AI meets 3D in Fashion | Style3D | PI Apparel

Good afternoon, everyone. I am Alfie from Style3D. As some of you may already know our company, we are a technology company founded in China, in Hangzhou. So actually, it's a high-tech company in China. Before I start, I want to say thanks to Stanley because LevelStyle is also a very valuable customer. We worked for them for more than 3 years. They give us a lot of valuable input for our product development in last few years. Also, to the last question you asked about Xiying, maybe I could address because we also worked with Xiying for more than 4 years. So basically, right now they have more than 300 users. Running on our software. So regarding the automation digitization, in case you are interested about Xiying, maybe I could help you to figure out a little bit. So today my topic is about AI and the 3D. So in nowadays, maybe not just in fashion, also such as in a computing graphic industry, such as filming industry, gaming industry, you will also see some AI background technology person, they are learning about 3D, and some 3D background person, they are learning about AI. And we believe there is some reason behind. And for us, we're going to share a little bit progress as well as some practice for us how to apply AR and the 3D in fashion. So before start, I want to share a little bit update about our solution in AI. So in the video, you may already see that for 3D and AI, they actually could be the mutual in, uh, umbrella for each other. So maybe, uh, such as for AI, you will see there is a lot of problem. No matter you are do some development based on Stable Diffusion or maybe you are using Midjourney for inspiration, you will see there is some issues for AI because right now the most common application for AI is to create a draft, right? No matter you want to create a draft for design or maybe you want to create a draft for sales marketing material, or even what you want to do, some interaction with your consumers. So basically, it outputs some— we call the regular content. So based on that content, you need to do a lot of edit and narrow down. So AI right now, we think it is still not stable enough. And also, it could— how to say— if you generate some proposal based on AI, it is very difficult to fulfill. So which means you are build something based on your inspiration, but in the last minute, you cannot fulfill in your production, in your real product. And for 3D, there is also some complaint, right? When I talk to some customer, my customer said, can 3D have a very good communication with the PIM? Can 3D have a very good communication between CAD system, right? You do a little— you do a lot of adjustment in 3D, and finally, you will see it cannot carry back to the CAD system. So in one hand, you increase your efficiency, productivity. In another hand, in the 2D, in the pattern makers, actually, they are losing their productivity. So based on that, we think 3D actually is good, but there is some place to improve. Such, we think 3D should be much, much lower barrier for the new user to enter this technology. Also, we believe that 3D, maybe it is not realistic enough in case you want to do the e-commerce. I still remember 3 years ago when we started the project with Shiying, we actually made a 3-month POC because Shiying said, in case I want to use 3D, I want to ensure it could be do e-commerce. So otherwise, in case it can only be landed in product development, so ROI for them, it is not enough, right? But in the last minute, we think it's not good enough because if you calculate the productivity, you calculate about the rendering quality, it is not good enough for e-commerce. So actually, today, we think, start from 2 years ago, Based on AI, AI actually helps you to enhance your rendering quality. Also, it gives you an opportunity how you're going to mix your virtual garment together with digital human, as well as in case you want to have a real face for your model, or even you want to create some realistic background for you to do the e-commerce. And for AI, because as I mentioned, it cannot be controlled. But 3D, actually, it's a very good technology for you to do the structuralization. So it actually enhance both side. So basically, as I mentioned, our insight is to use geography engines to— as a 3D bridge between 2D pattern and the 2D image. AIGC has already improved its capability in 2D world. So our path, you will see, we think in case you want to 3D data to train your model, then you generate a 3D image. It is— it is actually cannot be do in the real world. So our path actually is based on 2D. Then you could based on 2D to create some 3D object. And based on that structured items, you could do whatever application in the last application scenario. And in the last minute, in case you want to enhance the quality of 3D, 2D, as well as for AI could help you do that as well. So in fashion world, we believe in case you integrate both technology, it could not just give you the creativity based on AI, but it also gives you the opportunity to produce a real garment, a real footwear in case you want to fulfill your customer needs. Also, when we— before that, we are talking about we want to do a lot of cost saving based on virtual digitization. But based on 3D, as I mentioned, in case you have pure 3D application in your application, actually it cannot save your cost because you need to have talented people to create 3D asset. But based on AI, it gives you a real opportunity to speed your product to the market. When I talk to some brand, she said, you are keeping telling me that you want me to save the cost for the product development. But for those big brand, this is not their first priority. She said, what I want to do is that I want to ensure the speed as well as agility for my product development, then I have sufficient time to deal with my bulk order, to deal with my supply chain, so that I could have my savings