AI-Powered Design: From Sketch to Stitch with VIZCOM | PI Apparel

Thanks for joining, guys. My name is Jordan. We came here from San Francisco. And yeah, I'm actually going to— before I get too deep into it, I'm going to start this presentation off with a quick video user segment thing we threw together. It's about 3 minutes long. So I'm just going to jump into that. Skip to this timestamp. There you go.

[MUSIC] I'm a curious person, and I always want to try out the new stuff. And AI was in the media, and it was heavily seen negative in the creative space. People would say it's not authentic, it's just copy-paste or copying an artist's style. I felt in the same way in the beginning, but I gave it a try because I wanted to know the enemy. I also wanted to, to give something back to my community, so I started learning AI. So I started using WizKomm. Actually, WizKomm was really the first AI tool I used. Very soon I realized it's not my enemy, actually it's my coworker. And it felt pretty naturally because the user interface is really user-friendly. And the learning curve was pretty easy. Design is all about communication. I never have the one-click solution while using AI, so I always have this kind of playing ping pong. When I use resource images, I always use different moods because then you achieve that kind of extra character which you didn't see before. This is something we traditionally don't do because in the beginning it's always like that gut feeling that there's something interesting in it. It's not perfect, but something attracts my feelings. And then you have to explore what is it. Is it the shape? Is it the material? Might be the functionality. Go to other directions where you didn't think in the beginning. So I don't have a linear workflow.

Awesome. Oh, there's more.

I'm a hybrid designer and also a hybrid human, and AI really boosted my hybrid being. I teach my son that it's great to be a hybrid by showcasing him our roots. So we travel to Indonesia nearly every year with him. I think in general that is the reason I try new things. I try to see new cities. I try to take new paths while running. And that's the reason I think I gave AI a chance. Yeah, being different is kind of a superpower.

Yeah. Awesome. So yeah, that's Dennis. He's currently at On Running and an advocate and early adopter of these things. So yeah, my name is Jordan, like we were saying earlier, co-founder of Viscom. And I'll let Al introduce himself. Introduce himself.

Hello, everyone. My name is Al Hanhauser. I'm head of sales here at Viscom.

Cool. So I'm just going to go right along here. We don't have much time. So kind of give you guys a little bit of background about the journey and how this all started. So my background is I went to a place called CCS to study car design in Michigan. So I have an art school background and spent a lot of time studying in Korea. And this actually all began in my living room. So I was working at NVIDIA. And that's the living room where Viscom originally started. My co-founder I've known since the 5th grade. And I was like, hey, man, we're at our houses anyways. Why not just leave what you're doing and do it here, but for something called Viscom? And he's like, OK, awesome. So I left NVIDIA, and now we're here. So it's been quite a journey, and it's pretty amazing to be talking to you guys here all the way in Italy now. So this is kind of just a high-level overview of what we're going to be talking about today, the vision of the company, a little bit about the team, the tool, and then kind of how we think about the future of fashion and design and what's next in our eyes. So starting out with our vision and how we actually approach these problems. So if I open up the question of, what do these three objects have in common, right? They're very different. They serve different purposes. And most of you probably know the answer to this already. But a lot of people don't realize is a lot of these iconic objects started out as drawings. And drawings is a very fundamental part of expressing your ideas, and all the way to the fact that a lot of the world-changing ideas started out as sketches, all the way down to the New York subway system. This was once a drawing before it was manifested into the subway system we know today. And like I was just saying, humans have always sketched. It's a fundamental part of how we communicate. Talking and text is just one part of that, but drawing and actually using our hands is super fundamental to how we actually communicate ideas. Ideas. And I recognize this, and I tried to make it a core part of how you interact with Viscom. And it's the only thing that's changing about that more so is where we actually utilize those drawings. It's more so the canvas in which we use those that's changing while the core principles stay the same. So this is super important to us. And it also kind of brings a few problems when you bring AI into the loop. With infinite creation becomes infinite possibilities, and you kind of almost run into this decision paralysis. And you can probably go on your phone right now on Instagram see an AI-rendered image, and you'll realize there's this lack of signature. You see this generated image, but you don't really get a sense of maybe who made it or who it's by. And there's kind of becoming this convergence similar to the post-Pinterest effect, I almost call it sometimes. So there's a lack of signature. So this is kind of where we kind of come in the mix and really trying to understand the problem more deeply, where we're trying to build these things for people like you, like literally for designers. And not for the crowd of everyone that maybe claims to be an AI artist post-AI. I think there's a very big difference there. So yeah, our mission is essentially to accelerate the process of visionary minds like yourselves in this room here. Super important for us. So moving right along, super high-level overview of what the workflow generally is. We originally found our footing in the industrial design process, essentially leveraging the drawing as the main communication tool that artists and designers are already good at. Then rendering those drawings immediately, and then creating 3D models from those renderings, all using AI in the process there, minus the initial sketch part, of course. And then there's kind of like a canvas aspect of that, which I'll go deeper into. But this kind of becomes the new converging and diverging workflow that a lot of people are adopting. And we're seeing this happen in many different domains. Now, a lot in the apparel and footwear space, like we showed in that video at the beginning. But it's super cool to see something like this being used used across many different domains, not just cars and physical objects, but backpacks and shoes and footwear and things like that and accessories. It's been super cool. So giving a little bit more context about us and the team, so we have a lot of designers actually on our team, a lot of people from KeyShot, nTopology, Gravity Sketch, from the CLO team as well. So I really am very conscious that the people we have on the team are also designers. And these are just some of the major challenges that we're dealing with in the industry. And Al will talk a little bit closer about this in a second. But you guys have probably heard a lot of these problems as far as, OK, well, who owns these ideas? It really lacks my personal personality. And I really feel like I don't have a sense of control of what I'm making here. So these are kind of the main challenges. And these are some of the core product pillars, I would say, that we truly define to kind of combat those challenges. One example is starting with the sketch. That already kind of debunks a lot of the initial authorship, for example. Something else I'll go deeper into is how Viscom learns your unique style is super important. And then also these abilities to control every detail. I think especially in fashion, that's extremely important. And I'll hand it over to Al to kind of talk more about the actual enterprise and company integration things here. Yeah. Cool.

