AI footwear design in FLORA, from colorway automation to tech packs

Lightly edited from the recording for readability. Wording is the speakers’ own.

My name is Nico. I'm the founder of Footwearology is an innovation platform for the footwear industry, for footwear professionals. So we focus on people that want to improve the way we make and design and manufacture footwear. One of the things that we do is training, online training, for people really from every corner from the globe. Anybody who works in footwear, from designers, to developers, manufacturers, brand management, a lot of different people. One of the things we do is AI. And already for quite some time, I think almost two years now, more than a year, I've been using FLORA.

So what I'm going to do today is to try and serve two groups, both people who are very new to AI, to show you what you can do, as well as people that already know AI and show them some much more advanced automated workflows. And we're going to decide the webinar in two parts. The first part is going to be what most people ask for. How can I speed up my workflow that I do on a daily basis? And the other part is the part that always excites me a little bit more. That is, what can I do with AI that I could never do before? That is simply impossible to do with AI.

Any new technology helps us do what we do now. But also brings new opportunities to the table of what we could not do before. So I wanted to approach it from both sides. So I have prepared a full workflow or two full sequences, actually, of things that we can do.

So I'm going to start with a workflow where we know that AI is going to take over a lot of the work. So for designers that have to do AI augmented flow designer, but we also have, of course, for people in leadership who have to think about how do I apply AI, what are our policies? And we just ran one edition of that one. And the first result where we saw that AI was really taking over full workflows is coloring.

So if you have to recolor existing styles, which is also not necessarily the most exciting job for designer, but it's something you have to do a lot. So I had decided to show you how to do that. Well, because let me show you this. So what I started with is just to take an existing shoe.

This is a salon sneaker that you might know. And I intentionally took this one because it's relatively complex in terms of how the colors are applied and the materials as well. So here you have this big TPU overlay in two colors. This is probably a screen print on top of this. These are not two different parts, but then there's another upper underneath. There seems to be another material underneath here. Of course, we have the sole layers of foam, a whole TPU frame out.

So we have them here. It seems like leather trimming going up. Then we have color accents for the eyelids and for the eyelids. It stays here.

So there's a lot in there. Now, usually when, of course, we start with the new season or new collection, we have a color palette. This one I quickly generated also in AI. It's not the prettiest one. Don't shoot me for that. I did this intentionally because it's very contrasting colors so that you can see what's happening in terms of a process.

But what I've done is that what you usually do making color palettes, we have a base section, which would be the neutral colors for basic color combinations. We have some. These are colors that you can have in your collection all the time. We have some fashion colors are brighter, might be more seasonal.

And then we have some pop colors or accent colors, however you want to call this for little accents. A lot of times I work with brands and also my own experience when I was still a trend forecaster, this is how I would have to make my palettes to apply. Now, what I see a lot when people are starting with AI, they're simply thinking, OK, well, I want to change the color of that shoe. What do I do in this case? Let me just disconnect this.

So say what I did to get here is to simply say, open a double click anywhere in the galleons. Image and then I could connect with this plus this one here to image. What I did here already is to connect the issue because I want to recolor the shoe and I also want to use my color palette. So this is what I had this back.

And then I simply said, recolor the shoe in image number one, only using colors from the color palette and image number to keep the background white, keep the lighting and the angle the same. These are already some good pointers when you start. Very simple. If you create an image and you want to have references, so a balance in existing shoes, multiple existing shoes, you simply pull them to that image with the plus and then you will see them here in your little problems window. Simply the way FLORA sees it is from left to right. Image one, image two, image three, image four. It's easy then to brief.

So here I say this is image one. The palette is an image two. So you can breathe it easily. What is it that you want?

So then the same thing also always mentioned, keep the background white because if not, I might get fancy shelving or whatever that I didn't ask for. And a lot of times, of course, just wants it simple. We keep the lighting the same and the angle the same. You could also say, should you notice here that it starts changing the shoe as such? You could say, keep the shoe the same. With the current models, you have to do this less and less.

