- 3 months ago
Comment l'intelligence artificielle transforme-t-elle le processus créatif dans la mode ? Aux côtés de jeunes talents de l'Institut Français de la Mode, cette session explore les nouvelles approches du design, de la création visuelle et du développement de collections. Entre expérimentations, nouveaux outils et réflexion critique, les intervenants partageront leur vision de la place de l'IA dans les pratiques créatives de demain.
Category
🤖
TechTranscript
00:22Bonjour, alors je vais me présenter, mais on va faire en anglais la session aujourd'hui.
00:28Je suis Giovanna Casimiro, professeur d'un institut français de la mode, responsable pour tous les sujets de tech.
00:35Et aujourd'hui, we have here Alexis.
00:38Hey Alexis.
00:39Hello, hi everyone. I'm Alexis.
00:42I am a designer at Balenciaga, and I was a student of yours, and I did the master's at IFM.
00:52Yes, in fashion. So today our discussion is, we're going to be exploring with Alexis a bit of his creative
00:59process in fashion design with AI.
01:03And I think it's really relevant that you are here today having this conversation,
01:06because globally, the usage of AI is increasing in the fashion ecosystem, especially for fashion creators.
01:15And I think there is a bit of friction sometimes, how to use it, why to use it, when not
01:21to use it.
01:21So it will be really incredible to hear your process.
01:24You're going to show to us a bit of those images, right?
01:27And I will start with one first question.
01:29So you already introduced yourself to the public, but how did you start working with artificial intelligence at IFM?
01:36You had a course on that. How was that?
01:38So I started experimenting first with mid-journey, when it was a bit less socially acceptable that you were using
01:45AI.
01:47And at first, what I would show to my tutors, it was not the best, and it didn't really show
01:52any kind of handwriting of what you were doing.
01:54And then I actually, we had one student before me that had a scholarship.
02:02And then in my year, we were actually six students that we were awarded this scholarship by LVMH.
02:08And they gave us access to a startup called LLIM.
02:12And then it was mostly just like a platform that would give us access to other AI models.
02:17And we literally saw the models kind of progress as we kept using it.
02:23And so that's kind of how I started my journey with it.
02:26And I was always very, like, wary of actually how far it could go.
02:30And especially thinking about, you know, this being actually just like the average of everything.
02:36I think that was like something that I was worried.
02:39Because it's like you're getting so many images from so many places that maybe it's like a color palette.
02:44You just end up with like a muddy brush, you know?
02:49But yeah, I think that's it.
02:51Yeah, and I think that at IFM, we do have what you mentioned, the LVMH chair.
02:56Where students, they have a chance to learn creation through text.
02:59So it's quite interesting.
03:00So now I want to move to the second question.
03:02If you could walk us through how your collection begins.
03:06So we understand a little bit more how AI really blends into your process.
03:11And if you have images you want to start sharing, feel free.
03:14Yeah, of course.
03:14We'll go through them later.
03:16So when I started working, I first, like, look at artists.
03:20And this is a work by Jeff, no, Mike Kelly.
03:23And I was very interested in, like, conceptually what artists try to achieve.
03:29And I was looking to Mike Kelly, Jeff Koons, and Jordan Wolfson.
03:35And what I was interested in is why were they adopting certain aesthetics, especially kids' aesthetics, to talk about work
03:43of arts.
03:45And I think this relates to AI because when you're also prompting AI, you're trying to make it think like
03:51you want it to think.
03:52So I think it's very important when you're designing fashion, you want to be critical and you want to understand
04:00what you're trying to design and what kind of problem you're figuring it out.
04:04So I think that's kind of, like, my starting point.
04:08And it's very important for me to know what are we trying to achieve.
04:14And, yeah, I think that's...
04:16And when you're trying to decide what you're trying to achieve, do you use, like, do you have an analog
04:21process before?
04:23Like, that is not digital.
04:24Like, do you do, like, I don't know, you visit places, you take photos, like, you get inspired.
04:29Like, what are the activities?
04:31Just curious to understand what is your process before prompting or actually creating your interaction with AI?
04:38I think there's definitely, like, a collection of images that we all have as designers on our phones.
04:43But also, it used to be that you would go a lot to museums.
04:47But now I think social media and even, like, TikTok and everything can be an inspiration, which is also, like,
04:53I think we're all kind of chronically online.
04:56So that's also, like, a way for us to have, like, a massive gallery on information that then we can
05:00even give to the AI.
05:01And even today, sometimes I would get things from my head designer and I'll just get, like, a screenshot and
05:07he'll be, like, can you give me some information of this and that?
05:10And, like, I mean, like, generate versions of this.
05:14So where I think it used to be more like, oh, you go to a museum, you go to an
05:18exhibition, but, like, today everything is quite digital, to be fair.