So yeah, I get to talk about some of the security stuff. But so I was here for pre-revenue with Viscom. So I've been with the company for almost 2 years now. And one of the things that really separated us from the early days was how we kind of integrated with security from the very beginning. So being built from the ground up with like SOC 2 compliance, ISO 27001, like really boring things, but a lot of stuff that your IT legal teams are going to care about. So yeah, I think— early days, we really started to work closely with enterprise companies because of these reasons here. So SOC 2 compliance, as I mentioned, a lot of like GDPR controls and things like that. So we could work with global customers. We have a lot of boring admin controls. So, you know, you can see who's exporting images, who's creating these images, who's doing certain things in the product. We also, for all of our enterprise plans, we will never train the models on anything that you give us. So we would 100% protect your data. Anything that you export out of Viscom is 100% your copyright, your IP, and so on. So we have large enterprise customers today using us at scale across multiple teams across the globe. So if you go to the next one. So now we started in automotive. Jordan here, I think you'll get into your story, but he worked at Honda early days. So we had a foothold in the market with automotive. Then we kind of naturally went to footwear. Then from there, kind of do apparel. And now we have over 150 enterprise customers. And it's been about 18 months, not even, of actually going to market in the enterprise. So we've seen some great growth. And then here's some customer quotes. I think a couple of things to point out here is one of them is around breathing oxygen into your design process. So we're really about accelerating that process, not automating anything that you guys are doing as designers. We don't want to come in and automate anybody's jobs. We want to come in and basically superpower what you're already doing and make you much faster at what you already are good at today. So I'm not a designer. I can't really use Viscom all that well. But if you are a professional designer and you know what you're doing, Viscom is just going to make you so much better at what you're already good at. So yeah. And then this is the team. This was us last year?

I think so. Yeah. This was us at our offsite.

That's the whole company.

