Unreal Engine Development Update | Develop Brighton 2024

All right, testing. All right, cool. That works. Uh, people are still coming in, so uh I'll just uh get started because we have a lot of stuff to go through today. Um, yeah, my name is Arian Brussi. I'm a technical director at uh Epic Games on Unreal Engine and I'm going to talk to you today about um what's been happening in Unreal Engine and what's going to come up next. So, we're going to go over like what's going to come in like 55 and future releases and kind of as a first glimpse because we haven't shared this uh broadly yet. So, this is kind of one of the first um showcases of like what's going to come in the next few releases. Um, as a disclaimer, because we always want to do disclaimers. Uh, you know, we generally want to underpromise and overd

deliver, which means that in the past we've told you nothing, right? We always surprise you like, hey, new release, all kinds of cool stuff. But, uh, but, you know, we kind of want to change that a bit and kind of tell you what we're working on, like what where we're investing. Uh, so that you can make choices as well, like where you want to put put your own efforts into the technology. Uh, uh, for sure. So but you know um the plans are tentative right we we're trying to release these things on this time frame but it's development like you all know it will take sometimes longer um so yeah let's start by uh you know just just quickly reiterating the pillars of Unreal Engine 5 and uh really the aim for UE5 is to uh enable this generational leap in uh in graphics fidelity uh and also dynamics supporting huge worlds uh seamlessly uh with like faster workflows, improved workflows, but also making a lot easier to use and extend and uh you know we are

hitting good strides on all of these areas but we're not done right these are kind of the the pillars for the entire development of the entire engine. So we'll keep hammering away at at at all these uh these things. uh but it's also dynamic plan right we will add new things when we see them come up for instance uh PG was a big new feature that we introduced last year that we found while we're developing all these things and these demos uh we had to kind of create different tools and uh add that to the list so uh what I want to do is it readable there's a lot of bright light it's really washed out you don't know okay if it's not readable you should uh put up your hands so we can maybe do something about it. But uh okay, so what I want to do is like really quickly go through kind of the journey of it, right? because the the journey uh shows where we learned things and what we kind of kind of developed and you

know really it started with this uh demo that was in what uh 2020 already crazy long time uh where we really kind of used uh uh the new rendering ideas and the R&D kind of showed it in kind of an early version of of UE4 to kind of see what what uh what what it could uh be like. It all was running 30 Hz on a PlayStation 5. Then a year later, we took all these learnings from that previous demo and kind of we we said like okay well all these things all the pain points we had in delivering that earlier demo like we need to make better tools right we need to make tools to handle world streaming and creation uh much easier. How do large teams work together on on large files? Uh how do we have dynamic worlds that can change with with data layers and all that? So a lot of time was spent on uh workflows and performance and uh and uh to kind of scale up the creation of those kind of things as well. Um then you know running into end of 21

uh we uh showed off this new demo which was Matrix Awakens back in the day uh where we really wanted to aim to make like large scale photorealistic uh dynamic worlds, living worlds uh where we used Houdini tooling to make this world and make like importers and and uh and processors we call them to kind of take that Houdini export uh seamlessly into the engine and and all that and then worked on like a new AI system to kind of run the big uh simulation uh of it as well and you know and uh you know many many other things. Um but I think also was important at that at that time that this was like peak co and uh you know we noticed that the the ways we were working was starting to get really different. So we had to invest a lot of tools in a lot of time in tools to make uh workflow improvements for this hybrid setups uh or even fully remote setups.

So then uh you know we released uh all this new tech on Fortnite right which is kind of our our proving ground always to make sure like is it ready for uh what we call like production ready right uh can this run at performance all these platforms do we have pipelines to support all that and uh that's what we did uh here uh we also made this tooling to make like uh upress old assets to like nanite quality using kind of pre-baked displacement and kind of make it all run at 60 Hz on on current gen consoles. Then um you know based on the learning of making that city right we we where we're using Houdini and kind of offline kind of import export uh uh uh type uh pipelines you know we uh started to invest back in the day to work on like a procedural content generation framework to kind of do it inside the engine because the thing is if you have worlds of this quality and this density uh you