05:23And I think, I remember that you have another colleague, I mean, on the same chair, Paul Biu, who is
05:32also another designer that now he is working for a large group specialized in AI and fashion.
05:37And he told me something very interesting that when he works with AI, he has these two categories that he
05:43puts AI into, like, the models.
05:45One is parrots, because they repeat more of the same, and another one is aliens, which are the large language
05:53models that make him go beyond.
05:56It's, like, really co-workers that contribute to his process.
05:59So my question to you is, like, in your process of research, how do you see AI?
06:04It's more of a sketch tool, it's a co-worker, it's a sketchbook, it's something else.
06:11Like, how do you think AI is positioned in your own workflow?
06:16So I think maybe it's worth to talk that, um, maybe I'll just refer back to my work on the
06:21masters, and when I was, like, first experimenting with, like, these kind of, like, child aesthetics and putting onto the
06:28mannequin,
06:28and then also, like, maybe then making them in, like, smaller scale, so they would be, this would be, like,
06:333D work that we would do at the studio.
06:37And when we were first using this for AI, actually, the models were, like, super generic, so we would get,
06:43like, very obvious and almost, like, first-degree work.
06:46So this was...
06:47The AI generated...
06:48And this would be kind of the result of, like, okay, I want this image from Jeff Koons, and, like,
06:54these kind of military references, and it would just kind of be a bit of, like, a...
07:00I don't know, like, an average of everything, so I think this was, like, my first experimentation with it, which
07:07was just, I don't know, even this image, he kind of looks a bit like Bad Bunny in a way,
07:12so...
07:13Um, it was really not accurate back in the day, it was just kind of very first-degree.
07:20I think that's how it first started, these models, like, yeah.
07:25So you would really just really test some possibilities?
07:28Yeah, and in a way, it helped me because it made me think of, like, okay, the AI that's coming
07:34up with, like, the most average things, this is the direction I probably shouldn't go.
07:39So it was kind of helpful in that way.
07:41It's a risk assessment to almost, like, where should I not go, and would you be sketching first manually, then?
07:47So you would do your manual sketches, and then you would, would you put your sketches on the large language
07:53models you're using?
07:54Yeah, actually, for these images, I was fitting, like, some of the 3D models that I was doing, like, the
08:00maquettes, and also I would give it some sketches.
08:02With sketches, it used to be, it used to struggle a lot with sketches, but now it's actually really good.
08:09Sometimes it looks a bit anime, but that's the only thing, probably because it has so much data from anime,
08:15but, yeah.
08:16And then when you were saying that you were doing, like, the 3D models, were you using a particular software
08:22like Clo3D, or no?
08:24No, not really. I find Clo3D a bit more for, like, people that work a lot on just things that
08:30are meant to stay on 3D, like digital, but not, I think the translation for physical to digital, it's kind
08:37of tough, and it's not really the French way to do it.
08:40In the sense of, like, the French, they work a lot with mannequin, and sometimes when you translate patterns from
08:47Clo3D to the physical world, it's not a one-to-one, to be fair, so.
08:52So you generally would sketch physically with your own notebook, and then eventually pass that as a reference to the
08:59AI, and then you would create these different iterations.
09:02Yes, I think.
09:04That's super interesting. Now, I think the other question that when we see a designer like you working with Gen
09:10.AI for the creative process is to think, how can you translate these sketches to the simulations, like these images,
09:18and then later on to the physical piece?
09:21Yes. So, and also because, I don't know, I want to hear your opinion, but I think that AI, it's
09:26really close to really get to the physics of the real world, but there is a lot of expressions and
09:32lingo and the traditional savoir-faire that it doesn't understand 100%, so sometimes the generations will not be accurate, maybe
09:41now you feel they're better, but then how do you do this transition from, okay, this is a simulation, visualization
09:46of my idea, to the physical piece? How generally that happens?
09:52I'm just going to quickly skim through this. This is actually, it kind of ties to what you were saying,
09:57when we were generating with Mid Journey, we would get these kind of textures, and like, they would be like
10:03super exciting, but then as you said, like, how do I make this, like, it's like completely inflatable, feathers, like,
10:10and then it's like, where do we go from this?
10:13Like, so this ties to what you're saying, like, sometimes super exciting, but then maybe it should just stay as
10:20an image, but then at some point, the models started getting a bit better, and this actually was like, all
10:27of these teddy-verse were like, not, this is a picture from the studio where we had one model, and
10:32all of these teddy-verse actually, you know, it would be super wasteful to buy that many teddy-verse, deconstruct
10:38them, and then put them back together.
10:40So when you can have the computer do it, just to get a quick visual, you know, this is not
10:44saving time, but also money, and probably, like, not being less wasteful.
10:50And this was actually the image that I prompted it first, so I did do, like, some physical teddy-verse
10:56legs, and then I was able to deconstruct to, like, this.
