- hace 12 minutos
Categoría
🤖
TecnologíaTranscripción
00:02Desde la primera vez que entramos en SuperSonic
00:06hasta aquí, ahora
00:09es muy bueno
00:10que en la celebración más grande de nuestro juego
00:13en frente de los grandes fans del mundo
00:18celebramos esa historia
00:20a todos nuestros amigos aquí en París
00:24y a todos los que están viendo todo el mundo
00:26esto es el futuro
00:29y esto es...
01:11...
01:12...
01:14...
01:24...
01:25...
01:26...
01:29...
01:32...
01:33...
01:34...
01:35...
01:35...
01:35...
01:35...
01:35...
01:35...
01:36...
01:37...
01:37...
01:37...
01:37...
01:37...
01:39...
01:39...
01:39...
01:39...
01:40...
01:40...
01:40...
01:42...
01:42...
01:43...
01:45...
01:45...
01:46...
01:46...
01:47...
01:48...
01:50...
01:51...
01:52...
01:52...
01:53...
01:53...
01:54...
01:56...
01:56...
01:56...
01:56...
01:57...
01:57...
01:57...
01:58...
01:59...
02:00...
02:00...
02:01...
02:33...
02:34...
02:34...
02:34...
02:34...
02:35...
02:36...
03:06...
03:08...
03:08...
03:09...
03:09So this critical change here is the move to diverse programming language for gameplay
03:13development alongside the low-level C++ engine.
03:17And the reason for this is to enable massive scale code and content interoperability across
03:22games.
03:23Today we have a lot of content interoperability features.
03:26You go to the fab content marketplace, you download a bunch of cool objects, you stick
03:30in your game, but then you have to wire them up and make it work yourself.
03:33In the Unreal Engine 6 generation, by having a unified set of APIs and a unified language
03:40that works everywhere, you'll be able to build smart assets that are completely portable between
03:46games and between projects and can be shared at any scale with the potential to have millions
03:53of developers collaborating together on a shared body of work that anybody can use.
03:58Another hallmark aim for the Unreal Engine 6 generation is to enable far larger game simulations.
04:04You know, the original Battle Royale movie in 1999 that inspired the Battle Royale genre
04:09of games.
04:10Fortunately, it just happened to have 100 people dropping onto an island.
04:14It's a nice coincidence because that's actually the maximum number of players our code base
04:19could support running on a single thread on the server.
04:23If they had 1,000 players, we wouldn't have been able to do it.
04:27But the goal of Unreal Engine 6 is to introduce software and transactional memory technology,
04:33which Curtis Schmidt will talk more about later, in order to enable scaling of games to any amount
04:41of computers that can fit into a data center, hundreds, thousands, potentially even millions
04:45in the future.
04:47And this is technology that's in the research lab, and the team is working very rapidly to
04:51make it into a practical shipping system.
04:55Another major aim of Unreal Engine 6 is something I first talked about at the SIGGraph computer
04:59graphics show back in 2016.
05:02There I talked about a potential future of gaming as an open system like the web is open, in which
05:08any developer can run a server, build a game, host their own content, and the user can go anywhere
05:14they want from place to place, bringing all of their stuff with them, and tearing down the walls
05:20and barriers that exist between games to have a completely interoperable ecosystem.
05:25And you know, the Rocket League preview of Unreal Engine 6, if you looked at the very
05:29last frame of it, there was a little URL bar on top.
05:32The URL was versic1 slash slash rocketleague.com.
05:35And we're tantalizingly close now to fully realizing this idea of an open gaming ecosystem with the
05:44ability to travel from game to game and independent developer freedom to build everything they
05:48want everywhere they go.
05:50And our aim at Epic is to work with all developers in the industry to build this system together
05:56as peers.
05:57We want a system with no overlord.
05:59You know, we've been spending some time fighting against overlords in the industry with some amount
06:04of success.
06:05And we don't want to be the next one.
06:07Rather, we want to be a partner to every company in the industry, including all game
06:12developers and also other makers of technology and services in order to build the best stuff
06:18and connect it all up together.
06:20And so a huge part of this effort is going to be Epic adopting open standards like the
06:25Pixar USD file format and GLTF and also working with standards bodies and releasing specifications
06:32to bring new standards into the industry where they're lacking and needed.