can't render it externally anymore. you have to be in the editor and also the workflows and the pipelines will become much faster. So we invest a lot in uh in um the PGD tooling uh for uh for the the editor and this was kind of our first showing uh at that time. We also showcased our new uh uh material system substrate very early uh view on that and uh and you know the the first release of UEFN that uh that is kind of the UEI tool set inside Fortnite and using Fortnite as a platform and then recently you know our probably our biggest release uh so far in the lifetime like really kind of uh uh uh dumping a lot of new things in there as well you know uh going from like nanite uh you know improvements with like real-time tessillation and displacement right that helps to kind of uh get much more detail out of smaller assets by just using uh uh texture maps uh you know doing compute and variable rate

shading uh to kind of get the performance up a lot uh lumen with hardware rate tracing but also excitingly what you see here uh finally we're releasing uh motion matching uh to uh really get get the animation quality up. You know, in the past, we always felt that uh animation in Unreal was uh was not up to par with what we wanted to achieve and with what what best-in-class games were doing. And with now the motion matching support that went through a lot of rewrites and all that, we now have have a really solid system. And uh we also released this as a sample uh uh to for people to to to uh to play with and test the systems with. Um yeah and a lot of other things I think uh you know interestingly enough as well um we internally now shipped uh games using world partition. So you know previously right there was always our tech demos and stuff but now uh the the Lego Fortnite game and Fortnite in the

in the near future will also kind of fully use world partition and that's running on all platforms. So you know this world partition with many layers and data layers and one file practice all the things is running on the switch right and it's running with PCG enabled to make the the the Lego game run really well. So this is uh quite ready for uh prime time. So uh that's all being said let's talk about what's coming next. So, uh, starting with rendering and, you know, I'd like to kind of go over the the aims, right? Well, obviously with rendering, we want to kind of push the state-of-the-art of of graphics quality and performance for uh, games on consoles and, uh, you know, support high-end photo realism uh, not just for games, but also for other industries such as film and VFX and architecture, product visualization. It's used everywhere nowadays, but, you know, we're going to talk mostly about games today. Um so yeah uh starting with

Nanite. So uh the big goal of the year for the team is we have full R&D happening on foliage. So the use case is really right like think about like a a dense forest you know up to the hor horizon. How do you render that efficiently? The problem is that uh you know the polygons are just an aggregation of like little leaves going to the far distance and NI is not handling that super well. So we we're looking at all kinds of systems to uh to make that uh uh perform uh on like gen 9 consoles. As part of that as well we are looking at uh skeletal mesh support for nanite uh which is going to come in experimental mode by 55. So you can have like internal hierarchy in it, animate uh trees, whatever characters uh uh using uh using nanite, even apply uh displacement on top of the the the skinned meshes. So there's going to be like really interesting workflows that

you can can think of. Um you know, and then uh for the future, you know, you see that image there. I'm uh these are one of the the the the directions we are exploring to render far away uh uh foliage you know uh you know maybe you can think about foxholes right we are not committed to that approach yet but but these are part of the the explorations we have into rendering this at efficient levels in the far distance uh lumen we have a big effort going into what we call like many lights so a lot of uh you know hundreds of like local lights that are uh in in the world reflecting casting shadows and and and all that uh on console in game a lot of R&D is going into it looking pretty promising but we're not you know setting any any any target date yet uh you know generally for the future of Lumen we're both looking at like the high end like

how can we ramp up the quality for some of these use cases and get to like even even better fidelity But also look at uh making it run faster because a lot of games are looking to do uh performance modes, right? Running at 120 Hz, 144 hertz, whatever. Uh you know, can we run lumen at like sub 2 milliseconds per frame is a uh challenge that we're looking at to uh pursue. Um so generally on the rendering uh thing uh area uh you know hardware rate tracing want to support that out of the box running at 60 frames per second. We plan to ship Fortnite using that uh you know this year likely. Um you know we're working on like path tracer and uh and getting that ready for enterprise and and linear content um you know experimental things there with uh running it on Linux and and those kind of things for big render farms and

whatnot. Um, substrate we are still in heavy development on that, right? Really the path for us to deliver substrate in a in a good uh way is to kind of make Fortnite run on substrate without performance uh degradation. And then we say like okay well it doesn't it's not slower than than the stuff that we have. And now we can extend the the the fidelity on top of uh of the base quality with with with all the new tools. Um yeah and then uh on architecture uh level of rendering right really the big goal is to enhance uh renderer parallelization uh you know really using uh all the cores and using all the new kind of systems like work graphs and enhance barriers and stuff like that to uh to make it uh scale much better over over all the cores. Um feature- wise we're also looking at like a new uh GPU profiler that can support like the

