epic is you get to see people break your stuff. But you also get to see people surprise you by making content that brings out the best in the technology that you make. CHANCE: Yeah, I was just chatting with Nick Penwarden and a few other folks about some of the different screenshots or videos that are coming out of the community. And it never ceases to amaze me what people are capable of doing with the tech. That's really awesome. DANIEL: So Rafael has a second lighting setup in here I shall show. This one is kind of like the middle of the day. But if I turn off directional light and the middle of the day skylight and I turn on the evening sky light and some spotlights, then voila. It is now sunset. CHANCE: Awesome. That looks so good. VICTOR: That is so good. I was just looking up from my notes here over to the screenshare and I'm like, is that a picture? CHANCE: That's great. VICTOR: It's a really nice sky sphere or a cube map that's out there, right? DANIEL: It is. It is very nice. Love it. So let's talk about some of the Lumen artifacts in here. As I scroll side to side, behind things, especially foliage, there's some flickering. Something that we'll continue to work on. It's not too bad in this level because I have the Final Gather Quality cranked up to 4, which makes it cost quite a bit more, which is acceptable for ArchVis but not elsewhere. So something that we're still working on. And then if you go and look at the LumenScene version-- oh, this is something I meant to talk about. OK, remember how I said that we never actually tested in this level before release? So you see how these walls are black here? That's the surface cache-- it's black. And that means that any light that bounces off this wall is going to only work with screen traces. And when it goes off screen, it's going to stop bouncing. Which is why as you look around, the lighting in this hallway is not super stable. The reason for that is that content limitation we have where meshes need to have simple interiors. And this level-- this mesh does not. So the entire interior of the level is all in one piece, which is normal modeling pipeline for architectural visualization. But Lumen has this extra requirement that-- and this wall piece needs to be one mesh and then this wall piece needs to be a different mesh. CHANCE: So basically, an easy way for you to tell if you're seeing that as you move around watching it kind of creep in and crawl in and update over time, flipping on that visualization to let you know if you've got something that might be malformed for how Lumen is today? DANIEL: Yeah. CHANCE: OK. DANIEL: Yeah, and if I go back to the directional light, turn it on somewhere-- so looking at the surface cache, see, it's correct on the wall because the wall was modeled as a simple mesh. But it's wrong on this piece because this piece is-- these are not in separate meshes. That's why the surface cache failed here, which is something we hope to improve. But this is where we are at early access. And as a result, the bounced lighting off of this will go away when I look off screen, which I'm not actually reproducing right now. Well, let's see. Yes. OK, you see it there. When the screen traces hit the edge of the screen, the reflection changes. That is because the surface cache failed on this mesh. So you can fix that by breaking this up into three meshes. CHANCE: Got it. DANIEL: Let's see, what else did I want to show here? Oh, yeah. Hardware ray tracing, this is our hardware ray tracing showcase. Or the one I want to show hardware ray tracing in. So if we go to Project Settings and search for 'hardware ray tracing'. So Lumen is-- hardware ray tracing is on and Lumen is set to ""Use Hardware Ray Tracing when available"". And we can toggle that and see the difference. So not a huge difference here. Let's go to another part of the house and see what hardware tracing versus software tracing is doing. Here, hardware ray tracing, the curtains, especially are more accurate with hardware ray tracing. If you go and you look at the Visualize mode, then you can really see what's happening. So in the Visualize mode, it matches the setting. So right now, we're looking at the hardware ray tracing. And if I turn that off, this is the software ray tracing. And distance fields are not great for thin stuff, so the curtains didn't become solid in software ray tracing. And in software ray tracing, we remove a lot of the little objects for performance reasons. Whereas, in hardware ray tracing, we don't. So all the little objects are there. Let's see, what else? Oh, yeah. Let's go look at those bathroom reflections. So this is our hardware ray tracing reflections off screen and then software ray tracing does not have nearly the quality of reflections. That's the area where hardware ray tracing really shines. You can't beat it at reflections. But we find for global illumination that software tracing is pretty close. CHANCE: So you can not really worry about it as much if you don't have a bunch of reflective-- fully reflective surfaces like a mirror or like some stainless steel or something that's going to have-- OK. DANIEL: Yes. And that's why software ray tracing works so well for us in the past two demos we made is that they didn't really have any mirror reflections. Whereas, hardware ray tracing, it's pretty much a requirement to make mirror reflections work well. CHANCE: So that would be basically, if you had like a mountain that was-- this was a question from, I think, the forums. If you had a mountain that was reflecting off of a pond and it's not showing up the way that you think it would, would that be because software ray tracing versus hardware ray tracing? DANIEL: Yes. Almost certainly, yes. CHANCE: And then on that same note, was it planar reflections? Any support for that? Would that be a way to work around that or would that be kind of left to the same kind of limitations? DANIEL: We don't support planar reflections. We could, but we don't plan to do it because planar