But if you find that's an issue, say that. Now, but then we get this result and that's crappy. But not just here. Let me double click on it.

Then you can see it a bit better. This is not just because it's not a pretty color combination. It's also because it starts applying the colors in ways that are technically not possible because here I suddenly have like a little blue bit in my soul. If I want that, it means that I have to have a separate compartment in my soul that I can put this thing in or I have to paint it by hand or whatever.

Here suddenly my whole TPU frame has gone away because it does not necessarily understand what this is, but this is an essential part of my mold. And also, I might have a certain idea of how I want to color this shoe where certain parts always need to have the same color and other parts always need to have the same color. So if you want to really do, let me go. Back to my screen.

So if you want to do proper coloring with AI, there are several things that you want that shoe to be the same for sure. You want the color system to be the same so that it respects the different elements of that shoe and you want to be able to direct better what color combinations in my making. Because I can imagine if you have to make a new sets of renditions for a shoe like this, that you would say, okay, I have a more basic range in neutral tones. I have something a little bit more out there colorful.

And then I have something really popping. So a really high fashion mainly meant for PR stands out in the store, which is much more colorful. And now, of course, you can go one by one to go like now make this now make that. And that's not necessarily how you want to do this. And that's exactly what platforms like this for.

So let me show you what I did. So and this is a more advanced workflow for it. So if you're new, just sit back, relax, look at it. Don't panic, because I'm not asking you to replicate this yourself right away. If you want to learn stuff like this, I would indeed suggest follow training for the training.

But I start from the start point. We have the shoe and we have the belt. And now what I'm going to do is in several steps. I'm sure that I follow this system when I color use these colors in specific color groups so that I get to color renditions and I can actually use and make.

So the first thing that I do here is just ask a text model. So the advantage that you have when you work in a platform like FLORA is that you can combine any image generation, all of the MLM. So the text models and also the video models and now also 3D. So you are going from one flow to the other and you can combine all the different elements and you can choose from a wide range of models that you can use instead of having. Do I have to subscribe to check GPT or do I subscribe to cloud? You have everything in here.

So I'm asking here analyze digital and make an overview of which part is which material and which color and then also analyze which parts have the same material in color. So the purpose of this is that we can eventually develop a prompt that would let us recolor the shoe but keep the color system that the example has intact. Use hex codes and textile Pantone numbers to describe the colors. So just for your information, we of course in the industry work a lot with textile Pantone numbers and the LLMs know these also and can convert usually a hex code.

So the hex color code that you have as a green color to a Pantone number. So describe the materials to the best of your ability. So do not yet create a prompt to recolor the shoe. Just clearly define the color and material groups in the shoe. If color differences in the image are just caused by light reflection, then ignore them. Just look at the part as a whole. As an outcome, only give the color groups nothing else.

Now, of course, this also took me some time to understand what to put in here. And for instance, one of the reasons why I say don't make different colors based upon lighting is because it's something that it started doing where it's oh, I see a shiny bit here and that's a lighter orange. Of course, that makes no sense for me to recolor because that part is orange and that it's shiny and then shows lighter in the picture. Okay. Nice.

But that's not a separate color application. So I decided to do this in GPT-6 Astra, one of the newer models, but what you will see, you can choose many different models here and that's the same for everything. So whether I put in a text node, an image node or a video node up here, you will find all the different models that you could use. So I will give you some general pointers as we go along. And other than that, I would simply say play around and try because that's also how you learn. Also how I learned what model does what well.

So literally for every note, if you want, you could use a different model. You need to, but there's a possibility. So say that I first then analyze, okay, how is this shoe even built? Because I ran the prompt, running a prompt is here with the arrow.

And then here next to this arrow, you see a one and a plus and a minus. This is where you can then say, oh, I want to run it several times or we don't necessarily do that with text. You do that more with images, but just to show you where it's at. It gives now a really elaborate analysis of this shoe of all the different things that it does. I'm a lazy user, so I usually first run it with an image to see if it's correct.