modern concepts just like um uh async compute and mgpu and those kind of things right old profiler didn't do that nicely uh we have uh support coming for bindless resources which is really exciting to enable kind of much uh you know more flexible uh visual uh visuals for artists to add uh and and and and use various textures uh across across all the other shaders and uh looking at like overall um improvements in uh shader permutation and uh PSO caching and those kind of things by by by leveraging all these new systems uh going forward and uh a big aspect of all this is that we're looking at a new material editor to really uh make it much easier for artists to make performant uh materials out of the box without uh uh automatically entering into like the thing that that has huge amount of permutations really slow you

know everything explodes etc. So we want to kind of fix that by our new material editor which is you know future very experimental uh you know just that that we're working starting to work on that right now. All right. Uh world building. So, um this is obviously part of our our our uh goal to make like large uh worlds uh efficiently with uh with uh a new engine 5, you know, and uh and have much better workflows, but also have out of the box um uh streaming capability uh without having to write it all yourselves. So, you know, this is used internally a lot right now and all the licences have been jumping on it. So we are we're we're making a lot of improvements constantly to performance workflows and uh and and and all those things. And we're also kind of scaling the whole size of the world that we can support up uh massively. So I think the the the Lego

Fortnite world is on the largest dimensions like 20 by 20 kilometers now. So that is helped by um large world coordinates by all these new systems by instancing landscapes etc etc. uh which kind of allows us to kind of grow the the the size of the world. Um then uh with PCG um you know really the goal again is to kind of keep people in the in the editor and have people make uh um really um uh rich content by making rules and without having to hand place everything, right? That this became really apparent when making a city like this which you can't do uh uh by hand anymore. So uh that's all runs in the editor uh right now. But uh exciting as well is that it's a runtime system. So uh it doesn't just run in the editor. It can run runtime. It can run at cook. It can run at load time. It can

run on uh on the GPU going forward. So this will be a nice system to kind of really kind of beef up the detail uh across the the um whole world. And yeah, you know, I think the other thing to mention is that uh when we are going to like 16 by 16 kilometer plus, right? Who knows how big uh and we have millions of actors, you know, maybe tens of thousands per cell that you're streaming in, you know, we're starting to hit a bunch of uh inefficiencies in like spawn time, load time, uh etc. that we're hitting uh uh constantly. So we're we're working on optimizing that like think about like add to world on like physics and and things like that. So uh with five oh yeah first um brief interlude. So uh this is a good guide. I don't know if you've all seen it. I've been sharing it every every time I talk to people. Uh this is our new world building guide. It's like I don't know a lot of pages like 50 pages or something.

It runs uh to all our best practices until 5.4. for and shows you how to use all the systems in the best way that we uh can uh tell you right now. So please uh give this to your team, to your tech artists, to your uh engineers, whatever. So you can find it on deficgames.com. A lot of uh cool content there. But yeah. All right. So um on world building specifically uh yeah a lot of work again on like the workflows customization quality and all that uh you know exciting things like external data layers right not everything is combined in the same object same level you can have like different files to kind of contribute to like data layer setups and uh and and whatnot. Um, level instance property overrides are uh kind of a basic early prefab system. You know, it's not like

our scene graph that we're uh proposing for like UEFN that we have a prototype of in UEFN, but it is a way to kind of instantiate objects uh big uh uh accumulated objects and change things on those instances and kind of build large worlds with lots of buildings and houses and whatnot uh uh efficiently. Um let's see. Um then um you know oh sorry pressed the wrong button. Um yeah then uh going forward um you know a lot of work on performance and and tooling to make all this stuff uh work well right there's so much data that that that you're streaming through it you know especially if you have like many many data layers and many world partitions stacked on top of each other it's good to have better tooling to kind of understand where things are are going and what is loaded and uh and where these issues potentially are coming from. uh you know the big thing is

towards our new scene graph and that's really the thing that we've been prototyping in UEFN uh that's kind of a full uh rewrite of our uh actor component type type system um that we want to bring back into UEIE proper but this is probably a UE6 thing but uh you can start seeing what we're doing with that stuff in UEFN so it's kind of interesting to take a look what's what's happening there because you can see about like okay if you have this compos positional object model uh um you know combined with verse how can you kind of apply that to to making making uh making uh games in the future, right? The biggest challenge we have right now is to figure out the backwards compatibility to to say like well how do we go from UE4 style uh actor components to like the new scene graph you know that's the thing we want to solve because we want to keep the engine uh future versions of the engine being able to read old content as well.