reflections have a lot of overhead-- a lot of performance overhead. They actually rerasterize the scene, including rerunning all of the dynamic lighting like Lumen again-- the second time on a bigger portion, often than what's actually on the final screen. So there's just a lot of sources of inefficiency there. It's difficult to ever make them anywhere close to as performant as hardware ray tracing. So the path to getting those mirror reflections is not planar reflections, it is hardware ray tracing. CHANCE: Got it. VICTOR: Cool. DANIEL: Let's see. What else should we look at? VICTOR: There's a lot for sure. Daniel, just let me know when I could start bombarding you with some of these questions. DANIEL: OK, I think it might have been in something else I was going to show. Let me just mess around with the level a little bit. Because it's cool to watch dynamic changes. [INTERPOSING VOICES] DANIEL: If you want to hide the sky, it gets all overexposed. So then you can just compensate for that. And it's cool to see just enormous lighting differences from scene changes as Lumen shadows the skylight based on changes to the scene. And it actually looks-- so I lowered the exposure greatly. It actually looks really moody in here now. CHANCE: And so you would just kind of light this scene like you're building a house. You would kind of light it the same way you would put lights in your own home, right? DANIEL: Yes. CHANCE: So you go back to that moody scene there. Well, if you want that kind of thing, you could put a lamp in the corner or some light that can kind of fill in that space. And it should start to feel closer to what you would see in the real world as long as you're setting the actual brightness and whatnot to kind of match. DANIEL: Yeah. Rafael has a bunch of spotlights in here. Place where you'd expect you actually have lights in your house. And they bounce lighting through Lumen as well. So this is with Lumen off-- it's just direct lighting, and then Lumen on. Let me just move the directional light around a little bit because it's always fun. CHANCE: Oh, yeah. DANIEL: The skylight's too bright right now. Let me make this a little less-- a little more dramatic. VICTOR: Do you know if there are any-- oh, yeah, I can almost see it there. It looks like some of the lights are using IES profiles. Oh, no, that's the skylight. DANIEL: I don't know if any of those are set up. I could go look. But we do support IES profiles in Lumen. Let me turn off these spotlights that are flooding the area with lighting. I want it to be really dramatic-- really dramatic sunlight. I even have the skylight turned off. So it's just all sun bounce. Hell, yeah. Am I allowed to say that? CHANCE: That's awesome. DANIEL: This is like if you're on the moon, there's no atmosphere scattering so there's no skylights. So everything is really high contrast. CHANCE: Yeah. DANIEL: Like the sun is incredibly bright because it doesn't get filtered through the atmosphere and it doesn't bounce around in the sky either, so it's all bounce lighting. Yep, so I think that's all I had. CHANCE: That's awesome. DANIEL: Victor? VICTOR: Yeah. You know what? This is one of the streams where we don't need to talk as much. You can just sort of play with the parameters while everyone's watching. DANIEL: It's so much fun that, dude, I just get lost. Just the editor controls have a whole new life to them. Just being able to hide and show things, it's so much more fun when you can see the way it affects the light. VICTOR: Except knowing that if you did any change otherwise, you would just get an error saying that lighting needs to be rebuilt and-- DANIEL: Yeah, let's go back to static light, OK? Preview. And change to static. Where the lighting needs rebuilt message? It's not showing up. VICTOR: It's not. [INTERPOSING VOICES] CHANCE: Move the chair. Just move the chair or something, right? DANIEL: Oh, you know what? It's because I don't have lightmaps built in the first place. And the message only comes up if you had any built and then you unbuild it. So it doesn't actually care right now. So I can't show the message. I deleted all the-- I made everything-- as soon I came in here, I made everything movable and I deleted all the reflection captures that had to be manually set up for it. VICTOR: All right. CHANCE: That's wild. VICTOR: Yeah, thank you, Daniel. Before we get to questions, we're going to take a quick break so that we all can be with you all here just for the questions. We will be back in just a couple of minutes. Thank you, everyone. VICTOR: All right, we are back. Thank you very much, everyone, for waiting. Hope you're sticking around there with us. It is time for some of the questions. And I would like to start with a couple of the ones we received on the forums prior to the livestream. Let's see. This one here is pretty good. From Eugene, who's wondering, ""Is it possible to get more than three lighting channels with Lumen?"" DANIEL: That's funny that you ask that, because actually you get zero lighting channels with Lumen. So OK-- CHANCE: Does that mean it's global? The word global kind of-- DANIEL: Yeah, the problem is that once the light hits the surface and it bounces, we simulate all those rays together. We don't keep track of which light it came from anymore because it's too expensive for real-time. I mean, you can do it. But we chose not to do it because it would be one of those costs that most users don't want to pay. So we don't actually have any lighting channel support in Lumen and it's not something we can really add either. CHANCE: Right on. Let's see. Oh, I have one here from ReticentOwl. Lumen usage with light-based particle effects. So if there's a particle effect that's got emissive on it, how is that going to influence the scene? DANIEL: It's not going to work well. So the combination of small brightnesses doesn't work well with Lumen emissive. And then also, Lumen uses a pretty strong temporal filter to actually make it possible to be done in real time. And any