Yeah. That's basically the whole company. And we're now about 40. 40 people, so a little bit bigger than this when this photo was taken. And yeah, all based out— most of us are in San Francisco, some in Lyon and Hungary as well. So pretty diverse team. So now I'm going to go more into the actual tool and what it means to use Viscom in a lot of ways. And one of those things you guys have maybe seen already, but it's based around instant visualization, but also respecting your initial idea with the drawing. So the classic tech pack and then essentially just making that a little bit more photoreal. To kind of communicate your ideas at a higher fidelity. And I think the main thing here is respecting the intention of the designer, which is a really key aspect and sometimes hard to nail with these AI tools. Another aspect is actually then animating those drawings. So once you do have that render, adding a little bit of motion and life and almost kind of discovering new aspects of it that you might not have thought of otherwise. Is super cool. And we call these the happy accidents, you know, these things that you originally maybe didn't intend for but then inspire you to go down a path that you might not have thought of otherwise. It's kind of a natural progression of integrating these tools into your workflow. So this stuff is super cool. It still amazes me this is even possible considering what you would have had to done prior to make this happen. And the cool thing is this is all happening in one place. You never really have to, like, leave the platform or learn multiple things to achieve this. It's kind of all happening intuitively in a single space, which is the most exciting part. So rendering it and then being able to try that on to a model or a photo and then seeing that in motion is quite a magical experience. Here's just another example. So you have your Illustrator renderings or your sketch. You can then bring that into Viscom, very intuitive node-oriented type program where you then can be on this canvas, pick the ideas you think are the most relevant to you, And then once you do that, you can then actually generate a 3D model from that to see it from many different perspectives. So once again, this is an opportunity, I would say, to basically fail faster is kind of what I say.

Yeah, I just want to call out, this all starts with just that input in that sketch originally, right? So it's like a suite of products that's all in one consolidated tool with Viscom.

Things that are a lot familiar with— if you guys are used to Illustrator and things like that, there's a lot of familiarity here. And this is just kind of like another— quick demo of what that would look like. It's— you can do a lot of different variations, exploring different CMFs extremely fast. Here I was kind of exploring what a gradient jacket might look like and creating a 3D model from that render, and then using Viscom Visualization Engine to kind of understand what would this look like if it were a gradient rain jacket of sorts. And being able to see that from multiple angles extremely fast is, once again, a pretty cool magical experience here. So I'm just going to go to the next slide. And then, of course, being able to see that on a model, like I was just saying, to kind of understand what those proportions and things might look like is another key aspect of all of this. So just to be conscious of the time, I'm going to keep moving along here. And another thing we really focus on is how do we actually enhance some of the details that come from these images? Because when you're generating things really fast, you inherently give up the quality. We're always looking and exploring ways of like, how do we enhance some of these details that are more familiar with something that a CMF designer might expect or materials you might have in-house to bring in? And so really defining those details is super important. So I'm going to go to the next slide. And this is another key aspect. So you guys probably have heard of prompting or even like a prompt engineer and things like that. But we really try to lean into the visual language of design. Making the world your swatch is kind of what we say internally. And it's this ability to use images to kind of communicate your intentions. So these are just some rough examples. But essentially, what would it look like if a leaf were— you wouldn't literally make the leaf the soul, but it's a form of inspiration. And you're kind of using images as a way to communicate those ideas rather than just words because it's really hard to communicate only using text. So this is super big for us, even all the way down to just— kind of encapsulating the mood and style of a particular concept. So here, we had this kind of fungi-esque project where we were trying to see, huh, what would it look like if fungi and bacteria and things were a part of the visual language of the shoe collection? And it's just a really interesting way to explore things that might have taken much longer to understand what this possibly could look like. So this was actually by Dennis on the On Running team. That you guys saw earlier, essentially using some of those style transfer techniques that Viscom offers to interpret the images as footwear concepts. So super, super cool exploration here. And another really important aspect of that as far as using visual language as a part of the design process is adding context to some of your designs. So let's say you had an apparel object or a footwear or a shoe. You can then use an image to kind of explain what the context is of that without changing the design. You can almost think of this as smart HDRIs in a way for anyone who's used your typical render engines and things like that. This is basically a generative way of applying context to your designs without actually changing the original design because it's like lighting and things like that can fundamentally change how you perceive a certain design and things like that. So we wanted to make sure to have these kind of capabilities as well. And this is kind of a really cool one. Similar thing of leveraging a 3D model you might have from your favorite tool and essentially using Viscom as the visualization engine all on the singular canvas. So a lot of this, once again, is just allowing you to fail much faster in your process, which normally might not have happened until weeks or a few months later down the road when you're developing some of these things. So this is just a rough example of how we're seeing people use the tool in the apparel context, which is very fascinating. And another thing you can do is bring in a bunch of different things and then just ask it, OK, what if I were to mix these two things together and just understand what this would look like in a mix-mash collage situation? So we mix those— let's see here— the moss with the jacket and this material and that squiggly thing. And then you can understand, oh, wow, maybe that could be a cool embossing. Some of them are pretty abstract sometimes. But it's just a fascinating way to explore things that are a little bit novel. Because with these AI things that are trained on massive, massive data sets, it's hard to extract novelty from them. So this is our way of trying to inject novelty into the software in a lot of ways. Cool. So then this is kind of this interesting thing. So collaboration is a huge part of Viscom. And we're actually going to switch to another window here. And I think some of my colleagues might actually be in here as well. So this is kind of what a Viscom file might look like. We have Nico. I don't know if Nico's in here. Is he in here? [SPEAKING RUSSIAN] Oh, where's Nico? There he is. What's up, Nico? And so these guys are actually working in here right now. So you can kind of have your teammates in here if one of you guys want to play a video or something to show them. But yeah, anyways, this is essentially the workspace. Space in which you can kind of work with your team and explore different designs together and easily share the files using links and things like that. So it's kind of this like Miro meets like Photoshop meets kind of like a lot of these generative tools. And we just really try to make it a cohesive experience here. So super cool. And yeah, I just wanted to kind of show this off to you guys real quick so you can see what it would look like actually in the application. So It's a really cool experience.