But of course, as soon as you notice that some things are wrong, if your shoe doesn't come out as you want to, you have to go back in here and see if there's something wrong with the analysis. Because it took me indeed a couple of tries, right? To make this question. So then it would analyze it, right?

Then what I've done is to go to a next text model. As soon as you go to more complex workflows like this, you usually have to work with several steps. So also for those of you that are trying to make things like this, don't try to put every question or every step that you want to do in just one note, because that's I tried that the lazy user and that got completely, went completely wrong. Because now what I'm going to do is the next step is which colors do I want to use?

Now, of course, I have this model because I don't want to specify now make this color combination. Now make that. I also want AI to surprise me, but I do want AI to pick from my palette and to do that with the system. So I was what? I want one color rendition that only uses colors from the base colors. I want one color rendition that uses one base color and then maybe two of the fashion colors. And I want one rendition that does not use base colors, but only uses two fashion colors and a pop color.

And of course, this is up to you how you want to. You might have different combinations in mind, all good. And then you simply go. You are a other pair or colors letter I gave here, which are the available colors. Select exactly two colors. I did not like the brown and light green combo. It was on some shoes.

It was hideous. So I asked it to not do that one. But pick any other color could even be two of the same colors if you make to fully monochrome and another thing important. I also like the same pair as the previous run.

This is still relatively difficult in AI. You might sometimes have it run a few times to have it pick different colors and not now it picks, for instance, gray and light green. Fine. But you might have to run it a few times sometimes to get then whatever brown and black or whichever combo you look for. Same I did here.

This is the color combinations where I take three colors. So there's two colors from the fashion and one color from the base. So again, specifying which colors are there. I did find that I did have to specify that otherwise it could confuse even though I did connect it to the ballot and then the final thing, same thing.

And then I made another step that says, OK, now take that color system. So how is the sheet built up? Then also take the colors that this thing shows and then look at the shoe. That's when I brought in the shoe and then recolor it.

So and then I got the result that I wanted because now what you see from this first row. So here you see already the three differences. Right. Let me show you the top. Do you connect from one to the other to the other?

So now, indeed, I have here it shows at one point two browns and it made me just this fully brown color rendition, which can be a good commercial basic color rendition here. And here, of course, it only has two colors. So since the original model has more than two colors and then needs to start coloring multiple groups in the same color. But works fine. And the reason why I do it in such a complicated way, of course, what you can do is go back to where I was here and simply brief it one by one. Right.

So if you are a beginner, don't sweat. Brief it one by one. But if you want to automate it, my whole goal, of course, is that this system that I'm building now that it will not just work for this buffer, all my shoes. Having it work for all pallets, I found is a bit difficult.

But my goal with this was, OK, I'm going to build this. And then my goal is if I were this shoe with another shoe that it would still apply colors in the way that I'm briefing it to do here and still have that end result. And if you work more often in those are the end goals that you usually have, because you would probably say. Click and go get coffee and then you want to come back and find your color renditions as you wish.

So you could run this whole system several times over to get multiple color combinations. I'm now just showing you the results of one rendition showing it works right here. You then have the base color, but then with red and orange with the green. Then here we have much louder because we have two fashion colors and then one pop color all that works. What I then did, and this is again to show how advanced I can be, is to indeed work towards an automated system.

So I copied the whole thing. And then what I can do is to build a technique. I'm not going to show you every step, but I'm going to show you the base principle. So because this system that I have now, I would simply like to have as an automated thing, maybe also share with my colleagues, share with other people I work with so that we have a similar system. Say that I select all of these. That's how you would normally also select something.

Then you already see here a button. Oh, do you want to make this reusable build technique? You also see build technique. Doesn't matter which one. If I then click on it's going to ask me two things. First of all, what is the input? What do what does the system need to run?