from supply chain, as well as, you know, I have more speed, more room, more buffer for me to deal with my sales marketing. That is your goal to use the digitization. So one last result we think is that, as Danny just mentioned, is because for our industry, the biggest problem is stock, right? You need always to do a lot of investment before you want to sell to your consumer. And right now, based on technology, to give you the opportunity to do some test, or maybe you could even more closer to do a sale before you have some real heavy investment. Afterwards. So my two parts, one part is very quick introduction about our company, another part is about the solution. Okay, so, so compared to current digitization suppliers, I think we are a much younger company. We did our market launch during pandemic 4 years ago. You know, at that time, I still remember in China, the government, they have the control the status of the pandemic or the COVID issue. And in, uh, in June of 2020, we did our product launch. Until now, we have worked with more than 2,000 customers already, 60% in China, 40% outside China. So basically, one of the deep differentiation for us to compare to other competitors is that we are GPU-based simulation and the rendering. So which means we give you a much, much more faster speed as well as much, much more stable performance in case you want to deal with some difficult task. And on the other hand, 2 years ago, we acquired Assist. And some of you know they are one of the key CAD systems in Germany. So before that, we actually heard the voice from our customer. They said, we want to ensure the integration between CAD system and the 3D system is real. It's not a fake story. So that's why we acquired Assist, a very famous CAD system for Dutch place. And start from then, we started to build the integration between 2D and 3D. And 2 years ago, we think it's a very good opportunity and also very good maturity for AI technology. And since then, we have started to build the integration between 3D and AI. And until now, we have already like 500 staff in our company and 400 people in China, another 100 outside of China, mainly in US and Europe. So as I mentioned, we are quite young. So we are still being fundraised from the top private equity. And most are very famous private equity in Asia, such as Lenovo, Tomasiki and Xiaomi and Baidu. And last year we have do the fundraising for around $100 million. So that give us more— how do you say— more investment to do our product development as well as our market expansion. So we want to build a SaaS platform basically could help you to do the smarter design as well as collaboration between technology person and the designers, as well as for the operation person. And finally, give you a— how to say— a system to do the storage for your asset, as well as to collaborate for your asset. So right now, our key segment including 3 parts. First part is about the tools. So it's about the asset creation. Creation tool, as well as asset editor tools. And all of them, they are AI-powered. And based on that, in case you have the asset, you need to do the storage, you need to display, you need to do the collaborate, also you need to do the light design. So based on that, we also create a cloud-based platform, help you to manage your asset. And finally, we have the asset we want to apply in different application. In Dawn Days, we think there are two ways. One way is 2D. Another way is we call the 4D dimension, such as animation. For 2D, it means after you have the 3D asset, you could have the rendering. So you have the 2D picture so that you could use your 2D picture for your sales marketing, no matter it is for B2B wholesale or maybe it is for B2C application. And another direction we call the 4D animation. So in case you want to create a computing graphic animation video, or in case you want to do a B2C try-on, right? You want to do some tailor-made system. So based on Unreal Engine as well as Unity, we create the plugin so that help you in case you have the 3D asset, you could do the edit and apply afterwards. So basically, what we want to do is ensure that the connection between consuming what as well as the industry world for the vertical fashion market. This is our key product matrix. You will see there are 4 layers. The first layer is about the digital creation, content creation tools. The second layer is about empowerment enhancement tools based on your asset. The third layer is about the collaboration platform. Final Asia is about application, such as a supply chain. You want to have some, uh, VR showroom, you want to have some digital chain book, uh, because you think you need to do all the collaboration online. So we have a lot of application to be plugged in, uh, in our, uh, cloud-based platform. So, uh, we are a technology company, so basically I rely on such as computing graphic technology and the AI technology, such as in these photos. You will see we have very good ray tracing system. So in case you want to display a very good result for your rendering quality, so we give you another way rather than V-Ray to do the rendering, as well as to better and faster to display your asset. And also, as I mentioned, because we are GPU-based simulation, So in case you want to deal with this kind of multi-layer complex item, we could give you a faster as well as more accurate performance. Even you deal with multi, multi-project, it will give you a very stable result as well as performance. And in case you want to have all the 3D assets, 3D garment already, and you want to ensure it could be applied in the ecommerce. And I could guarantee to you, in this year, all the big leading brands in China, they will have certain amount of their asset to do the ecommerce. The name I just mentioned, they have set up a very aggressive target in this year to ensure certain percent of their product could be used, 3D technology to use, to be used