Session: AI Production for Game Studios – How it's Going And What To Expect In 2025.

[Music] I would like to welcome to the stage if I may our uh our next speaker now representing leer AI we have uh a talk about AI production for game studios how it's going and what to expect in 2025 so please can I welcome to the stage uh y Raku from Le AI Round of Applause please for our speak good morning everyone um everybody's talking about AI today it's all over TV it's all over news it's all over your social media and today is going to be no ception so hopefully you got in the

right meeting room my name is YCO and I'm the VP of engineering at layer AI layer AI is a generative AI power tool used by game studios to produce 2D 3D and video assets before layer I used to work as a senior principal engineer at Unity Unity where I help to ship Unity cloud and Unity gaming Services overall I have about decade of experience in gaming industry but one could say that that I started my career pretty early I think the first game Studio that I founded was when I was 8 years old along with my friend which we call Dragon arrows my uh role in that startup I guess if you want to call it that was to come up with the stories for the games and draw sprites on paint uh so you could say that game development today looks quite a bit different than uh it did 25 years ago um but yeah let's jump into the content of today malfunction okay there we go okay

that's I I guess you didn't see the slide before that's where we go cool so I'm going to get us St by talking about the current state of AI very briefly and thereafter we're going to focus on some of the unique challenges that game studios today are facing and what is ler doing to address some of those challenges lastly we're going to finish off with some future look ahead in terms of what we're planning to do in the future to tackle some of the challenges of today and of the future hm oh sorry my bad good um you might have all heard this quote I said before AI is no longer a competitive Advantage but it is a necess necessity that this is more true so than ever before both for us as individuals and as companies in order us to be competitive and on the market we need to use these tools to get that extra Competitive Edge we've also know there's a couple of like interesting uh themes happening in the industry the first one is that everything is moving so fast there's new new models popping up every single week

and it's really hard to keep up with what's going on the second thing is that the legislation is actually lacking behind which puts a bit of turbulence into this industry as well the third which one is very interesting is that the base model layer so this is the llms like GPT or llama uh and then also the image generation tools like flux or stable diffusion and more they're be becoming a commodity they're no no longer viewed as Novelties but they're becoming more of the building blocks of the future and today at ler we've had the privilege of partnering with a lot of game studios here's a couple of those um and we discussed them about their challenges related to Ai and here are the four key things that we noticed the first first problem that they're struggling with is that AI is moving really fast oh actually the first one is that AI tools are not Enterprise ready so this means that they're missing the mark on some key security compliance requirements like having so to comp compliance or ISO 271

the second one is that this this area is moving so fast that it's really hard to keep keep up to date with what's going on as soon as you're done procuring one tool there's already the next one available and you're spending all these Cycles procuring tools instead of actually building your game the third challenge is that AI tools are still complicated to use AI itself is not novel but it's very novel in the use of consumers like GPD came out a couple of years ago but we're but is that going to be the end or final way of interfacing with AI probably not and we're still trying to figure out the best way of getting that done the fourth one is that gaming demands more than just a general purpose AI in gaming we have very unique tools that we use blender God do unreal uh unity and so forth and the best tool seamlessly integrates with these things beyond that you also need access to collaborative uh features such as being able to comment and collaborate with across developers or an artist on that tool as well so let's dive in the first

challenge so AI tools are not Enterprise ready so we work with a lot of larger game studios like Zinga and what they need is Rock Solid security from the platform that they use this means having things like so 2 compliance and ISO 270 certifications beyond that they need direct integration with active directory or the employment management software they also need the ability to manage permissions on the platform as well the third thing is what they need is access to copyright safe models for produ use but they also need IP safety which means that the data gets processed in a way that it doesn't get exposed to the public the fourth thing which is very important is having access to audit trails and observability so if there's a dispute about a generation for example infringing on a copyright you can always Trace that generation back to how it was being generated where copy unsafe words used and so forth the fifth thing is that as AI is taking over more and more there's going to be more and more content being created the challenge is going going to be how that content gets

managed which why why which is why asset manage becomes even more crucial at layer we' imple we exposed some uh copyright safe models such as Pria we've also implemented a lot of these features we have a drive to manage assets we've implemented audit Trails uh and we also have a way to we have scheme integration and we have a way to for the users U to manage teams uh and uh members and we are are also so to compliant the key thing here to know is that like okay we're here to talk about AI why am I mentioning Enterprise features uh they're not that sexy honestly but like that is a requirement for Enterprises to actually use these tools it's Al it's always nice to be very flashy and do the all the all the cool stuff but you need to get the basics done right as well the second challenge that a lot of these companies are faced with is that it's really hard to keep to up to date with the state of innovation everything is moving so fast and like I said as soon soon as you're done procuring one tool and Having learned one tool there's all already the next one available it