So yeah, PCG a lot of work going into PCG. Uh you know we have released like the what we call like the biome core. So it's like the first uh kind of uh meta system that's that can handle biome. So more complex PCG setups uh to make make larger systems. And we're also working towards GPU compute. So the first version of that coming in 55 we're planning. So that means like small PCG graphs can run completely on GPU compute. So really kind of uh and kind of interoperate with nanite. So they can spawn nanite meshes completely on the GPU and really get to like much higher uh detail levels on um on the graph execution. You know that's uh first getting faster but also uh you know get gives you much more uh control over where the PCD is running right. So is it a baking process? Is it running at runtime? Is it running at load? Is it

running at at at uh at at the GPU? And trying to figure out if you can find uh overlaps in um in computation, right? If you already have calculated it for a far away LOD, can you uh use the the result of those calculations for lower LODs when you move closer, right? So you don't calculate all the things again uh you know when you're moving to the world for instance. So a lot of work on that uh as well as like interoperate better with um spline meshes Niagara uh those kind of things. Um in the future I think an exciting thing what we uh are working on is um called mega roles which is a initiative coming out of the quicksol uh uh group and really we want to uh think about like well you know assets should become like smart assets right so if you have a tree maybe it's not not just a tree but a tree knows about the biome it's part

of it knows how to spawn other things around the tree rocks bushes, uh, moss, etc. So, all these Quicksol assets, all everything in that library will become like smart assets, uh, as part of our Mega Worlds, uh, initiative. So, we're going to kind of release kind of various biomes through the through the through the next couple years to kind of showcase how you can make these kind of rich uh, assets and uh, you know, as examples, right? Because we expect everybody to make their own uh, um, you know, extensive biomes uh, etc. using all these tools. So, you know, really the goal is for um content developers to not have to write PCG graphs themselves, right? Maybe your tech artists are making a PCG graph, but then your environment artists are just placing objects that have embedded PCG graphs uh for instance. And then um you know also kind of more future uh uh like you know landscapes

are are a big focal area for us. uh terrain uh you know you call it um you know right now uh often the landscapes are you know if you look at like Fortnite or something are the lowest quality uh uh uh geometry on the screen. So we really really want to ramp up landscapes to like uh um you know nite level of of quality uh without entering like in our kind of previous tech demos in like huge overdraw and really kind of in a nasty type type situations again. So uh we're working on like a completely new system that can combines uh things from you know full 3D landscapes and height fields and uh and think about everything as like a mesh and uh and uh and all that. So a lot of R&D is going into that. Uh we're working like first prototypes and uh and all that but a lot of work going into that. This is probably uh beyond 2025 but yeah and uh you know man it's hard to see uh

really uh exciting as well. We have used this in Fortnite since a year or two or something. Uh we have a time of day plugin inside the the the the game or the engine coming uh that's using sequencer. So you can use sequencer to uh to uh you know make make animations on and and settings for everything that that that you can attach to sequencer uh to make like really rich time of day visuals. All right. Uh animation. So, uh really our our goal here is to push for a full uh animation altering pipeline, you know, bypassing the need to use uh external DCC's. And why are we doing this? Well, the really the the goal is to improve that roundtpping time, right? Uh but also what we're finding is, you

know, people can actually animate in context in final pixel. So, if you think about like making a movie where you have lots of explosions and lighting and it's all dark, like why are you animating certain things when they're completely in the dark, when they're completely hidden by all the VFX and and all that? So, um that um has been going on for for for some time. We've been dumping a lot of energy in this and um you know what's what's part of this as well is that it's really it's an it's one of our most open and extensible frameworks in in the engine. So you can use Python, you can write your own tools on top of this uh to make like a really custom uh pipeline for your animators and you know ultimately we'd like to get more people be able to animate quickly inside uh inside Unreal. So a lot of areas uh on animation um you know on deformers right we are enabling that into into sequencer and then uh making that more modular uh going

forward um you know that whole uh control rig and modular control rig where you can like easily make make components and and attachments and and and uh dynamics on on on a skeleton that's now uh moving to beta and that will uh in the future get a lot of work with interoperable with uh physics and uh uh the former graph and skeletal editor kind of improvements in that in that realm. Uh you know the skeletal mesh editing was this missing aspect of uh the engine where uh you still had to use an external tool for that. So that's now completely built into the Unreal editor and quite capable. And um you know we're we also have like a lot of work going on to like retargeting where you know you can move these animations across different style assets and kind of keep the the the movement of the feet and all