kind of fast moving lights are going to cause a problem with it. And in fact, if you have a muzzle flash, you probably want to set Indirect Lighting Intensity to zero and remove it from Lumen's consideration so that the lag is not visible. CHANCE: Got it. OK. VICTOR: Couple of questions that are in regards to the possibility of being able to exclude certain meshes from Lumen. And this is in regards to, say, one of the suggestions-- or examples was a sci-fi corridor with a lot of emissive lights going on. Is it possible in any means now or potentially or in the future being able to exclude emissive primitives from Lumen? DANIEL: Yeah. We need to do something about that because if you want to use Lumen and you're trying to make a sci-fi map with lots of small bright emissives, then there's going to be a lot of noise. And you don't have any control over it right now. We do not have that ability right now because of screen space traces. They're going to hit what's on your screen and they're going to pick up that color. And so while we do have like this separate Lumen representation that we could hide emissive from, screen tracers don't respect that. So we need to do some work there-- just work that. CHANCE: Right on. I've got one from Laurish asking about potentially making Lumen calculate and converge faster for things like the Movie Render Queue since it's not-- it's happening kind of frame by frame. Is that something that is possible at this point? DANIEL: Yeah, it's something we know we need to do, especially on camera cuts. Since Lumen is all targeted at real time and it needs multiple frames to converge, when there's a camera cut, we have a noisy image. And the image changes after a couple of frames, too, which shows up as a pop. So what we need to do is do some warming up, but also allow more budget for those early frames to smooth it out and make Lumen work better with cinematics. And if you do a full offline render-- can't remember the name of that in the Movie Render Queue-- I guess that is just the Movie Render Queue, right? Then, yes, we should be using some much higher quality setting since the performance is not an issue anymore. But we don't have that implemented right now. CHANCE: Got it. VICTOR: On the same topic, GulloPBR was wondering, ""Are there any plans to transform Lumen GI into lightmaps for ultra fast baking?"" DANIEL: No, there aren't. We actually have a really good lightmap previewer through GPU Lightmass, which-- so Lumen is operating on the screen. It's not going through the lightmap textures, which makes it not a great preview for lightmaps because it doesn't accurately represent lightmap artifacts like UV seams-- lightmap UV seams. Or your lightmap resolution is too low or you're just missing UVs on this mesh or they're overlapping or something. You want to see those artifacts as part of the preview so you know what to fix before you click the long build. And Lumen doesn't have the ability to reproduce those artifacts. But the GPU Lightmass Previewer does. CHANCE: All right, got one from PatientZero, I think, if I'm understanding this correctly. So the surface cache-- if you have a mesh that's got a bunch of different planes and whatnot inside of it, and we saw the black on the screen earlier, that's basically because of the mesh itself, not necessarily the layout of it? For instance, the question is about building floors and walls inside of a Blueprint and then putting the Blueprint in the scene, it would still work great with that, right? DANIEL: Yeah. Yeah, and both of our recent demos did a lot of that where the meshes were assembled in a Blueprint somewhere and then placed in the level. That works fine. CHANCE: Great. DANIEL: Hey, we win one. CHANCE: This is really amazing tech. VICTOR: Yeah. VICTOR: There are some other questions in regards-- I think you touched on this a little bit in your presentation. But might be worth diving into a little bit further. What are some things to look at to optimize? You mentioned sort of overlapping meshes. What else would one like to start diving into if they're looking to optimize their scene for Lumen? DANIEL: Yeah. So let's see. The main performance stuff is the resolution. Making sure temporal super resolution is set up, and that Lumen's rendering at a much lower internal resolution. And the output, not 4k. Content wise-- like the way you build your content, if you get up to two million instances, there's going to be quite a lot of cost with that of Lumen trying to keep the surface cache up to date, especially if you fly fast. The instance overlap is probably the single biggest thing for Lumen, especially if you want to use anything higher quality than just the global traces. Like this scene that Rafael made performs very well because architecture stuff, you tend not to overlap too much. But if you just copy paste scanned rocks all overlapping, then the tracing cost would be quite significant. Let's see what else? Reflective materials cost more. That's kind of unavoidable. Not reflective, but smooth materials, roughness less than 0.4. It's kind of unavoidable because we have to trace more rays to give good quality reflections. But it is true that scenes with more diffuse materials cost less. VICTOR: That's not only true for Lumen. That's true for other features as well, right? DANIEL: Well, our reflection meshes that worked with bake lighting, like the reflection captures, are very fast. So you don't really notice that reflective material costs more. It does cost a little bit more. But it's not enough that you really notice. Whereas, with Lumen, it actually costs significantly more. Because just like ray traced reflections, Lumen needs to trace a ray per pixel to give good quality reflections. VICTOR: So resolution is once again, probably one of the most expensive. Yeah. CHANCE: Generally, Yeah. VICTOR: Yeah, it's just in general. But it's good to know that it's specific here for Lumen as well. DANIEL: Yeah. CHANCE: I got one-- I'm sorry. Go ahead. DANIEL: With