In this case, we have a few options. So technically I need the shoe that I want recolored, and I need the palette. If you also want to make changes to these prompts all the time, then this is also input. But generally I don't want that. Generally, I just want these prompts to run in the background because these are good. They analyze the whole shebang by themselves. I just want the end result, which would then be those color renditions.

So here I would say click one, the other. Then you can see here selected here on the side. You can fill out what is my input. Right.

And then if I were to say continue, then it's going to ask your output. So the purple bits are the inputs. The green bits are the output. And then I'm only selecting these three images. And this whole circus that is in here, all my LLM stuff, I formulate it in a way so that I can automatically make it that I don't have to do this, of course, manually every damn time.

So let me show you. Of course, I'll click it off for a second because I already built that technique because, like I said, my goal was I don't just want to do that one. I want to be able to do any shoe. I will apply this color palette because having it automated so that it also works with very different colors is possible.

So what I did now was just take another shoe. There's an Adidas shoe. This is the also how the technique shows. Right.

So if you have that technique, you will find that here in your sidebar. Oh, where does my thing keep jumping? Sorry. Here on the sidebar, you see the stamp logo and that has your techniques.

So you can click on that also to select techniques made by others to do techniques that you made. And then you see in the here, this hides all this LLM stuff, which is in the middle of just says you need a base shoe, you need a color palette. And then we're going to get this base color combo, fashion color combo, fashion and accent color combo. So that's what I did. I just took another show, like that in here, like the color palette in there.

And then I just say run. I'm not going to do it now because for the sake of the webinar, we don't have that time. But then you see it worked. There are some comments that I have because here it suddenly made this piping, which is probably not it.

So it might be that I have to go back into the analysis. But this is most definitely not bad at all. So for the complexity of the operation that we are doing, is this clear so far to people or do we have a ton of questions? I do see some things in the chat. Do I find that building a technique requires to shoot to be almost identical? No. In this workflow, no.

But that's only if I do all these steps in the LLM, if I don't do that, then it does not get this complexity of all the different shoe parts, the type of color, then applies it to now say. Because we're now going to work towards as a next step towards a tech pack, which is, of course, the main thing that comes into play in terms of boring administration for this is there are some designers who love it. Fantastic. But many designers do not like that.

So I'm going to show you how to automate the process immediately towards tech pack from here, because you have your color conditions. And so the first thing you might want to do is to show the shoe from different angles. Again, we'll first show you the beginner version and then the technique. The beginner version could simply be I have say I have my image. I pull my little plus I make an image node or I double click choose image node and then connect them.

And then you can simply ask if the shoe the same, but show it straight from the front. There you go. That was the results. Say that I want to show it from all angles. Just a quick trick to show you want to show from all angles in an animation very fast. You can ask it to make a three quarter angle for which I usually advise for animation. And you can see how well it does this.

And just a few months ago, this was not even possible in this way that would really do keep the shoe exactly the same show from different angles. No problem. So what you can do very quick trick for those of you that do animation say that I want to make an animation of this shoe and I can just show it from all angles. I either double click and choose video or I drag and choose video.

Now, if I want to have a looping image that simply spins the shoe, give you one trick, duplicate your image so that you have two of them. Why? Because usually if you bring in a video model, the reference images in and there's two of them, for instance, then it often automatically sees the first image at the start frame and the last image as the end frame. And since I want to do a full 360, having the end frame and the start frame the same helps.

And then you simply get this. So I simply asked, make a looping video that spins the shoe 360 degrees, showing it from all sides. Keep the background wide. Image one is the start frame and image two is the end frame. I asked for no sound, but unfortunately it did not do that.

Here you go. But look how well it does that. Even with lighting, you can ask for lighting effects if there are certain things you want. There's a little tiny cookie. When it restarts, you can probably fix that in a video editor.

But this is so quick to just show from all angles. I've had people join my 3D course who simply just wanted this. You don't need to be 3D modeling for this. So this is just animation. Very quick, very simple.

So stop that for a second. But it's going to get more fun if you think this is right. Because, of course, I am way too lazy to make one by one all these angles, please. If I want to make a tag track, I want to show it from all angles. I cannot use animation. I want to use pictures.