And just to— sorry, real quick.

Yeah, go for it.

Yeah, just to add to that. So this is built in. This is called Workbench. So it's kind of like a Miro board that's actually in Viscom. So a lot of our teams that are customers that have global teams or dispersed teams, they'll all put their ideas here. It's a very easy way to collaborate globally, cross-functionally, marketing, whatever. They can come in here and see what's going on all in one spot. So it's pretty cool to collaborate at this scale.

Yeah. So we can— so yeah, these are kind of like the videos that would play. But OK, there we go. So and then, yeah, it's really cool. So yeah, you can see kind of like how these videos are coming from this one node, trying a bunch of different poses and different things. But yeah, these all derived from an original design that someone had. So it's just cool to kind of see it all happening in one place. Awesome. So yeah, thanks, guys. So I'm just going to move back to the slides. So that's a really cool aspect of all of this. And then go to the next slide really fast. Oh, that's me. Oh, yeah, OK. So oh, yeah. So palettes. This is kind of our way of allowing designers to define and own their own unique styles. So what does that mean? I don't know if you guys know who this is. This is Bob Ross. Probably has the most unique style, a lot of style. And you know, it's uniquely his in a lot of ways. And how can everyone else be their own Bob Ross and have their own style, right? So this is kind of like an abstraction of how that works in the context of Viscom. Your data, your own data, you can then sort it, arrange it, and then kind of create these unique bespoke models that then understand your perspective on a particular subject matter. This is kind of a quick example of what that would look like in practice. So we have this thing called palettes. In the workbench that I just showed you. You can essentially upload a pool of images that captures an aesthetic or a design language. And the next time you render something using the software, it will kind of follow that unique language that you've defined. And this is basically your way of adding your own novelty into the software that might not have been on the internet or might be uniquely to your particular company brand. And these are kind of the 3 areas of focus that we found to be the most useful for something like this? So your own style, the brand language, and maybe a visual style of sorts.

Yeah, and we have some referenceable customers like New Balance or something like that. They've actually trained a palette that we helped them curate the data for. Then they'll actually have performance palette, running palette, maybe just leisurewear. And then all those designers are using old styles that they've had in the past, new sketches, maybe ideas. To influence that brand language. And then they use that for the next way that they develop the next line or whatever it is. So it can be done at a global scale, or it can be done as a personal aesthetic as well, just for a little project that a specific designer has.