dynamic lighting, we have to do so much more processing for every pixel. Dynamic lighting pixels are not the same as static lighting pixels. It's significantly more expensive. So temporal super resolution is really the key to making all of this real time. VICTOR: Go ahead, Chance. CHANCE: Got one from our old friend Tom Looman. Hey, Tom. The question is about, let's see, more improvements during early access or is the next release immediately 5.0? I don't think we've made any noise yet about-- we haven't said anything about any interim releases in between the two. But if you're looking for some of the latest and greatest, and Tom, I know you have the guts to go do this. Main is updated pretty much regularly. So if you watch out for specific areas there, you might be able to cherry pick some things down and try them out. DANIEL: We did fix one bug in Lumen in early access, which was if you had the software tracing set to global tracing, it would break hardware ray tracing. That's fixed. Well, sorry, will be fixed in the next update. But no new features in early access, no development. It's stable. Just fixing bugs and, yeah, like Chance said, check UE5 Main if you want-- if you want to drink right from the fire hose. CHANCE: If you want to go through the pain of Main. I said this on Twitter a couple of times. Main is generally, not super stable for a number of reasons because you're getting basically the latest as we can check it into the engine ourselves. Having said that, a number of folks do feel comfortable in that space and want to kind of see what's coming or grab a fix or two that they might see that affects their lives. So if you-- yeah, if you've got the guts for it and want to go fight the dragons, go nuts. VICTOR: There was another question here from unknown. I don't have the name, unfortunately. But also, ""Is it possible to disable emissive from the computation entirely?"" DANIEL: It is not because of the screen traces. Because we're always going to pick up what's on the screen. But that is a very commonly requested feature to be able to control the intensity of the emissives. So we'll see what we could do. VICTOR: Thanks. Another question here from VirtualH-E who's wondering, ""Can the distance field mipmaps' distance thresholds be tweaked?"" DANIEL: Yes, they can. OK, so right now, it's all in a bunch of CVars. I don't remember the setting right now. But basically, with this one, you can actually just force-- so you can force it down to-- just here's the lowest mip of the distance fields. And you can see what the different mips look like. And there's the highest resolution mip. VICTOR: Let's see. We still have the PIP up there. Can you repeat what the console command was, Daniel? DANIEL: It was r.DistanceFields.-- I crashed it. Good timing. CHANCE: That never happens. DANIEL: That was one of my debug CVars and I didn't check the bounds on that one. Nobody would ever input a wrong value into that, right? VICTOR: Sure, no. Why would they? CHANCE: Yeah, and early access doesn't crash at all. DANIEL: Except for me in a livestream. CHANCE: I think back in the day when we were first doing a DFAO, way back in the Fortnite days, I think we had crashes back then, too, if you recall. DANIEL: I crashed it? CHANCE: Oh, I think everybody did. I think we all have. DANIEL: Well that's my first crash for today. VICTOR: On the topic of mesh distance fields, there's another question from 1Millinis who's wondering, ""The limitations that mesh distance fields had in UE4 with non-uniform scaling, are they still relevant?"" DANIEL: For the most part, no. We don't have significant problems with non-uniform scaling anymore. We've greatly improved that. If you go too far with it, like let's say, more than like 4x scaling in one dimension versus the other one, then you'll start to see some artifacts. CHANCE: And I also see that in your art to begin with, right? DANIEL: What's that? CHANCE: With scaling. You would see it in other places, too, other than just lighting, right? DANIEL: Not always. A lot of times, people make these really simple template levels with just a box. And they just scale it like crazy. And it looks fine because the triangles didn't have any detailed texture map to them. CHANCE: Yeah, I was speaking more after grayboxing but yeah. DANIEL: Yeah, real world assets, if it's not procedurally textured with world space, if it has actual texture map to it, then it's going to look pretty terrible once you non-uniformly scale it by a crazy amount. CHANCE: Got one here from RendermanPro, what about sub-surface or translucent materials? I know you showed the glass in there earlier. But what about subsurface? DANIEL: Yeah, we want to add support for the subsurface shading model. We haven't got around to it yet. And we also want to add light scattering through using Lumen's traces, but we haven't implemented that. VICTOR: Let's see here. We had a couple of questions in regards to combining Lumen and baked lighting. Lumen for some areas, baked for others. DANIEL: So we looked into the feasibility of say baking lighting indoors where you have a lot of lights that need top quality, and then dynamic with Lumen outdoors. And we came to the conclusion that it wasn't actually a useful combination. Because most of Lumen's cost is screen dependent-- screen resolution dependent. And so you're paying that even when you're in a baked lighting area just to have a seamless transition between the two. So we decided not to support that particular combination. But we do really want to have Lumen Reflections supported with Lightmap GI. The same way that ray trace reflections work today. That's something we plan to add in the future. VICTOR: Cool. That's a good note. Thank you, Daniel. On the note of Lumen reflections, we have a question from CarloALT who's wondering, ""With Lumen Reflections, the reflective surfaces reflect the distance field scene. In the future, will it be possible to reflect the lit scene?"" DANIEL: Yeah, that's what this reflection