So I have created before a pull orology angle generator. And that is simply bringing in one image and then it creates it from all angles. In one go. So here you go. All of them. And look also how one does even the sole profile. Of course, the sole profile is guessing, right?

If you want this exact, you will have to bring in a reference image of the soul. If you don't have a reference image, it will just assume something. But this is really fast. I just say I have my color rendition finished. Let me zoom out a bit.

So I was here. I had this whole system. I create my color renditions. I then say, OK, I like this one. Connect here. Run.

And then you have in one go print. I then added to that the color palette. But why? Because I'm going now in one goal backpack. And what did I do? Let me zoom in because I think people that have not seen that are going to probably lose their pants now.

But this is an actually full functional color palette. I've asked for all the different things, all the different hex codes, color pantone, textile pantone. It shows you all the different things and shows you the details, everything. So what I did, I took in one go all these images, all these seven images that for most of the slides and I took also the color palette and I said, make a tech pack that features images one to seven and the colors from the color palette in image eight.

But only the colors that are actually used in the shoes in images one to seven. Make a tech pack that features all these shoe images and a table that contains all the different parts in their hex and bento colors. Run. And then you get this. And I think it's awesome if I see how much time this usually costs. It even puts in the logo, everything. And of course, me being lazy, I was I want to do that for my other shoes as well in one go.

So, of course, I could actually do combine all these things together. Let me see if it wants to let me walk through here. Yeah. So what did I do just to show you that it works also with other shoes? I took my other color rendition that I had here to the base color in this case, plugged it into my technique, made all the color angles in one go, added the palette.

And then in this case, I also added the tech pack that I already made because I really like this layout and was like the same as this. I can imagine if you work in a company that you have a standard layout of how you are, how you want to take back. So and then I simply connected everything. Of course, I can make this also as a technique if I want to.

But just to show you the same thing in again, logo, this is Adidas, everything in there. So now if you also want to learn how to make this automatically editable so that if I smack a shoe in here, it's updates it and I can actually type in it. That's why you come to the heart rate because that's not what I can do inside FLORA. But I think this completes my first little bit about how to speed it up. I will very quickly, before I start answering questions, show you what gets me thinking. And that's the more experimental workflows. Because when you work in AI, you now suddenly have a way to cross over text, images, music, video in ways that you could not do before.

And in our heads, for instance, traditionally, we might think, well, let's do a shoe sketch. Then I'm going to do the render. Then I'm going to do the 3D. Then from there, I'm going to do video. That's the traditional workflow.

But what if I say I have a fantastic video that I love the atmosphere. Most designers are not very good at describing words. What is this? What is this atmosphere?

But you do want to have it as an inspiration for what you make. So you can do that now. So let me show you. This is Alia the alien. You can follow her on Instagram. One of my favorite. I love to real AI.

So here we go. This video. I don't know if you can hear sound with it, but it's not. An alien woman and she's dating alien guys. And it's a very fun, a bit surreal type of atmosphere. I really love the visuals.

So what I wanted was I want to have shoes inspired by that. So say what if I want to have shoes for her and shoes for him for the dates that she has in this case in the video. What you can do is ask a text model to analyze your video. And it's actually a muscle kind to inform with a Gemini is the best one to analyze a video.

So I did that. I simply connected it to text node and say, watch and analyze the reference video and then create four different prompts. Each of these prompts will describe a shoe design inspired by the color, the atmosphere, style, and storyline. The video. I want two prompts that will describe a shoe for the female character and two prompts that will describe a shoe for the male character. You have completely to define the prompts and what the design should be as long as it matches with the video. Make sure that each prompt is written as a paragraph and that there's an empty line between these prompts.

I will show you later why I want this. Also, since the image generation model will not be connected to the video, each prompt needs to be described in such a way that it can create the shoe without actually seeing the video. Finally, make sure that issue is don't straight from the side, lateral side. And to make sure that each shoe is meant for an adult, not for children.

