Unreal Engine 5 | How Good is Lumen?

hello everyone and welcome back so before we begin today's video i'd just like to uh thank everyone so much for watching my last video and if you just subscribed welcome i don't know entirely what happened but uh well it got a couple of views um it basically now makes up for about 90 of my entire channel watch time and it's allowed me to meet and chat to so many people across the industry and beyond so thank you it means a lot that you chose to watch now in my last video i had my mind blown by unreal engine 5. it took us one step closer i feel to that dream of raw creation where you're able to just kind of work with an incredibly powerful tool and make things so for me the absolute long-term dream is to basically just be able to work at the speed of thought which means you have an idea you get the idea out there pretty much as quickly as possible and we are a very very long time away from that but i think unreal engine takes you a step closer and it promises to remove a couple of the really important slow technological barriers that slow down fast iteration and testing of ideas now initially i was going to start spending some time looking at virtual reality with unreal engine 5 and spoiler there are some really good things and some things that probably still need to come out in the future but while i was doing this i was really thinking about lumen now as a vr developer i don't get to work with real-time gi very much and i had such a good time with lumen last time that i really wanted to put it to the test and see if it was in fact this next step in basically just giving artists really quick fast accurate feedback for the way their scenes look so i devised a way to test out lumen and push it to its absolute limits as well as scratch a itch i've been having a little while to just make something fast and dynamic and high poly that isn't necessarily optimized and as considered as you have to be when you make virtual reality things but i think perhaps really i just wanted to make something fun so we're going to make a virtual club scene the reason i'm going for that is because typically in a club or bar environment you have a lot of different moving dynamic lights and moving objects by which i mean people now i want to see how lumen handles all of this dynamic lighting so so far in the demos i've only really seen lumen running with a few different light sources and i want to see how it handles more quite a lot more as well as a lot of injector global illumination from surfaces so i adapted the observation platform from my space city that we saw in the last video and i basically just modeled a central decks area and a few different elements to go around the scene things like a giant uh what's going to be led tv to push as much dynamic moving and also changing light as possible into the scene then i nicked and repurposed a security camera asset from my vr game and used it as a movable projector that was going to start basically just waving around in whatever direction it fancied throwing again more light into the scene kind of like this idea but with a little bit less uh jack which i then just kind of iterated on and it was it was fairly straightforward to get it working okay now one of the larger challenges of this scene was the people and to be honest i think they kind of look well hilariously bad for the most part i used a number of uh miximo pre-made assets and animations just because i didn't really want to let this what was effectively just meant to be short simple quick test to satisfy some curiosity spiral into a huge project so i had to draw the line somewhere and i was quite happy having fairly janky terrible people if they still interacted with the light okay and it was kind of funny so it's kind of good to watch so yeah i would say the mixomo assets are they're absolutely fine for certain kinds of games but for the new kind of ultra high fidelity things that you might want to start going for uh they probably are quite oversimplified and that's okay there's a reason that animation is its own ultra high detail very complex industry and it can take a lifetime to master animating so i wasn't going to try to take all of that on in a day now initially i wanted to populate the entire scene with metahumans but in the end i didn't even try metahumans are really really good they're ultra high fidelity but they really are still quite expensive in terms of performance so what they ended up doing was using a few metahumans for key shots and then much much simpler characters for background and if i was going to put more than you know a few hours into this project i probably would have looked into using some kind of niagara style particle system to drive all of the animations for the background characters but that is a rabbit hole for a future video now a couple of videos ago i actually created the metahuman to try and represent myself inside unreal engine and i did use it for this video but i thought it'd be a little bit self-centered to have it front and center so what i ended up doing was creating a ridiculous and fairly terrible helmet to cover it up i thought it'd be a good chance as well to test some of the reflections that lumen was capable of but really this was a pretty ridiculous idea and it's probably one that if i'd put more than two minutes of thought into wouldn't have made it into the final experiment but then again this is just an experiment so this is the time to do ridiculous things that and not feel that afraid of making slightly questionable decisions fundamentally projects like this are about learning and experimenting rather than