quality is doing. The reflection quality of 4. So this is my bad for not documenting this. Kind of came in hot in early access where we were trying to figure out what the scalability settings should be. But the Lumen Reflections > Quality 4 is a magic value, which gets you the hit lighting, actually evaluating the light at the hit point in the reflections instead of using surface cache. VICTOR: So not in the future, it's right now with the number 4? DANIEL: Yeah, it's a hidden feature. VICTOR: And that's why we have you here, Daniel. DANIEL: That's why you guys need to watch the livestream. To find the secret. VICTOR: I say that every week. DANIEL: We saved it for the livestream. It was intentional. No, it wasn't. VICTOR: Another question. And also something that we talked about a little bit before the livestream. This one comes from Caradeco who was wondering, ""Any tips of how to use Lumen but also make our game scalable for low-end PCs?"" DANIEL: Lumen doesn't scale down to low-end PC right now. We have some prototypes of basically generating irradiance fields with Lumen. And that can scale down pretty well. But they haven't gone further than prototypes yet. And I don't actually know whether we'll be able to scale down further than we are scaling down today. Can't make any promises. VICTOR: Galen, I saw some questions earlier in regards to when we were talking about changing the base color to better match Lumen. Is that something that folks who are already using Megascans might have to redownload some packs to change that? Have you talked about how that workflow might look like? GALEN: We don't have an official announcement on that today from a product perspective. It's a complicated issue, obviously. Yeah, we'll have a proper announcement about that soon. VICTOR: For now, you can go ahead and open up the material, pop in a little scaler in there. And go ahead and tweak it yourself. Most of them do have a master material, right? So it isn't necessarily a one on one work that you've got to do. GALEN: Yeah. I mean, you can go and literally just crank albedo tint to-- if it's at 1, 1.5-- [INAUDIBLE] --or something like that. So you could do whatever. But yeah, I mean, there's definitely more mathematical ways that we want to kind of sync up with Lumen. DANIEL: You do have to be careful when you're multiplying albedo, which is base color, you don't want to get anywhere close to 1. So if the input is 0.5, you cannot scale it by 2. 2 works on the rock textures because they're starting out like 0.05. But real world materials like the brightest white wall is like 0.997. It's not 1. 1 Means 100% of energy gets reflected, physically impossible material. And the multi-balance will just explode and wash out your whole screen. So just like while you're doing this, look at albedo-- the diffuse and make sure that it's not 1. GALEN: Also we have-- so yeah, I agree with all that. So then there's also just an albedo tint that we sort of pass in through the master material that you can change the color of a rock or something like that. And that by default, just passes in 1s across R, G and B. And so if you were to change those values, you can also get you there, too. But yeah, I mean, we'll definitely look to get a mathematical solution in front of the community soon. CHANCE: Awesome. Daniel, earlier you had mentioned altering Lumen through post-processing-- Lumen settings through post-processing. Do we have access there to all the same checkboxes that we have in Project Settings? Or is it just more of the drop down to use it or to not use it? DANIEL: No, it's not the same. So the Project Settings, we put stuff where we expect you'll want it in every level. Kind of like a project wide decision. And you don't want those test levels to get out of sync. For example, the choice to use Lumen or the choice to use hardware ray tracing. Because it greatly affects the look, you wouldn't want like someone in test level somewhere else that's verifying your assets to not have it enabled. So the settings that are on the post-process volume are things we expect you want to change just in this one part of the map or this one hallway. CHANCE: Can you query that information from Lumen in a post-process material? DANIEL: No. Post-process materials don't have access to any of the Lumen internals. CHANCE: Trying to find ways to really dig in. VICTOR: Another question here from MallChad who's wondering-- a bit open-ended. DANIEL: MallChad? VICTOR: MallChad, yeah. I guess, Chad of the mall. A bit open-ended. ""What considerations was taken into flexibility wise for people who want to extend the behavior of something as big as Lumen?"" DANIEL: I'm sorry to say, we haven't got to any extensibility yet. We've been just absolutely hammered with just getting the quality that we want out of it and performance and making it work on all the platforms and shipping early access. We do really want users to be able to extend the geometry types, especially if someone's making a voxel plug-in or a procedural mesh plug-in to be able to have-- to extend Lumen to be able to ray trace that. But we haven't worked on that at all yet. VICTOR: Another good question from Raiju who's wondering, ""Any thoughts as to the application of Lumen with regards to a stylized art style as opposed to realistic?"" DANIEL: I don't have any thoughts on that. We're pretty much always just trying to match the path tracer when we're working on this stuff. Basically, the path tracer is the reference. And if Lumen doesn't match it, that's a bug. I guess it mostly comes down to our art knobs-- artistic controls, which we're sorely lacking because of the difficulty with screen traces making it hard for us to have the GI not match the main screen. We're looking at what we can add in the future. And we definitely know we need to add more art knobs. But right now, it's very limited. GALEN: Well, one of the things that this brings up in my mind, too, is we've shown some photorealistic examples, obviously, in the last several