06:37And I'll be talking more about that later.
06:39But right now, I'd like to introduce Marcus Wasmer to talk about the architecture of Unreal Engine 6.
06:45Thanks.
06:58Okay.
06:58Thanks, Tim.
07:00So let's just take a second look at the recent history of Unreal Engine.
07:04So Unreal Engine 4 opened the engine to everyone by making the tools free to use and the source
07:10accessible on GitHub.
07:11And UE 5 reinvented how worlds are built, helping you go bigger, add density and dynamic detail,
07:18and do it faster than ever before.
07:20UE 6 has a different proposition.
07:22It's about how evolving how we ship and operate our games.
07:26That said, UE 6 will keep doing the things you want an Unreal Engine to do.
07:31Rendering will keep getting better.
07:32Cook times will come down.
07:34Iteration loops are getting tighter.
07:36Mobile is increasingly capable.
07:38However, we believe for UE 6 that there are three things about game development that need
07:43to change at the same time.
07:45First, gameplay programming needs to get simpler for creating persistent, live, large scale worlds.
07:52And Verse is the foundation for Epic's future programming model.
07:56It's a next generation language, purpose built to power massive, persistent game worlds at scale,
08:02where global state just works, and transactionally correct concurrency is handled by the runtime.
08:09We've built a modern high-level gameplay framework on top of Verse that gives you a true foundation to build
08:14games and experiences easily,
08:15and to share their interoperable components between games.
08:19You'll hear more about our goals and architecture for this programming model from our technical director on the project in
08:24a few minutes.
08:24But for now, let's talk about the second guiding principle of UE 6.
08:29Content and code should be portable across games and even engines.
08:34Our goal is to give the games industry a whole new way to grow our ecosystems with cross-promotion, portable
08:40player value,
08:41and to really lean into all the positive-sum dynamics that Metcalfe's law predicts for connecting experiences and social graphs
08:47together.
08:48And we're going to do this with a determined effort to move UE beyond just extensibility and into open specifications
08:55for interoperability.
08:57Where existing standards such as GLTF or USD are capable of fulfilling our needs, we'll make them first class formats
09:04within the engine.
09:05And where a standard doesn't yet exist that serves the needs of game ecosystems,
09:09will open up Unreal's own systems as open specifications with first APIs, defined asset conditions, and documentation that any engine,
09:19tool, or studio can implement against.
09:22And Fortnite Cosmetics will be our first real proof point of portability.
09:26We're going to start by moving the base system to an open UE 6 module that's usable by all games.
09:32And this means that you'll have the option to use a player's entitled Fortnite Outfits in your own games,
09:36and you'll get the tools to build outfits for your own games that work inside Fortnite.
09:41And we're tackling this problem first because we want to prove things out with a system that's complex enough to
09:46be a meaningful existence proof of the idea,
09:48and one that inherently comes with a ton of player value by respecting their purchases across an interconnected ecosystem of
09:54games.
09:55We see this as the first step towards building a shared economy for smart assets.
10:00Functional assets with logic and functionality that work across games to recognize players' time and spending in a better way.
10:09In the end, this isn't really a Fortnite story.
10:11It's just about proving that such a mature, complex system can work at scale,
10:15and that every game that works with these systems will immediately benefit from them.
10:20Happily, we're taking the first concrete steps to such an open E system right now.
10:24MetaHuman has been adopted by some of the most successful games in the world,
10:27and is rapidly becoming a standard for digital characters.
10:31And that's why today, we're announcing the MetaHuman Dev Kit,
10:34opening up our RigLogic and MetaHuman DNA tool sets under the MIT license,
10:39so that you can deliver MetaHumans in any game, any engine, any platform.
10:44To learn more...
10:46Thank you.
10:50So, to learn more, please come to our talk right after this.
10:54All right, moving on, no tech presentation would obviously be complete without a conversation about AI.
11:00So, to get one of the fun talking points out of the way,
11:03there's some speculation going around that game engines will be superseded by promptable world models.
11:08And we don't actually think that's what happened next.
11:12What does, we think, is much more interesting.