that in in check. Um then uh you know last but not least on this one um ML deformer has been um um you know getting constant updates right it's a really powerful system and I don't see a lot of people using it yet but it's really uh performant right it is uh running uh on Fortnite we have it on on various characters and it's running like sub millisecond on you know full muscle sim kind of output on like the Hulk character and and stuff like that. It's uh it's really cool to look at and it's basically it's one of these nice use cases for ML where you uh train simulation like hugely complex uh simulation and kind of you know output into this really simple model that you can uh run in uh in runtime and um much more of this is completely able to run inside the editor. So you don't need full muscle sims or simulation tooling

in external tools anymore. um on sequencer. So, uh sequencer is getting a bunch of love especially in like the UX uh area, right? Uh you know, we really want to aim to broaden the appeal of sequencer across uh the the um the uh editor. Uh making it more more clean to use, but also we are seeing it being used for kind of any of the timeline tools inside the the the editor. So uh you will see sequencer pop up in in more and more areas. Um we have a uh effort called dynamic sequencer that uh we we we we plan to use uh that to extend a sequencer into more dynamic uh uh uh operations to basically make interactive cutscenes easier. Right? So go from your game to a to a a sequence more seamlessly without having to do a lot of hacky uh effort right now. And

really that's kind of runs into our kind of future efforts into like a cinematic dialogue system. So can we make a basic uh dialogue system uh you know branching dialogue system in the future that as kind of part of the engine that you can all extend obviously you know um a purely animation in the engine um you know like uh really uh you know for the animator um you know a lot of work has been going into that uh you know animations layers and also we're getting this focused animation mode you know which is kind exciting because we want to switch the UI of the editor because it's getting complex again uh um to be more focused around the persona that's using the editor for a certain uh purpose. So animation will hide a lot of things and kind of focus on animation. You know, also what's probably interesting to to mention is um for um

uh Lego Fortnite, for all our Fortnite cutscenes and also starting to run in Fortnite itself. Uh we're all using this internally. So all our animators at uh Epic are using this tooling to make old animations, right? For Lego Fortnite, we have not not used anything but Unreal to to animate all of this stuff. So that's great because we have a lot of animators screaming at the programmers to make it make the tools better. It's this is terrible like please uh you know this and and you know we we make a lot of progress that way right. So that's now all in 54 and you know further 55. So yeah uh motion matching uh yeah really happy that we are now releasing this the or we we have released a sample. You know, obviously these are complex systems that only run with a huge amount of data. So you can only evaluate it when you have a huge amount of data. So we have hundreds, I don't

know how many crazy amounts of uh of animations as a base uh version to for you to test the system with and you can use them as kind of your base layer and then start layering things on top of that. But also say like, well, this is actually working really well for our type of uh uh game. Maybe now we can mo cap this ourselves and clean it up ourselves to have a more unique uh set because I'm sure not everybody wants to use the same animation as we do. But uh you know we actually use this for Fortnite as well and kind of made like a stylization pass on top of that to kind of turn it into Fortnite animations but it's using the same underlying uh animation mocap data actually. Um and we're expanding that a lot right the team is continuing to work on like multi-acctor support uh you know uh tooling around picking the right trajectories and uh and such things uh as well. Then on uh the systems side of animation, you know, previously we talked about like the tooling. Now we're talking about like more of the the

gameplay systems uh side. Um you know, we're working on this new gameplay camera system to really help, you know, interpolating between like uh gameplay and cutscene kind of uh uh cameras alongside with giving debugging options like what's uh you know to make that happen easier. Um a lot of work on like various debugging things, right? That's that's again right we have people internally using this you know that have problems like how do we debug this stuff so we have to improve these toolings around uh uh debugger so you now can save and load the state of your full animation uh diagram and setup but also kind of connect to a console to see what what is happening why are certain animation being picked uh you know why are the blends in the blending like the way and scrub through that which is uh quite amazing Um and mutable we're also getting the mutable uh middleware that we have acquired a couple years ago uh as part