projects that we've been working on with regards to Lumen and Nanite. But it's actually really cool to see now, especially I've had several friends that I've actually reached out to me to do more of like concept painting and environment design where they rough environments in, in 3D. And they're able to just now use Lumen and not have to mess with a lot of the settings they otherwise didn't really understand in UE4 with regard to lighting in general. So it's really cool to see that. That's kind of like a tangent off that question. But it is pretty awesome to sort of see now that you have a lot of-- different types of artists, a lot of different types of artists that are actually kind of engaging with the tool and making just entirely different kinds of art. And Lumen is one of those tools that just empowers artists in such a cool way. So I don't know. I just figured I'd throw that in there on the side. VICTOR: And I'll say it's great for those of us who aren't too good at lighting. CHANCE: That's right. Yeah, that's what I was going to say. Yeah, for me, it's in many ways, a workflow thing. The final output, it's great and all. But kind of knowing what-- Daniel said earlier, knowing what your scene's going to kind of look like without having to wait for somebody to come click the button and then get your lighting scenario loaded up or that sublevel that has all that lighting in the scene. You just make the changes and they're there. And that happens regardless of what style your game is. DANIEL: So I worked on CPU Lightmass back in the day. And it was always pain and suffering. I made it the best that I could. But at the end of the day, it's baked lighting and it's going to take time to build. And it has these pain points. Very poor artist iteration. It is definitely like personally satisfying for me to now be part of making this tool, which greatly just addresses all of that poor artist workflow. CHANCE: Yeah. Well, and the same thing for a designer. If you're gray boxing and trying to figure out what the flow of something's going to look like and then somebody comes in and lights it and actually arts it up. And you need to make a change to something, you can know what that change is going to affect. And what other problems that might pose for your lighters or your artists without having to just do it. And then wait for them to come back with a solution and say, hey, this will or won't work. Or a response that this won't work. I mean, I think it's just tightening all those iterative loops that is something that really, really, really helps iteration time. VICTOR: I've seen a couple of questions in regards to-- oh, where did it go? Seeing a lot of questions. One particular that I was looking for. Apologize for that. I'm scrolling too fast. Jesus was wondering, do you guys have any example which plays with water? I would love to see how this works with water. DANIEL: Lumen reflections don't work on single-layer water yet, which is the shading model you have to pick for water. But we're going to make that work. I'm not sure if that will be in 5.0 yet. But it's been on our minds. CHANCE: There's a bunch of these questions we've answered, so I'm trying to pop through it. Is there scenarios that you wouldn't use Lumen that you can think of? Other than hardware? DANIEL: So there's the whole question of, what do I use for my project? Is it dynamic lighting or static lighting? So if you have gameplay reasons that you need to use dynamic lighting, that makes the choice easy for you. And there's the platform constraints of Lumen. The alternative is our baked lighting pipeline, which is very mature and high quality, especially on lower end platforms. Platforms where battery life and heat are a big concern, like handheld platforms, lightmaps are a much better choice. VR, lightmaps are a much better choice. And ArchVis, where the quality has to be perfect, lightmaps are extremely good at that. The diffuse quality at least, together with ray traced reflections. And probably some of them that I'm not thinking of. CHANCE: I think that's good though. It's a lot of good options-- considerations. DANIEL: The important thing is that in Unreal Engine, we have both of these paths covered with extremely high quality tools. It's basically up to you to decide static lighting-- static, global illumination through lightmaps or dynamic through Lumen. GALEN: Boris. CHANCE: Or what? GALEN: Boris for now. CHANCE: Oh, yeah. VICTOR: I found a question I was looking for. I had seen a couple of them. This one in particular comes from Mohammed Edri who's wondering, ""How can I render Lumen in a big forest scene? By default, it disappears at a certain distance. Is there a parameter to tweak to solve this?"" DANIEL: We have two problems with forests right now. First one is the range of the LumenScene. There's no well-exposed parameter. I mean, actually, everything in Lumen is just like a CVar. So it's there. But you're going a little bit off the beaten path if you start tweaking those. And the bottom line is that the techniques just don't scale up to a giant open world. LumenScene, the surface cache specifically, is limited to about 200 meters. So screen traces work past that. And in Lumen in the Land of Nanite, we had a different technique called the distance scene, which is Lumen working up to about a kilometer. But that is like an experimental thing that's not on by default. So if you have a big forest, for now, there's going to be noise around all the leaves, which is something we're working on. And Lumen's only going to be active to 200 meters. And there's no control to extend that distance other than tweaking CVars, which I don't want to say which ones they are right now because it's been a while since I tested it. GALEN: And that 200 meters isn't as scary as it sounds, I think. I mean, the distancing and then switching SSGI and stuff like-- that was something that when Daniel mentioned it to us, originally, we're building that demo, it's like, oh, well, we're