11:14That AI-assisted game creation will tighten iteration loops, reduce time-consuming manual setup of levels,
11:21character rigs, particle systems, getting bone weights, adjusting lighting,
11:25all of the manual work required to translate professional creative intent into performant and cross-platform games.
11:33And crucially, Unreal Engine is uniquely placed to be the efficient, cross-platform, high-fidelity runtime
11:40that will remain the substrate on which much of the world's best games run.
11:46Thus, for UE6, we see LLMs, Gen.AI models, and tools like Claude, Gemini, and so on,
11:51as playing a central role in building content faster with the creative control that you need.
11:56And as you saw earlier, the future of these workflows is a fluent translation layer
12:00between natural language intent and your production pipeline.
12:05Our goal for UE6 is to greatly reduce the tedious work in authoring content
12:10to leave more time for creative exploration
12:12and increase the amount of iterations a team can make to polish their content.
12:17UE6 will ship with these workflows battle-tested against internal development and in UEFN.
12:24And while we've mostly been showing and talking about content creation,
12:27coding and engineering are important as well.
12:30Internally at Epic, we've been doing a lot of investigation to see what works and what doesn't.
12:34And we recently opened up pretty broad usage for code generation and AI analysis
12:39across our backend, engine, and game development engineering teams.
12:43We've had a lot of great success stories.
12:46We've had particular success with people writing custom tools for their own work,
12:50fast code indexing tools for helping LLMs deal with large code bases like UE,
12:55fast incident response analysis,
12:58automated root cause analysis for crashes,
13:00automated test generation,
13:02and of course the wins you'd expect on backend service development.
13:06But what's interesting is that most of those cases
13:08aren't even generating mainline Unreal Engine code yet, though that'll come.
13:11So you can expect Unreal Engine 6 to include all the key learnings
13:15for using AI for engineering along with content workflows.
13:19Okay.
13:20So we know you obviously have lots of questions about the transition from UE5 to UE6,
13:25so let's talk about that for a second.
13:27UE6 converges two threads of development we've been running in parallel.
13:31UE5 for high-end standalone game and content development,
13:35and Unreal Editor for Fortnite as the live environment
13:38where our new programming model is getting battle-tested.
13:41UE6 will combine these two products into one.
13:44As UE5 and UEFN merge into one editor,
13:47you will be able to ship traditional games and projects
13:49exactly as you do today.
13:52You'll be able to ship directly in Fortnite
13:54or ship your own ecosystem
13:56and optionally make it compatible with ours.
13:59You'll have an easy path from one to the other.
14:02Actors and Blueprints will be in early versions of UE6.
14:07Eventually, these will be deprecated when the new framework is sufficiently mature,
14:10and you'll have conversion tools to move projects from one framework to the other.
14:15Also, a new UE6 mainstream will be on GitHub later today,
14:19publicly visible, just like the UE5 stream has always been.
14:23Any remaining UE5 changes will merge to UE6 to not get lost,
14:27but not vice versa.
14:29And in addition, Fortnite development will be attached to the UE6 stream
14:33so you can see all of our ongoing work live and cherry pick any changes as needed.
14:38However, this stream is not meant to be any kind of alpha.
14:41It exists just to be transparent with where we're going.
14:44And while we're not currently planning another official UE5 release after 5.8,
14:49you know, we do reserve the right to ship a 5.9 if needed.
14:55You never know.
14:57So when can you get your hands on UE6?
14:59We plan to release in early access at the end of 2027-ish.
15:07We know how game dev works around here.
15:10However, Epic will be shipping things on the tech stack ahead of the early access release
15:15just to make sure we get it right.
15:176.0 will be released 12 to 18 months following the early access release.
15:22And we expect just about everyone should be able to move their projects to 6.0
15:25as it will still have all of the mainline UE5 tech.
15:29Except for Cascade.
15:30We're killing that one.
15:31That one's not going to make it.
15:33So get on the Niagara train.
15:35That said, we will be taking the opportunity to make some big architectural
15:39and other breaking changes that we wouldn't normally do in a point release.
15:43So you can expect a similar upgrade path from UE5 to 6 as we had for 4 to 5.
15:49We'll keep everyone up to date on our progress as development continues.
15:52And now Curtis is going to come up and give us some more detailed breakdown
15:55of the gameplay architecture.
15:57Thank you.