This is because in my first renditions, I got kitty shoes for her, especially which I did not. Now, why did I want four prompts as paragraphs with lines in between? What you can do if you want to make four prompts in one go, but normally, of course, now I have four prompts in this window. If I now simply click an image model, it's going to try and run all those four in one image. That's no way you want.

So what you can do here is use something that we call an action. So if you double click, you can also choose whoops in the bottom like my window in front of it. Take my word for it. But here in the bottom, there's something that's called action. And in this case, I chose split text and then I can have a split mode. In this case, I chose paragraphs so that it split me in one go in the different prompts as separate. I like this as a workaround.

I found that it doesn't work very well in techniques, but as quick workaround, this works because then I suddenly have my four, splits the prompts. And then I can connect to an image window and run. And here I also show you, I have a bunch of renditions for each one because here you can then really quickly say, OK, run it in that banana, bro. Run it in another one. You can choose here whole lists of different models.

This is really the place where in ideation, you see the effect of the different models. I think style-wise, it's really well video. Love also what they did for the male character. This was actually my first one where I forgot to ask the white background, but I really love the shoe.

So I kept it. This results in shoes that you could not have just thought of yourself. It's a specific style. It's more complex, even here.

But and that's my final thing that I want to show. Because at Phorology, people who know me and know how I work, I don't like A.I. For just making weird stuff that we could never produce, because I'm always thinking of how can we make it and how can we make it better? And there's definitely ways that we could probably create stuff like this, but expensive, difficult, blah, blah.

So what I wanted to do, I set myself a challenge last week for this webinar to see if I could make a reconstructor, meaning that take this shoe, take any shoe as a star. And then I want this thing to make different constructions. So I want a variety that is different parts, but not glued together. So that could be stitched or clicked. I want a variety that is one shoe fitting inside another.

So maybe like an exoskeleton and an inner shoe inside so that it's modular, designed for disassembly. And I want a shoe that is fully mono material. So maybe 3D printed, maybe molded in one go so that we can recycle it more easily. And then I tried, OK, can I make something out of these extreme shoes? And I can.

So look here what I did. So here is another rendition that it had made just to suggest. And then I indeed created prompts asking for the different things. So this first prompt is asking for a more sustainable version without glue, use anything with stitching, with textile. The bottom cannot be attached with glue.

Then I asked this one, which is a more difficult one, longer, for split it up in an inner shoe and a skeleton over that is not assembled. And then the last one is the easier one, which is mono material. Pretend that it's printed or molded or not. And then we got let me just show you is really fun.

So here so it translates a lot of the upper shoe knits. I know that we have Vasi here from Athens Knits. So I'm eager to hear what he thinks of that. Not everything will be fully possible.

But nonetheless, very interesting. For such a elaborate complex upper, I have seen knits or woven knit combinations. These could be separate patches. Interesting also to indeed make these modular ideas where you indeed have some type of looping or stitching. How can we do this? Not each of them might be fully possible, but it does give you a lot of inspiration of how you might be able to do stuff like this in a much more sustainable way.

So also here, the skeletons, not the easiest thing to do, not the most pretty ones, but it gets the idea. That one's also not bad. That one's also not bad. This will still be multi material, right? Imagine combinations between this knitted upper from the first one, the exoskeleton from this one, and then looking at the same thing, but full mono material.

This also interesting because this could be full molded with a textile in sock in that's been come out. We see more and more of these shoes being made. So elaborate will be complicated with separate flaps for molding and printing. Always think of separate pieces that you might click in or do.

So then finally, just to show you fun of that, I also did that as an, I made this as an automation. So I took the other shoe that I liked, yanked it in, and then it will create automatically now three, no glue assembly, three modular design and three mono material. And look at how much fun we're having here. It might be fully possible, but it's a good starting point. These types of constructions are interesting.