building this giant world. You're going to see where that bubble basically ends. We never really noticed that. I mean, Daniel and maybe some of the other rendering engineers-- DANIEL: Yeah, I noticed it. GALEN: --could pick out exactly where they saw it in certain screenshots. But for most people, I don't think that that was a huge factor, especially on the last demo. DANIEL: Yeah, so to correct when I said it doesn't work with huge open worlds, what I meant was, specifically the surface cache. Because it's only simulating what's around you. It works with large open different worlds, but it doesn't cover the whole view range. Whereas, basically Lumen falls back to screen traces after the 200 meters. So the final result works with large open worlds. It's just not this one ingredient of the recipe. CHANCE: So it's coming up quite a bit again. I know you covered it in your slides, but I figure it might be worth for the folks that might not have seen that or just hanging on for a Q&A to cover again. What are the current supported platforms? What are the planned supported platforms? And kind of where are we maybe not going with it at this point? Again, could you just reiterate that one more time? DANIEL: Yep, sure. Let me go back and find that slide. OK, so Lumen is next generation consoles and high-end PC only. And we do not currently have the ability to scale down below that. We might develop that in the future. But it's kind of like an area of open research because it's very difficult to scale dynamic global illumination down while keeping a quality that is actually acceptable. You pretty quickly get to the point where you wouldn't have wanted to ship with that quality level. For games that need to use dynamic lighting and ship across platforms, like Fornite, the solution is to use our other dynamic lighting technologies that do scale down like distance field ambient occlusion and screen space GI for that mid-spec PC and last generation console. So Lumen on next gen console, distance field ambient occlusion and SSGI on Playstation 4 and last gen consoles. And then unshadowed skylight on mobile. And that way, the same set of content and the same game can ship across a wide range of hardware all with dynamic lighting. CHANCE: Got it. Cool. VICTOR: Chance, is there anything else you found here? I feel we've-- CHANCE: Yeah, I think we've wrung the rag. All the water that's in it in many ways. VICTOR: Wrung the rag? CHANCE: Wringing a rag out, like squeeze some-- VICTOR: Oh, ringing. All right. OK. CHANCE: Yeah. VICTOR: Yeah. Sorry. I don't-- not a native speaker here. Some things are still new to me. CHANCE: It's OK. I forgive you. DANIEL: Can you switch back to my screen real quick? While we were just doing Q&A, I didn't even notice these chairs before and under this table. But it just looks really nice. I was just sitting here looking at it. I never even looked under the table before. But like this chair looks real right here. CHANCE: Stream ended. Are we good? Are we still good? VICTOR: Let's see. I think we're still seeing Daniel. The TeamView instance died. DANIEL: OK, it's fine. It's not important. VICTOR: We're still live. Getting some artifacts. They can still hear us, so don't say all the things. DANIEL: I'll try not to. DANIEL: Victor, what? I can't believe you said that. VICTOR: To not say all the things? DANIEL: I can't believe you said that. CHANCE: It's OK. Oh, there we go. VICTOR: Let's see. It's going to be-- oh, I haven't had this happen to me before. CHANCE: Yeah, those are [AUDIO OUT] chairs. 9 or 10, what's it? DANIEL: We save our technical difficulties for the end of the stream. VICTOR: If that's how it goes every stream, that will be great. Trying to get back up. It's kind of interesting that I'm not able to access the PC but we're still live. CHANCE: Yeah. And chat's talking about the chairs. So I think they can see. VICTOR: No, and I can see Daniel showing the chairs. It's just like 8 seconds delayed. GALEN: Look at that chair. It's hot. CHANCE: Look at that. It's a good chair. GALEN: It's a great chair. DANIEL: You model a chair-- a really high quality chair. You put it in a level and it looks real. CHANCE: Yeah, that's fantastic. I can't wait to see where we go from there. I mean, the techs been out for just a couple of weeks now, right? So I'm sure it's going to be exploding with all kinds of beautiful chairs. GALEN: Well, I mean, that's one of the cool things, too. I'm sure you guys have been doing the same. But just jumping on ArtStation, jumping on YouTube and start seeing people update scenes that they maybe made in UE4. Flipping on all the bells and whistles with UE5. DANIEL: Yeah, how easy it is to try out the UE5 features. Lumen is just one project setting and you can really quickly select all your static meshes in the Content Browser and convert to Nanite. And you can be trying it out. And Virtual Shadow Maps are enabled by default. And there's just one project setting to turn it on if it's an existing project. It's like minimal effort maximum result. CHANCE: And on that same note, too, there's so much free content in the Marketplace. If you want to try these things out, you can grab them into a UE4 a project and then bump it to UE5 and take a look. This is so great. Yeah, it's so fascinating to see what people are able to do here. As not an artist, this right here is just magic to me. Or a rendering programmer. I'm just a guy in between. So this is really special. Well, I am good on questions I think at this point. We've got other things. If there's anything that didn't get answered that we missed on the doc here, I think we can take this back to the forum. Right, Victor? VICTOR: Yeah, the forum announcement post, which is where we announce the streams every week. You can go to forums.unrealengine.com. The Events section, that's where we post all of them. And the forum announcement posts for the episodes of Inside Unreal is where we continue the conversation after we have gone offline. Daniel, is there