takes a lot of time what custo what our customers need and what game companies need is uh access to the latest models as soon as they're available because that is a competitive Advantage they also don't want to spend time on proc like spending money and time on procuring these tools and they just want to build the game they also need somebody who's going to make sure that as the regulation uh evolves over time they stay compliant with that process Gallum Studios which is a smaller studio uh is building the next generation of Sims so on the left you have some of the assets that they were already generating with layer before and as soon as 3D models and videos were available they were able to solve for new use cases on layer so uh I guess the 3D one is not playing there but they B basically what they did was they took some of those images and then turned them into 3D models directly on layer but now what they're doing as well is generating these videos uh based on those two images as well back and yeah yeah

okay there you can see the 3D video yeah cool anyway to jump to the third one so AI tools are still complicated to use so what uh what the game studios really need is an intuitive ux so that artists can directly tap into the value without having to spend a lot of time learning a new tool further they need deep integration with their existing software so this means what I mentioned earlier Unity go do unreal blender maybe digital Asset Management tools that you might be using uh they need help onboarding their teams onto the platform of their choice and often they also need uh specific AI expertise so in the best case you get really good results from AI by itself using the tool that you have but the reality is like to get the best absolute best results you still need specialized knowledge that we can also help with uh huge games uh there example of huge games but something that we made uh very easy at layer is the ability for you to create your own custom style so

what you can do is you can come into layer with like 8 to 10 images upload them in we create a custom F tune what they call it and based on that fine tuning you can create hundreds of thousands of images in a matter of seconds in the style that you originally had right uh which saves a lot of time the fourth challenge is that gaming demands more than just a general purpose AI it's not enough to just be generating assets but you need a tool that is going to specifically work in the context of game gaming has these unique unique needs like you need absolute style consistency when you're building a like high quality game the second one is that you need the seamless tool integration and another thing which is interesting in gaming is that we have a lot of these repetitive tasks like you're generating mobile banner ads and like I'm sure the game artist don't want to spend time like resizing and moving some objects around on a cannabis like in 100 different different sizes across different like um you know devices like Android or iOS it's going to be more convenient if you can automate that task

of like basically generating those banner ads and spend time doing more creative stuff right uh and the another important thing is that you need the collaboration tools directly on the platform in gaming you have you're going to be interfacing across artists game developers game directors art directors and more and you need those collaborative collaborative tools there as well a game uh is benefiting benefiting from the style consistency so they trained various different styles on layer but here's a example of a couple of them uh and what they do is they create uh different assets using these Styles and then they leverage uh layers canvas tool to basically compost them into game uis and once they're done with that they can export that file as a PSD and then do future uh further refinements on a Photoshop as well right but yeah but beyond those customers there's more assets being game asset being created on layer but an interesting thing is that it's layers is just not used for creating game assets

uh and and and another use another another broader point is that like AI today can be used much more than just game Assets in general it can used for creating marketing content or even liveops content right one interesting detail note is that if you think about the industry today the cost of AI is going down the speed of AI is getting higher and the quality is getting better we're getting more modalities like video so I foresee that uh in within the next couple of years it's going to be cost efficient for us to create personalized content for players using AI to help Drive higher LTV right so today that might be not might might be cost prohibitive especially there like things like hyper casual or Super Hyper casual but eventually it's going to be like uh efficient so at layer these are some of the things that we're uh betting on uh in hopes for the future so the one thing is M uh multimodality so this this means that we can produce assets across 2D 3D

video and audio which we can already today but it's also about having the ability to go across these modalities being able to convert the 2D image into a 3D or a video or the 2D image into an audio clip for example but being able to go back from 2D 3D into an image and so forth so it's the ability to go seamlessly across these different modalities the second one is that we want to make it make it possible for game studios uh that have unique needs to automate their creative pipeline so this means being able to create kind of like a workflow engine uh to automate for example the generation of those banner ads and more you can do variation testing and so forth the third thing that we think is going to be very important is the use of generative AI in liveops so possibly creating personalized content for players based on their history or usage uh to drive up that LTV the fourth one is that as more assets get created and Enterprises adopt the AI tools and you have more users you need better tools to manage the content as well which is why Asset Management

asset management is going to be very important the fifth one is that uh we're also looking into how we could leverage extended reality uh as a way to interface with AI uh maybe VR for example right uh here on the left we have a quick like a video of like how that workflow engine builder might work uh and we're hoping to ship this soon overall whether it's layer regard some something else we think that the best AI tool is going to be be a seamless layer across all of these different modalities binding into a single tool for your users to use so in summary uh you should maximize your studi productivity with a single Enterprise ready tool that evolves with AI empowering your team to focus on uh game development uh instead of uh lengthy procurement processes and learning new tools and U yeah that's it all folks and uh be happy to take some questions uh from