This is actually quite possible also to attach this little flap here like that. This might have to be embossed in it. Look at this skeleton. This is making me so happy. You can make minor adjustments, right? You could say, okay, this part maybe needs to go. We have other tools that you can learn when you work with us for painting and stuff.

This could even be a separate full piece inside where you can make it custom. I love this skeleton. It doesn't have to have all the hinges. This can just be a one. Also here that you have this entire spine like decoration on the side as part of the exoskeleton. These are things that are doable. It's a good starting point.

It needs more design work. So don't worry. You still have more than enough work at the designer. More, because we can start doing this complex things.

And then here, all the ones that are, this would have to be printed in resin. I don't see any other option or stuff like that. I cannot believe that, but yeah, this is things that I think are really interesting and make me tick. So now I think since we're in entering the final 15 minutes, start answering also questions. Let me open what I have so far.

So also guys, if there's other questions, let me know. Nico, I have two questions for you. Tell me. Yeah. First question is about the positioning of the shoe. We play with the lateral view and wondering how this would work to see the full upper view. I believe you covered it already, but just curious to hear.

Yeah, we have covered it. So here again, I could also say, I'm gonna hold on. Oh, my screen is getting away from me. Here again, I would probably do to either connect my technique show from all angles, or you could simply say, if I want to see this one from the top, I say, go here, image, and then whatever you could either go show this show straight from the top. And that, by the way, for those of you also interested a new demo also out on FLORA's YouTube channel, which shows how to change the last shapes of shoes. And so say that you generate this from the top.

So let me just run this. And you don't like the last shape. There's actually several techniques that you can use to change it to a shape that you want, including an actual last that you have. Derive a last from a model. That's other technique. Well, here we see then from the top, here, we would have to do something with that fan like thing here, because I don't, that's too, that's making the opening too small, but that's simply showing it from the top is easy. If you then, since this is non-existing shoe, want it to be more slim, you can always say, okay, connect me here.

And say, keep the show the same, but make the toe shape pointed. So, and you can try stuff here, but that's questionable. Another question you had, Sasha. - Yes. The second question was from Lily. How does FLORA know the materials for the tech pack? - So the way it knows is simply from training.

So what we see now, also why we see that this is now even possible to do these tech packs is because it has been trained with so much information and we shifted from having to train the models with metadata. So with tech photos to having neural networks that are intelligent and teach themselves what things are and connect text and imagery. And that's where we make the big steps for our data. That it starts understanding also a little bit better, not just what this is called, but also what it is.

Here you see that the change in last chip, but it can also be that it has, that there's a material that it doesn't know. And then you can write that in your own description of the image where you say this material, that whatever, say that I want to make it clear that this is a, knit of a full plant-based yarn that I could then write that's the information. Or I have to explain it. This is material, but this is that. You are also able of course to do the reverse to replace materials to have maybe little scans swatches of your suppliers. You can select parts of the shoe and say, put this material in there. And it does that better than Photoshop does.

So I think for a lot of these things, we are going to also move away from Illustrator. If it were me, it would go out the window tomorrow because these things, okay, I'm now looking at some extreme shoes, but this photo realism is more clear for your factory to understand what it is that you want. Other questions, shoot. There's a few about 3D, creating 3D models. Could you give us a look on how to create 3D models in FLORA? Yeah. And I'll show you a couple of different ones.

Also for those of you, cause I don't have the time because the generation takes a bit more time of a 3D model, which is understandable. So much data. So good news also for those of you who want that, there is a demo coming out in the coming weeks where I demonstrate some 3D also on FLORA's channel. And also for those of you interested, I have a full course for full 3D AI, a workshop that you can follow in your own page online, where you actually also make a full 3D model that you can 3D print, you can subtract the last, I mean, all of that.

But let me start simple. Okay. Couple of options. First thing that you need to understand, which is limitation if you're doing shoes in 3D, is that of course, when you have a shoe, we want to usually see the toe shape, right? The top, you want to see the bottom, which is the sole profile.