of the the the the engine. So that means you can use mutable to make custom customized characters uh easy and making as an integral part of the of the engine uh going forward. This also improves cook performance and support content on demand with mutable. So and then um future on systems we have an effort that we call NM next internally. It will probably get a better name uh the moment we uh we actually productize that. But really this is a way to make a new kind of unified uh animation pipeline where um we want to make the animation graph logic much uh faster by leveraging all the work we've been doing on rig VM for control rig. So we want to use run all these things on the same uh hyperefficient uh VM that we built for control rig and kind of integrate all

those things. So um yeah we plan to run all our next generation logic on that and then we expect to not have to nativeize animation graphs anymore. Right? right now you know that that set graph I don't know if you can see it but blueprint right now uh gets really slow uh if you have complex animation graphs with Rick VM we I think we can get it to at least five times faster maybe 10 times faster at some point it doesn't uh require you to nativeize it anymore so okay then on developer iteration so a lot of work going into that right games are ever expanding in size complexity data uh etc. uh teams uh you know we hope for them to uh to uh uh make it make games faster and iterate more which ultimately gets to better quality. Um you know exciting that we have uh Unreal built accelerator now supporting shaders

but also Windows, Mac and Linux uh coming with 55. We have uh Horde CI coming into production readyness. So that's cool to run all your continuous integration, right? Horde does much more than continuous integration, but this is the part of Horde that does that. And we also have a new thing which you called the UEI studio telemetry that you can hook in all your tools all the horde and UBA and editor uh stuff etc into a kind of a telemetry uh pipeline that you can hook up to your own telemetity backends and kind of show developer efficiency crashes uh you know spot trends in like coupe time and load times and editor stability and whatnot. So yeah on uh on the editor and target iteration we have a lot of work going into uh all these various things uh you know like Zen loader is really kind of optimize uh

loading in the game but also in the editor. So can we get the startup time of the editor much faster by async loading things without having to read through all the all the world partition stuff you know shared DDC and cloud DDC and those kind of things and cooked output store is really about like well if some person on your network on your team uh transformed an asset or cooked an asset whatever can we uh leverage that that effort to through the entire team so you don't have to do that work double um And then you know uh a bunch of other things. I think also important um the Zen target platform streaming. So you know for the longest time Unreal was really efficient in the Pi setup, right? So you're on your PC going into Pi going back etc. But really running it on a on a on a target device or a console, it was not great. And we're trying to fix that with uh the target platform

streaming where you don't have to package and uh and cook and uh and make monolithic kind of things that you copy to your def kit but really you stream it from uh you know the editor from all these toolings that we have directly to the to the uh target platform. Um yeah and then down the line um you know we do want to have a big rethink of like how we cook etc. you know, we are digging down into uh you know, making incremental cooking instead of iterative cooking uh the default. So that will improve a lot of the the speed there. But also looking at like well can we have like cooked editor support out of the box for uh for a UEI. All right, cool. uh simulation and I'm I'm starting to run out of time so I kind of have to accelerate a bit uh simulation both physics and uh particle

effects uh you know a lot of lot of uh improvements coming there with uh physics um you know we have uh you know most exciting things perhaps just talk about like the ML flesh you know we're expanding that a lot a lot of this stuff is going to run completely in the editor uh going forward so you can have full muscle muscle skeletor skeletal setups rightly editor that can uh run on any metauman. So they have full deformation running ML deformer etc. uh we have the chaos visual visual debugger coming online in 55 which is really exciting to kind of actually see where where things are going and you know generally we're working a lot of performance right across the board to help uh help even more complex things um uh Niagara a lot of things in flight um a lot of a bunch of things going into production ready with the data channels

and lightweight emitters that are that that are really really interesting you know stateless emitters coming down in the in the in the future. Uh but also looking at like uh you know improved iteration time in like the editor and the cook aspects of it. Uh you know performance you know can you run these things at uh at frame rate on on devices and all that and you know looking at like things like running nite particle renderers and those kind of things. Um yeah and then uh a lot more work in like performance and inter interoperability and and and all those things. And I think you know you start to see the benefit of having like our own physics engine and our own uh advanced VFX system and uh and PG starting to kind of all kind of interoperate and kind of get to much much more capabilities when they all are working together seamlessly. uh then framework um so that's gameplay

AI and uh and networking so a lot of work uh going into uh two areas like uh really like workflow improvements right like can we make it uh uh you know easier to use inputs new movers gas modularization state tree and all that but also performance right so uh much more um lightweight actors uh uh mass gets gets extra functionality, you know, instancing by default. Uh, you know, so everything is an instance instead of just an an actor.