being too scared to try out an idea that you might later cut okay so we've got our concept we have got our zed list 3d actors let's get them all into the scene and get it all built out and see what we can come up with and then we'll render out a few shots and we will see how they perform [Music] [Applause] [Music] [Applause] [Music] [Applause] [Music] [Applause] [Music] okay all right so i couldn't resist just taking it a little bit further um one of the main things i wanted to understand with this experiment was scale i wanted to know if you could zoom right in and get the nuance of a detailed lit rich space and then zoom right out to the city scale now i don't think even with the incredible power of nanite we are able to have a environment where every single room or every single piece of an interior has been realized yet but i think we can get pretty close to some really really amazing stuff at least in the relationship between small internal levels and great big cities and that's not something new that's been done in triple a video games for years but i do wonder if this tech is going to make it a little faster to do both faster for indies and also faster for aaa teams so that effectively as a player we can just see more of it but i will also have to acknowledge that for now this project did kind of achieve what i was hoping for it to achieve and that it was able to really push nanite to the point where it started to glitch out um i think i either maxed out memory or there was just something quite off with the scale but i encountered this uh directional light casting bug so when i triggered the retractable roof to open it was starting to look really really exciting and all of this incredible light was starting to diffuse through the fog but then all of a sudden the directional light once and started entering the space started flickering on and off now i did find some allusion to this inside the unreal 5 documentation and they had a few potential quick fixes that might help i tried a few of them and none of them were really doing it so i decided just for now to cut my losses and get on with the video and this is either something that i could fix as a creator or something that is you know just a result of the fact that lumen is still undergoing development however if we put aside this potential bug or potentially user error for just one moment i think that lumen is a really really powerful tool um it really was good at just kind of injecting bounce light into a scene and i think we definitely did highlight a few ways that it does it and a few limitations it might have for example when you add a light it takes a few seconds for the gi to catch up and this seems to be a way of spreading over that kind of bounce light calculation over a number of frames maybe it's a frame rate thing i'm not quite sure about the tech behind it but what it basically means is it's at least feasible but if you have a lot of really fast moving lights interacting with your scene you may see some slow artifacting but apart from that small bug and you know as long as you know how to use lumen i'd say basically is the real deal it's real-time global illumination and we now have real-time ray tracing and we have lumen that's two really really good options for building up high fidelity worlds and now as the tech develops over the next few years i think my hope and also expectation is that real-time gi will only get better and more integrated into more games it's definitely something that's existed as an ultra high-end feature in fact a couple of other quite accessible engines like cryengine and unity have had various types of real-time gi as options already but now it's really great to see that unreal engine not only has one but has a very good one now let's see what people do with it and maybe one day one day quite a long time from now i would imagine maybe i'll get to play around with real-time gi a bit more inside future virtual reality projects but for now i think the last thing i'd like to do is just give a serious and sincere heartfelt thank you to everyone that has watched my last video i don't know entirely what happened i think i got incredibly incredibly lucky with the algorithm but usually my videos get you know a thousand views if i'm lucky and the last one at the time of recording is around 225 000 views that is absolutely absolutely insane especially for a really tiny channel i don't expect that to keep up in the future i think i have really hit a lot of luck with the algorithm but the fact that so many people have been able to watch and comment and engage has it's really meant a lot to me um so if you choose to say thank you so much and welcome but for now i've got to get back to uh some vr development in unreal engine 4. it feels pretty strange going back to ue4 after playing around in ue5 but it's still a really really powerful tool and it's still a great amount of fun to work with so i'll get out a new video soon but until then thank you so much for watching and take care [Music] you

Channel: Alistair Hume

Views: 43,797

Keywords: Epic Games, Unreal Engine 5, game trailer, gaming, video games, video game trailers, upcoming video games, new video games 2021, upcoming video games 2021, unreal, unreal engine, unreal engine demo, PS5, Xbox, VR, Virtual Reality, UE5 VR, Valve Index, Ready Player One, Metaverse, VR Metaverse, Future, SWO, UE5, UE4, Game Engine, Game Development, Gamedev, Environment Art, World building, VR Game, Next gen, next-gen, next gen game, sci-fi, scifi, Oculus, future city, new features, Lumen

Id: rNOK9RdJDcI

Length: 11min 58sec (718 seconds)

Published: Tue Jun 01 2021