on the ecommerce. So for these 3 pictures, actually, you could see the first picture is produced, is created by 3D software. So basically it is not good enough, no matter for consumers or maybe for designers, right? So after your AI, you could have like the third picture. So you have everything, you have the real model pose as well as the background. So that gives you, you know, much, much more closer to do e-commerce platform. And this is about one AI and the 3D application. So I will save time to give you a quick look for the video. So this is our iCreate tools. Our AI tools help you to generate the print as well as the fabric. After you have the fabric, you integrate the property as a visual effect. And you could apply in the 3D garment so that you have a dress. And once you have dress, you could tailor-made the pose. And in case you use AI for 2D, you cannot tailor-made the pose. And only you could have the picture in the front angle. You cannot have the angle for other pose. And then you could do the e-commerce. And based on our integration between digital printer And the 3D, you could very, very quickly to produce just for fashion on demand. And basically, there are 3 parts. One is about how you're going to digitize your material. And after you digitize your material, how you're going to create a 3D asset. And once you have the asset, how you're going to apply them. This is also another situation I— foresee in the market. When you talk to a lot of customers, they said, yes, we have a lot of 3D assets already, but I cannot apply them into the real application. Because for designer, single item is not enough for them to make decisions, right? They want to create a look. They want to create a display. They want to create a virtual store. So in case you only have a 3D item, Sorry, it is not good enough. So that's why we also create a lot of tools to ensure you could do the mix match. You could create a collection. You could create a display based on your real needs. So this is our key design for how we're going to apply technology in the application. You know, such as I mentioned, such as for Midjourney, right now it is only for inspiration, for brainstorm. But in case you want to do the edit based on the draft, you want to narrow down your design proposal, also if you want to do some blending, sorry, Midjourney cannot help you to do that. So for our concept is that we do not want to fight against Midjourney. Midjourney is very good for inspiration. But once you have the draft, how are you going to deal with them? So that's why we create iCreate tool to help you to do the second part, how you're going to edit your original items, no matter it's 2D and the 3D. And once you have the design proposal and you want to make it into a pattern, so actually, we are also working on how you're going to generate the patterns. Based on 2D styles. So in future, we believe for pattern makers, they are no longer pattern makers. They actually— their key job is to use the tool to edit the patterns, not to create patterns. Such as for the designers, they use AI tools to create their design proposal, right? This is the same methodology. And also, after you have the patterns, Yeah, you could use AI to do the automation to create a 3D asset for you to do the e-commerce or maybe some— any other application scenario you want to use the digital asset. So this is a function. I will not introduce them by detail. So this is our application detail in different application scenarios. So we believe— No matter such as a technology company as us, or maybe such as a brand, you know, a lot of China brands, they also have the IT team to working on the AI application by themselves. No matter based on Midjourney, ChatGPT, or maybe some, you know, the local, the local AI big models, they are working on that. So we are also trying to find where is the best application scenario for AI to link to 3D. So basically, our value proposition is for 3 parts. One is that for the design and planning phase, how you're going to make it more agile, more productive-culottes, more productive-culottes, so that it helps you to increase efficiency. And secondly is how you're going to produce, right? In case you could already create virtual asset, we think the value is also not big enough. For brand as well. And finally, it's in your operation workflow, how you could be more faster and flexible. And so we are targeting to building a digital engine to drive the future of fashion. So what we want to do is to provide a best tool in the fashion world to help you to ensure that the digital asset could be really applied as well as to be managed. So some use case to a quick look. And this is how you're going to— in our software, how you're going to generate a 3D image for sales and marketing. So you will see right now, no matter the detail for the clothes or maybe such as for skin, for the face, it could be generated very easily, not use Photoshop to do a lot of edit afterwards. And this is about how you're going to do the inspiration as well as to the edit. So we actually, we have 3 functions for our Create, iCreate tools. OK. And this is another tool for us to engage designers as well as the merchandiser planning. And another one for how to do your— do the posing. And so rather than 2D AI, you could also have multiple angle for your model and the clothes as well as to mix with different background. And one more is about our MixMatch. So it's a real-time rendering. So you have a very high quality. Rendering as well as to mix, match different items based on 3D. So it is actually, it's a, you know, a plugin together with Unreal Engine. Let's go! And we also have a demo version upstairs, so if you are interested, welcome you to inquire with us. So, okay. This is basically my sharing today. So in case you have some question, welcome to chat with me. Thank you. Thank you. [SPEAKING CHINESE] Ah, okay. Okay. [Speaker:JIAOLONG_WANG] So