anything else you wanted to leave the audience with before I do my outro spiel and we say goodbye? DANIEL: Nope. That's it. VICTOR: All right, well, I'm going to thank you in just a bit. But before we get to that point, I would like to let everyone know, thank you so much for joining us today. Again, it is my time to talk. Because I don't say enough every week. If you tuned in and you just found a channel here today, thank you for joining us. If you're interested in getting started with Unreal Engine, you can just go to unrealengine.com, download the Epic Games launcher. From there, you can go ahead and download both UE4, as well as UE5. There are no reasons why you shouldn't be able to start learning how to use Unreal Engine. If you think, oh, but UE5, it's in early access-- not ready. Just go to UE4. If you go to learn.unrealengine.com, you can find a whole wide array of courses available to learn everything from lighting to Blueprints to programming to VR. We got it all. And if I'm able to scroll up to my notes, I will be able to continue. Make sure you use the communities.unrealengine.com. That is where folks host regularly physical meet ups. But during the pandemic, this has not been happening. However, folks are still getting together virtually, showing off their projects, talking about what they're working on, as well as helping each other out. And so if you're interested in finding a community that might be a little bit closer to you than anywhere else in the world, go ahead and go to communities.unrealengine.com. You can also find others on-- I already mentioned the forums. We also have an unofficial Discord channel unrealslackers.org, as well as a Facebook group and Reddit. All good spots. Make sure you let us know what you're working on. Clearly, it's super exciting for all of us to see all the UE5 experiments that you all are doing. Make sure you mention us on Twitter. You can add us on the forums. Send us PMs. Or if you have something that you'd potentially like to let us know but you haven't let the world know about yet, go ahead and email [e-mail]. We're still looking for more countdown videos. We've been going through some of our short film Australia videos lately. But they're going to run out. And so we're still looking for development videos. They're usually about 30 minutes of development. Fast forward them to five minutes and send it to us separately with your logo. And we will go ahead and add the countdown to them. If you stream on Twitch, make sure you use the Unreal Engine tag, as well as the game development tag, if you're doing game development. If not, just use Unreal Engine. Maybe there's something else there specifically for what you're doing. We used the creator one. That's a good one. And those filters are the best way to find other creators that are working live on stuff using Unreal Engine. CHANCE: Lots of them up there, yeah. VICTOR: There are, yeah. And today is a great time to do it. Because we try to raid someone after we've gone offline. That is if you're watching this on Twitch. If you haven't already, make sure you follow us on social media. That is where all news related to Unreal Engine and Epic Games go out. Or where you see it first. And make sure you hit the notification bell on YouTube if you want to see when we upload new videos. Most of them are not live. And so there's a lot of content there that you might be missing if you're only looking for our live content. Next week, Chance, what do we have next week? CHANCE: I think next week, we're going to cover the open world features with the tools team. So that ought to be great. I'll be chatting with Jean-Francois on, well, partition, data layers, one file per actor support. All the great things that allowed us to really build this expansive space in a short amount of time. VICTOR: Yeah, we were hyping about this earlier this week. Just sort of talking about how cool some of these features are. And while they're not as flashy and pretty and we can't just have Daniel sit here and turn a slider back and forth for everyone that's enjoying the content, they're extremely useful for some workflows. So suggest you tune in next week. CHANCE: I think this is the one that on one of the YouTube comments was like, oh, these folks are talking and Chance is just nerding out about all the features. I think this is the one I probably nerded out the most on that livestream about-- because I worked on it in the project myself and so I immediately saw some of the benefits there. I'm really excited to talk about it. VICTOR: Cool. And with that said, my outro spiel is done. Daniel, Galen, Chance, thank you so much for coming on this week. Hope you had a good time. DANIEL: Yep, thank you. [INTERPOSING VOICES] VICTOR: Oh, you're so welcome. I'll make sure you get the updated slide template for next time. Or maybe we'll tweak it a little bit so that you'll like it better. DANIEL: Do I-- do I get anything for being on the livestream? VICTOR: Send you a t-shirt? CHANCE: You like ice cream? DANIEL: You guys should have a live stream t-shirt. VICTOR: We can take a walk in the rain maybe. I don't know. Socially distanced? DANIEL: I like the t-shirt idea. VICTOR: OK. What size are you? DANIEL: Large. VICTOR: All right. All right. I can make that work. CHANCE: It's happened. VICTOR: Yeah. Now, you all saw that I promised Daniel that live, so yeah he'll haunt me for that if it doesn't happen. Cool. Thank you all. DANIEL: I have witnesses. CHANCE: That's great. VICTOR: You do. About 477 plus 927 as of right now. Anyway, thank you all for hanging out. And thank you all for staying out there. Stay safe and we will see you all next week at the same time. Take care, everyone. [MUSIC PLAYING]

Channel: Unreal Engine

Views: 178,172

Keywords: Unreal Engine, Epic Games, UE4, Unreal, Game Engine, Game Dev, Game Development, ue5, valley of the ancient, early access, global illumination, ray tracing, lumen

Id: QdV_e-U7_pQ

Length: 154min 51sec (9291 seconds)

Published: Thu Jun 10 2021