16:09Thanks, Marcus.
16:11And good morning, everyone.
16:11I'm Curtis.
16:13I work on the gameplay architecture in Unreal Engine.
16:15And today I'm excited to talk to you about the new gameplay framework for UE6,
16:19which begins with Verse.
16:21Now, Verse is a next-generation programming language, purpose-built to power massive,
16:25persistent game worlds at scale.
16:27We originally released it three years ago in UEFN, where we could grow it to become
16:31the foundation for UE6.
16:33And Verse draws ideas from functional logic and imperative languages, and it should feel
16:38immediately familiar to anyone who's worked in languages like C Sharp or Python before.
16:41But it also has some unique features aimed at solving the complexity and scaling problems
16:47of modern game development, starting with this unique software transactional memory model,
16:51or STM.
16:53Now, all functions in Verse run as part of atomic transactions, which can be rolled back and re-simulated as
16:59needed.
17:00And this lets us do a lot of interesting things in our games code.
17:03So for example, to check if an object can be successfully moved in the world, we can simply set its
17:08position, then do our spatial queries to see if it's hitting anything.
17:11If it is, we abandon the entire transaction, reverting the entire operation, and pretending the code never ran at all.
17:19Now, Versus transactional memory model can also help us address another big problem in game development,
17:24building enormous worlds with larger player counts.
17:27And today, one of the big limiters to really building larger games comes down to server technology.
17:33Scaling a game to run on multiple game servers requires a fundamental rethinking of your entire game's architecture.
17:39But with UE6, we're working to build a full distributed software transactional memory system to address this.
17:45So here's the idea.
17:46We take your existing single-threaded style Versus game code, and then we distribute it across multiple servers.
17:53When an object's needed in one server, the Versus runtime rolls back the current transaction, migrates the object over to
18:00the server,
18:00and then reruns the transaction with the object now present.
18:03And this is all done in a way that retains the transactional consistency of the original program.
18:09And finally, with Versus, we wanted to make saving game data way simpler.
18:13So the Versus runtime can automatically synchronize and save any global state for the program,
18:18and then share it across all running instances of your game or an ecosystem.
18:22And to the programmer, what this means is that setting up something like save state for a player
18:26is as simple as setting a single global map of player to save state.
18:30No more having to set up databases or coordinate schemas between your game and other back-end services.
18:36You can do all of this without ever leaving Versus.
18:40Now let's take a look at how Versus fits into the bigger UE6 gameplay architecture.
18:44So at the foundational level, we have the Versus runtime, along with all the UE systems that exist today,
18:49like rendering, chaos physics, meta-sounds.
18:53With Versus, we're also introducing the Versus scene graph framework.
18:56This is a modern game component framework with full support for prefab-style workflows and runtime composition.
19:02The scene graph has been designed to utilize all the power of the Versus language,
19:06including modeling the entire prefab data model through Versus type system.
19:10Above this, every core system in the engine is exposed as concrete Versus objects.
19:15Each material, mesh, Niagara system that you author becomes a type of Versus class with a statically validatable API.
19:23This lets us use the Versus compiler to validate all of our data in a unified manner across the entire
19:29engine.
19:29For example, if you change a parameter in a particle system,
19:32you'll immediately know where encode needs to be updated to match.
19:36And finally, we'll be introducing high-level frameworks for common game systems like characters, weapons, and vehicles.
19:43Now, building with these frameworks helps speed up development,
19:45but it also creates content that can be shared widely across games and ecosystems.
19:50So imagine grabbing a car from fab,
19:52and instead of simply getting a skeletal mesh and some animations like you might today,
19:56you get a fully working verse module for a car that functions across all of your games.
20:02With the UE6 gameplay framework,
20:04we're envisioning a world where enormous amounts of code and content
20:07is being built and shared between these projects,
20:09with Versus providing strong backwards compatibility guarantees
20:12to help the ecosystem grow sustainably.
20:15Of course, we'll need a place to store all this code and data,
20:18and here to talk about that is Matthias.
20:31Thank you, Curtis.
20:33I'm Matthias Jansson. I work on developer tools,
20:36and I'm here to talk about our version control system lore.
20:41As we continue to scale our platforms and services,
20:45we run into many challenges when it comes to data storage and version control.