for the mechanical, for the fabric, currently our solution is quite similar as what you could see from the market. So basically— I would say there is no rocket technology for these fabric solutions. The second question is much about AI-generated videos. So you have already seen Sora, right? But you could also see there is a lot of mistake in Sora. So basically, we think if you want to create a runway animation for AI, it is still too early. So basically, before that, I still remember some brands, they want to create the animation, such as in case you do the boxing. So you will have different fitting situation for different garment. And we do not think no matter for computing graphic as well as for AI could help you to fix that. So at that time, I remember the solution is that you create some continual pictures. So you combine those pictures to give you understanding about the fitting performance when you do the way of unboxing. So right now, in case you want to do the runway AI, I would see it could only generate some basic quality. Stuff such as Sora, such as Pika. 3D designer and the marketing team have not worked together in the past. How can this— OK. So for this question, my answer is in two perspectives. One perspective is about technology. So we think the final deliverable for our technology, or maybe for other technology, is a picture. As I mentioned, there are two directions for the technology. One is 2D. Another direction is 4D, right, the 4D animation. So in case you want to do the 2D, the final deliverable is a picture. No matter this picture is for B2C e-commerce, or maybe this picture is for wholesale, or maybe this picture is for such as for the ODM to provide to the designer buyers in the brand. We think the deliverable is a picture. So in case this picture, the quality is good enough, as well as the efficiency is good enough, I think no matter for sales and marketing, they will accept, and the designers, they will accept. Current situation is that once the designer sees a fake face, the CG face, or the digital model, they dislike it, and they will not give any further comment based on 3D object. So for the technology perspective, we think right now you could already deliver a very good quality picture, and the efficiency is also good enough, and your stakeholder from each department, they will accept. And another perspective, maybe it is much about organization. Such as the question you just asked to Stanley, what is the differentiation for SHEIN? We worked with SHEIN for more than 4 years. We also work with a lot of traditional brands, such as Urban Revival, Anta Sports, Li Ning, Peacebird, JNBY. And I would say SHEIN is the most, most fast company. So that is how they run their project. So that is how we are running our project when we deal with China local brand, because they make really fast decision. And when they want to make decision, they have to balance the good or maybe the speed. They always choose speed. So they think, in case we are fast enough, we could conquer a lot of problem. And this is also my mindset. To development our company as well. Except for create 3D soft item, is the software also able to create a 3D hard item? Actually, it can, but I think it is not effective enough. So for our software, also such as for Crowe Brosware, they are more common to create like the garment, backpack, handbag. And in case you want to create footwear, you have other software, such as Rhino, 3ds Max. They are quite mature for the hardline product creation. So you will see also, like, the performance brands right now, they have the trend to integrate the hardline and the softline. So we think in the asset management angles, it can be done. So which means you could have one platform to storage, to display, as well as to collaborate those digital items, no matter from hardline and the softline. But in terms of the tool, how you create a digital asset, we think still we are more good at softline, and Rhino So those kind of software, they are better for hard line. The same, similar. AI, AI, because AI right now, it is still not easy to control. So we do not think you could have a common standard to evaluate the AI asset from different company. So right now, as I mentioned, the key application for AI is to provide your inspiration as well as draft. So it cannot be the standard items or maybe it's a narrow down item for the real applications. Is there a way to utilize existing 3D Can we layer to AI model? I think AI is not for you to edit the item. AI basically give you the 3 capability. One is the enhancement, so which means the quality for the fabric. So you could enhance the quality detail for your garment, for your fabric. Another usage is to do the automation. So you have a lot of different function in your software. So if you want to speed up the efficiency, you could use AI. And the third application is how you're going to use AI to mix with different items, such as I mentioned, a 3D garment, how you're going to blend it with a digital human. As well as a digital background. So AI actually is not OK for you to add some layers or maybe to edit some layers. It is not OK to do it right now. Which department in Shein use Style3D? What is the main reason? Do they satisfied? For Shein technical department. And all the pattern makers are using 3D software, all the pattern makers. And I think for them, all the reason to drive them to use automation is speed. So speed is a consideration for them for all the supporting behind the CADization. And do they satisfied with the result? OK. You know, see, actually, they subscribe every year. And they also buy more lessons every year. So I will see they are quite satisfied with our product as well as our service. So except for create 3D soft item— OK, OK. Is the software also able to create 3D hard item? I already addressed it. Yeah, okay, I think I already answered all this question. Okay, thank you. Thank you.