Totally. Yeah, that's a good shout. And here's kind of like an example of that would look like in the Dyson context. Now, Dyson probably wouldn't make shoes to vacuum your house. But it's just kind of like to show you like, oh, whoa. We can really kind of stretch the context of what could be done here. Because you're not going to have time to explore what a Dyson lamp or a tea kettle would look like. But it would just be cool, right? So it's just another way to explore the portfolio of design languages that you guys maybe have historically defined over the years to land on something you might not have otherwise. And that's kind of the whole point of a lot of this stuff is to be an explorer. And it's interesting. We've actually started to see— emergent job titles derived from this, people calling themselves a tastemaker or an associate data set curator. You're almost becoming a professional mood boarder in some ways. And this becomes a big part of the creative process, knowing what images are relevant and look good to create palettes then becomes a new practice to be good at. So this is a LEGO example, a bunch of LEGO houses. And then you can turn your drawing of a LEGO or a house into a LEGO house. So this is a really exciting capability that I'm super happy about. Yeah, just going to keep moving forward. And then, yeah, we're almost coming to the end here. And then, yeah, a lot of this stuff is built just for you. A lot of the— very much thinking of designers from the beginning. And this is kind of how we see the future tools. It's like you and your AI, Viscom more so being a colleague as it's currently in this intern state where you're having to tell it a lot of things. I love interns, by the way. Nothing against interns. But we very much see this turning into a collaborative experience with— as if it's almost like this digital colleague in a lot of ways. So that's how we see the future here in some ways. So now we're going into this what's next conceptually. How do we think about viscom in the future or just AI and design in the future and some things we're thinking about? And there's this idea of manufacturable intelligence, right? You have your ideas and then they actually materialize them. There needs to be this kind of intermediate interface when you're kind of thinking about generative design being this thing that has to actually enter the real world at some point. And the way we see that— and this is just one idea— is Viscom being this kind of translation layer between the bytes and the atoms. A lot of people and things we show today, we're just kind of thinking about the world of bits, just things that stay on the screen. But what does it mean to actually materialize a lot of that? And this is the concept maybe where you can add a cart, an object you designed in VisCom. And then maybe there's this rise of the personalized manufacturing. And it actually just comes to you. And it's interesting that you'll actually start to maybe see the world around us reflect the tools. Cars in the '80s and '70s were very boxy, not because people couldn't draw curves, but the tools were more so the limiting factor of what you could actually make in the real world. So maybe we'll derive to this world where the objects around us are a reflection of the tools that we're using. And it's enabling this new era of generative products. And I always tell the team, oh, man, success is when I can go to Foot Locker or IKEA and I look around and I could easily tell this was— Viscom was used in the process there. So it's a really exciting time. And this is just a conceptual idea I always share. I'm like, oh, man, what if you could go to the Viscom store store. There's a bunch of these. And these are these new generative product lines and things like that. So very much a just high-level concept. I used to love Build-A-Bear as a kid. I don't know if you guys know what that is. It's a place you go to make a bear. But what if there's a build anything? I can just go there and build stuff. So really excited about this kind of future here. And yeah, so I think custom and accessible doesn't need to be this mutually exclusive thing. And I think some of these new technologies around AI and accelerated design can really enable a future like this where our kids— like, I don't have any kids, but maybe some of you guys have kids. And they could be the designers of the future at their current state. And yeah, that's kind of the thing— enabling physical imagination. Thanks, guys. Yeah. [APPLAUSE] Cool. We're going to go over to the— Slido. Answer some questions here. Awesome, awesome. That was great. That was really cool.

See if there's any questions.

Oh yeah, we still have like— oh man, I was talking really fast. We still have 7 minutes left.

It's good. We needed the time for Q&A. But no questions.

Yeah. All good. Submit your questions. We have 7 minutes. Yeah. Yeah, no worries. No rush. This is where we need Gary Gay. Is Gary Gay back there? Oh, OK, cool. I have a question.

Hardware?

Yeah, yeah, of course.

First of all, congrats.

Blow it away. Oh, thanks.

Could this be applied to other industries as well?