20:50We need something that can handle large binary data sets,
20:54scale easily across the globe,
20:56and provide secure data isolation in a shared multi-tenant environment.
21:01We also need to make it easy to use for everyone, not just programmers.
21:07We looked at the available tools and came to the conclusion that none had everything we needed.
21:13Git, for example, is excellent at what it was built for,
21:18distributed development of source code.
21:20But game development has huge amounts of binary data and content
21:24and rarely needs everything on your local machine.
21:28Git doesn't really work well under such requirements.
21:33Other commercial and proprietary systems were also considered.
21:38But our foundational tools have to be open.
21:41Open specifications, open protocols, open formats.
21:45Anything less, and the people relying on them can't extend, audit or trust them.
21:50So we decided to build Lore from scratch,
21:54and we designed the data model to meet our requirements from the ground up.
21:59We made scalability and performance core tenets from day one,
22:04and this is where most of Lore's technical innovation lives.
22:08It combines an efficient large asset handling with a free branching model
22:13so teams can iterate quickly without sacrificing workflow flexibility or storage efficiency.
22:20And Lore will grow with you from a zero-config, single-service setup for small teams
22:27all the way to multi-region, multi-million user deployments like UFM
22:32without requiring your team to fundamentally change how you work.
22:39At the heart of Lore is an efficient content-addressed data store.
22:43Every file, every directory, every revision is identified by what it contains, not where it lives.
22:52Identical pieces of data are only stored once.
22:56Lore gives you direct access to that storage platform
22:59so you can build your own applications and pipelines on top of it.
23:04Lore's version control uses this storage platform to let you work on multiple things in parallel,
23:10switch between them freely, and collaborate in safe, isolated spaces without worrying about how it scales.
23:18Lore's sparse binary-first storage architecture fetches only the data you actually need
23:25for the revisions and branches you work with, supporting projects of any size.
23:31Large binary data is stored in smaller fragments, allowing partial deduplication and highly efficient,
23:38parallelized data transfer of any file.
23:41That same data model is what lets Lore scale out across the infrastructure.
23:46As you grow, Lore lets you transparently add layers of additional servers from the local machine,
23:53out through edge and regional caches, and up to a durable global store.
23:58And because access is authenticated and authorized at every layer, that scale never comes at the cost of control.
24:06Every server enforces the same access policies.
24:11For a globally distributed team, that means that users in different regions can hit their nearest cache,
24:19and the global source of truth is never the bottleneck.
24:22Everything Lore does, data storage, version control, is exposed through a single complete API.
24:29And on top of that API, we built the tools you would expect.
24:32A command line interface, a desktop application, and in the near future, an integration in Unreal Editor.
24:40And these tools are all built on the same API and library.
24:44This API-first design makes Lore easy to extend and integrate into any development pipeline.
24:51Your tools and applications can easily use Lore through the API in the language you want to use.
24:59Though we're just scratching the surface here, I hope this gives you a sense of the power and potential of
25:05Lore.
25:05And it's already used in production environments.
25:08It powers the UEFN ecosystem, where it's been known until now as Unreal Revision Control,
25:15providing version control and data storage for everyone's content pipeline.
25:20We have already started transitioning to use Lore as the primary version control system for development within Epic,
25:28and we plan to be fully using Lore by the UE6 launch.
25:33We built Lore to address scalability and efficiency challenges we have at Epic.
25:39And as we talked with many of you, we heard similar challenges echoed across the industry.
25:44And that led to a bigger realization.
25:47If we're trying to foster a truly open ecosystem, one we all build together,
25:52then a system shape behind closed doors would fall short.
25:56And that's why we're announcing Lore as open source today.
26:08All the source code, along with documentation of data formats and protocols,
26:13are now permissively licensed under an MIT license and free to use.
26:19This release shouldn't be viewed as a perfect or complete package,
26:23but as an invitation to you to extend it, challenge it, and help shape it into something even more powerful
26:30together.
26:31So, if you're interested in learning more, head to lore.org and come to my session later in the show
26:38for a deep dive discussion on Lore's architecture.
26:41And with that, I will hand it back to Tim.
26:43I like turn on wollen.
26:46Hear where I go.
26:47Gracias.
Comentarios