Then you have all the sides. 3D AIs are now developed mainly for gaming characters. So they look at the object as if it's a person. So they look at front. Like face, back, left, right, the side of the person.

So that's not ideal for a shoe because we would want buttons. And that's not something that you can easily do. So the simplest way to try, so say that I want to do this shoe that I have here. Say I want to do this one in 3D. Doesn't matter. I click, I'm first going to create this in a three-quarter view. Why?

Because if you show it straight from the side, what a 3D AI usually does, it just imagines the other side mirrored. So that's not always what you want. So I would first say I create a three-quarter view so that you see a little bit more two sides. View of this shoe.

Okay. I'll be going quite quick. And then from there, you could go into a 3D AI that can use one picture. You have for many of the AIs that you can use multiple views. There's a couple of complications with that. And I don't have enough time at this moment to show you that.

So for that, you will have to watch that demo or come follow my training. Because of course, the more views you have, the more executive views you'll be. But at least here you see a bit more of the last shape and that there's also a fan on the other side and whatever. So now if I want to go into 3D and go to 3D, and then I'm going to use the one, actually there's the one on top, the Hanyuan 3D. Why do I want to use Hanyuan 3D? Because it uses one image and it's very fast. You have other 3Ds that can do stuff really well.

So I'm going to go through the 3D model and see how it works. And then I'm going to go through the 3D model and see how it works. And then I'm going to go through the 3D model and see how it works. And then I'm going to go through the 3D model and see how it works.

And then I'm going to go through the 3D model and see how it works. So I'm going to take a second. So I'll let it run and then answer other questions in the meantime. We'll get back to this to see what it looks like. Other questions? Question from Jose. How many iterations did it take you to get to your final techniques honed?

Was Astra a game changer? Sorry, what was how many steps? How many iterations did it take you to get to your final techniques honed? That was half a day of work. That's also why I'm not releasing this workflow yet into the FLORA community because I want to improve it a bit and work on it a bit with my participants until it's perfect. Those are complicated.

But this is also one of the most difficult things to do in AI, but I wanted to try. One for the angles took me literally 10 minutes. Super easy. And here you go. 3D.

So you can see it now from all angles, but you also see already is that most of the models already now make the shoe hollow, which is a protocol. So because technically, yes, we wanted to be hollow, but say you want to do 3D printing. You might want to determine by yourself how thick this shoe should be. And that is not something that you could do in here.

So that's why we have a special app that we developed that helps you do that. So here you can then indeed see what you want to have wire frames. You can see other things you can see that actually, for those of you that are experienced 3D, that it makes usually not so much this one, but the other 3D eyes make very high res meshes, which is not necessarily bad, but usually makes them too heavy for a lot of other processes. But that can also, again, we have an app, but we also have an app.

But do you see now how simple this is? And then you can indeed download this. You can see that a little bit. Then in my top bar here, I can choose download, and then it will give me the different options. The same way you actually have for image. If you would say download that you can choose, okay, the standard.

But here you then get standard depending on what it is that you want. So if you want to, if you want to do 3D printing, usually STL, but you'll be also work. So that's how simple the 3D can be. It's a little bit more to it. And also, of course, you see here that it has the kind of doctor, a sole profile.

So let's give it a few more months. Oh, yeah. Cause actually I saw just for those of you for information, cannot do it in here yet. But one of these 3D eyes now has an app on the phone where you're sitting in front of it, like a scale, and then take it to here.

And then you have immediately a full 3D model. So the developments go very fast. I met with all of these companies in China, cause they're all Chinese and to talk to them about the full word developments. There's a lot coming. Yeah.

So we just rolled out a 3D node last week. You can import 3D objects onto the canvas and there's a wide variety. Of different 3D models for you to use and toy ran with. I've been really liking meshy seven. It's been really good.

But yeah, you can turn any image into a 3D model. You can plug in multi views and produce pretty excellent 3D models. But with that, we are up on time. Thank you so much and take care. Thanks a lot. Thanks guys. See you on the next one.