[Applause] you great thank you right so do we have any questions from the uh from the audience we can get a mic to you oh you've got a mic down there did you want to did you want to ask a question no no I'm uh in case they have question oh okay great you got a mic good oh there's one here behind you there we go hi uh do you in your experience see these tools mostly being used by artists to speed up their workflow or non- artist to get stuff done so today uh we started off in 2D so it was specifically used by artist first uh but now with the introduction of 3D for example and once with with audio I we see that probably Indies and game developers could would benefit from from this very well much as well if that makes sense right right thank you do you have any oh there's one over that

side I'll I'll run oh I think someone's bringing a right to you there thank you hello uh thanks for sharing um just just for a background my previous uh job was I was head of strategy at Gina a global publisher and uh I used to work on sell uh publishing games or cell development projects and what I saw was that usually G developer they spent a lot of resources on initiating new projects but it took a lot of times and you know eventually didn't pass some some Metric test and he got killed uh but I just wonder that do you see these kind of tools would help uh for not just for existing projects but also for a new G development project as well in in which case that you can uh save a lot of time in terms of have a protype quick quicker and have a test in global scale test the graphics in within not uh two years but within several months or weeks and save a lot of resources uh I think that would have a huge impact uh for these uh bigger sale company so so just curious uh your uh

experience uh maybe not for now but you know what's the potential in the future that for these strategies that can make quicker decision and Save A Lot resources for this Cate yeah is that let me just summarize that question so I make make sure I got it correctly like the it's more of can this tool used be more in the conception stage to like speed up development so that the project doesn't K get killed before it actually gets delivered yeah or like if you have to kill it you can kill it earlier not wait until oh yeah yeah yeah no no ABS 100% 100% I I I obviously don't have time to go into all the things that we're thinking about doing uh but some of the things that we're planning on doing is like the ability to create mood boards and this type of stuff so that you can collaborate on the idea of the game very soon uh you don't need to come into layer and with existing assets for example there's already like existing styles that you can do to use to like create uh game assets for your game for example so you can get started very very fast without even having an artist in place so you could have that's why I said like for even for Developers for example like I'm not an artist but like I can come and draw a couple boxes and get some assets that you know I could

already use in a game right I'm not trying to make a triple A one right uh so I think overall uh we're already doing this today but I'm sure we again we'll do better better with this in the future uh but I see that like AI is going to allow people to ship games faster and with less resources and it's not just games like it's it's a more broader thing like across the industries that's why I think like AI is going to enable startups to really Thrive because you have way more productivity with a less amount of people um so so yeah and say yes I appreciate it yeah great thank you I've got one question at the back there I think we might be at time but so do you want to thank you very much hi I'm Primos uh we're 10 person Studio making mobile games we used your service and scenario and my qu so we we were doing a game like uh think of travel town or merge Mansion where you have you need 300 assets small size something perfect for an AI right and we made a couple ourselves like two sets like 30

assets uploaded them to the to these platforms and then see if we could get a new asset out of it a pack of assets out of it and we just didn't really the the style didn't get transferred the line thickness was not there the color schemes were all over the place and so on um so it took quite a lot of time to go over everything by an artist manually where do where do you see this going like are we actually going to be able to produce like a 100 Assets in a day with it took us now I know 10 assets a day manually it would be maybe five so it doubled the process how far do you see this going in the in the midterm uh I'll my my quick summary I'll let B answer this because she deals with aot a lot of these customers and these types of concerns specifically but in in short I would say like I was like we're moving fast and like enhancing the AI Generation by itself but beyond like what we're doing at at at our layer

level the base model layer is getting better and better as well so if I think about like where AI was like a year or two years ago like six months from now already everything is going to be higher quality and so forth but I leave it up to B because she deals a lot with a lot of these kind of questions from like customers hi uh Primos right yes nice to see you again uh we actually funny you mentioned we work with both Travel Town team and more Dimension um so what you're facing right now and we're working relentlessly on on optimizing that is the self serve on boarding uh both teams who found success uh and making those games were actually on boarded by us uh so as soon as you know we did a workshop and explain how these like differences between model models how to caption these even though we have default uh configurations in place they're not necessarily giving you the best results for like specific things you want to do that's why we interfere there and really help the customer so I think um you as a small Studio come in

and try didn't get optimal results this is a common uh situation we're facing however we exactly found what needs to be better and we're currently working on improvements uh so what are some of these things um initially of course types of images uh you upload and how AI perceives that art style and how that art style should be communicated and configured in the system so that um you can regenerate those assets um with like just a few words uh hopefully without any hustle uh but yeah it's not uh that straightforward at the moment yeah thank you for your answer we have to leave it there we've got our next speaker a round of applause please for L speaker y Ando [Applause]