For example, set design, film, pictures? Totally, totally. So yeah, we've seen this being used in many different contexts. So set design, some of those environmental context things I was showing you guys where you can use an image. To kind of set something in a physical space. We've seen some game design studios use it in that way where they're kind of creating these fantasy worlds. So you can just already start to imagine how that would be used in a set design or a more physical space creation. So we got some other questions here. How complicated in terms of paneling, hardware, et cetera, can Viscom support? OK, so in terms of paneling hardware, so yeah, we definitely have seen a lot of the hardware stuff being used— Viscom being used for. That was kind of our original bread and butter, like the industrial design space. So tons of use cases there. We have a lot of the traditional Photoshop selection tool type things to kind of be very specific about where you would maybe want that to happen. So there's a lot of use cases there and a lot of tutorials online about how to do exactly that. So very much an applicable use case. Can we get patterns from generated 3D apparels? So yes, all of the 3D models you saw here that we were generating come with these texture maps. So you can actually export a GLB file— or I think it might be FBX that has that— that comes with the actual texture itself. And then, yeah, you can then use that in any of your own software and maybe bring it into Substance Painter to edit that texture and see what it would look like on the— the generated 3D model. So a very valid use case there. Can the products rendered be exported as a tech pack and manufactured? Oh, man. He's got the viscom crystal ball. So this is something we're definitely exploring here. I always say, yes, it would be great if you could export these as a tech pack as it belongs. But currently, you can't do that. But it is something we're definitely doing a lot of research on and trying to understand the complications of making it real. But currently you cannot do that. How fast can you implement new features? Show text— how fast can you implement new features? Fast. Show text in images. Example, show text. Ah, okay. I would say, I mean, we can— I would say we implement pretty fast. I guess it depends on what we consider fast, but showing text— text in generated images is notoriously a difficult thing to do. Not something we're super great at right now, but we understand it's a problem we need to improve on, but Yeah, we're constantly uploading this, like updating the software, but almost trying to like almost not do it too fast sometimes. Because as soon as you get used to something and you change it, it's like, aw man, that's annoying. So we're always trying to balance that there. Could you improve the brushes even more so we have more options while drawing? Adding more pressure sensitivity? Yes, yes, yes. This is huge, especially on the iPad right now. Since we don't have like a native iPad app, it's like this little bit of a clunky browser experience. It's still very doable and it works, but I totally agree here. To improve some of the brushes here is kind of one of our next big things on the roadmap as we're building out the team. So it's a good shout. Is there a way to keep branding consistency, for example, applying a specific badge or label and having it being accurate and consistent in every iteration? This is a great, great question. There is a way to do this. Now, it's not super intuitive, and that's one thing we're always trying to work on. It's like, how do you balance the— introducing these very specific features that maybe are only relevant to one or two— like one particular thing, but also having it be applicable to everyone. So maybe there's this thing where the badge is also something that works in car design. Or the way you would currently kind of do this is using kind of like selection tools, cutting and pasting a badge onto where you want it to be, and then using Refine to kind of like actually make it part of the image you're trying to add it to. So that's one way to do it. It's a little bit of a craft. You kind of have to do currently. But there isn't kind of this automated way to do it. And then palettes is maybe another element of being able to keep a little bit of stylistic consistency when you're creating designs using Viscom. Pricing model? What is your pricing model? Is it per subscription on my team or— oh, so yeah, it's just like a seat-based thing. So something you would see from Autodesk or I think CLO does this as well. So very much a traditional buy seat. We don't have any concept of credits and things just yet. We're just now testing that with animation that's currently in beta. So to make those videos, that's where you'll have maybe a limit in credits and such. But everything else you guys saw today is very much just at a flat kind of price of a per seat basis. Do you focus mainly on designers or also on the marketing team? We very much still focus on designers, but sometimes we find that inherently doing that naturally solves some of the problems that the marketing team might also have. So for example, setting some of those designs in context, that's something that the marketing team might find useful. But marketers aren't necessarily an ideal customer profile that we're laser-focused on. We're very much thinking of designers here.

Yeah, and we do have certain— footwear and automotive companies that have started to adopt us in marketing, but it's at a very light scale right now. But it's something that we are interested in looking at. But right now, it's not heavily used by marketing.

Totally. Totally. How do you prevent a user from creating something that looks like a brand? They do not run from passing it off as if it is theirs, I'm guessing. So yeah, we have a lot of filtration. So I think, yeah, a lot of things— we have a huge list. We're constantly updating it, and we actually allow enterprises to create their own whitelist of words they don't want to be used in the program. So you can't really query a lot of famous artists' names and brands at times, just to kind of prevent some of those potential IP infringement things. So we're very, very careful about that. So that's kind of a lot of the things we're doing on the fly to make sure that people don't create images from other brands and things that might influence their designs and allow them to claim it as their own. So there's a lot of things that go into that. Yep. Can we leverage our creative 3D assets with the AI solution? And will the info be protected? So yes. Yes. So everything you upload to Vizcom is encrypted. And we do not get any access to this. It very much stays within your workspace and instance of Vizcom. We don't train the model on the fly on your data or anything like that. So anything you upload, like this 3D asset question, will very much stay within the garden or the— what is it? What do they call it? The wall garden. Yeah, yeah, yeah. Of your own facility.

And it doesn't get trained. The model does not get trained on anything that you give us either.

Yep. Oh, sorry. Can I ask one last question? Whoever had that, I'm so sorry, but we're a little bit over time here. I know we got a break to launch here, but yeah, thanks for coming, guys. Really appreciate it. Thank you. Yeah.