11:00And maybe it wasn't precisely what I wanted, but you get a quick visual, and I still can get the
11:05continuity of the same model in the studio, same lighting, and it's very efficient for, like, the process, just to,
11:12like, okay, see if this idea works or not.
11:16This is interesting, because I feel that this also is a bit of a match in the cinema entertainment industry.
11:22I was doing a panel yesterday in a different event, and one of the things that they brought was the
11:27idea of a reshoot.
11:28Sometimes you do have a shooting that worked really well, but there are a few details that you need to
11:34rework, and then Gen.AI is really great for that, so you really don't have to ask for extra budget.
11:40In your case, it's more a practicality, not somewhat budget, but it's really interesting that you could really have the
11:45sort of multiple variations of the same idea, and really test, almost like an A-B testing.
11:52Yes.
11:53And then, let me ask you, for you to do this prompting, like, you initially started with the other image
11:57that has the bare legs.
11:59I mean, this was the initial image, right? Like, just learning your workflow.
12:03The initial image was this. This is, like, straight from the iPhone, and then I drew over it, and I
12:08told it just to, like, add more Teddyverse around, and, like, I was kind of trying to tell it that
12:14I wanted them to be deconstructed, but also, at that point, it didn't really understand what I meant with deconstructed,
12:21and, like, and then I would try to be a bit more explicit, being, like, maybe just some legs hanging
12:25around, or, like, a bit of the forest coming out.
12:30But, but, so, there's also, like, now I think also the translation of ideas, it's a bit easier.
12:35So, you feel that now, when you prompt, how complex are your prompts in general?
12:40Sometimes you have to be a bit bossy with the AI, like, I, sometimes, when I do use it today,
12:45you have to tell it, don't change anything else, because if you just leave it open, it'll imagine things, or
12:52when I'm sketching, and I just want to change the length of something, or the cuffs of something, and it's,
12:57I know it's going to be faster with AI,
12:59I'm just, like, just change this, and not nothing else, but, yeah.
13:05So, it's strict, but also, not be super creative, sometimes very concise, and try to have it not confuse things.
13:13When you're referring to different points of data, like images, it's, like, use image one as a starting point, take
13:20the shorts from image two, and then the jacket from image three, do not change anything else.
13:26You know that, it's, like, just putting AI in a box.
13:30Yes.
13:30But, you know, what I love is that, when I listen to designers, I feel that everybody has their own
13:36recipe in their mind on how, they're, like, what you just said, okay, reference image two, and then capture this,
13:41and use image one.
13:42One, I feel that you just create this instantaneous workflow that you learn to replicate.
13:48It's really hard to put that on paper and teach others how to do, but you kind of create that
13:53on your own, and you're using, to create these images in your current collection,
13:58are you using just one large language model, or you have a combination of them?
14:02Because you learn with Limn at IFM, and also with Midjourney, so today, what is your approach?
14:09I think models got a bit centralized, in the sense of, like, it is true that Google's models, whatever it's
14:18called now, Nano Banana, or OpenAI's model, is, like,
14:24they are the best at producing the best result, so I think that those are the two models that do,
14:32and they are also the more, they consistently get it right more often.
14:36Right.
14:37Yes.
14:37I think that there is also one that is becoming quite popular here in France called Owen.
14:42It's a very strong one that I think big houses like Louisville are using, and I don't know, have you
14:48tried?
14:49Yes, actually, we do use Owen at Balenciaga, and that's the one.
14:53Because I think they have also some sort of agreements regarding IP, which is kind of critical for the houses,
15:00and I know that other houses have tried their own AIs, but sometimes I think they don't have the same
15:10level of success,
15:11probably because they don't have the best data that, because Owen is, I think they call them wrappers,
15:17because they actually use the OpenAI models and Google models,
15:22so they have some sort of little magic they do on top of it, and that's what makes it more
15:28fashion or more design.
15:30But those are the most successful ones, I think, yeah.
15:34And I think that you're right.
15:35The thing is that when you look into traditional large language models, a lot of them are not trained for
15:40fashion terms.
15:40So I was doing an exercise with another colleague, Benjamin, who is actually the director of the LVMH chair,
15:48and we were trying to use multiple models and use the term raglan, which is the cut that you do
15:53on the back, you know better than I do.
15:55And that's a very specific term for couture, for savoir-faire, and most models did not understand, except Owen.
16:05So I think Owen, I think Nanobanana kind of understood a little bit, but Owen really understood.
16:11So I think the main difference that maybe we're looking into a future where some models will be really niched
16:17into specific needs,
16:19and the sort of glossary, because they will be trained into more specific, but I don't think right now we
16:24have any model that you say,
16:25this model is nailing the needs of the fashion industry.
16:28Do you think we do have one?
16:30I mean, yeah, as you said, I don't think there's one specific, but sometimes I do would say Owen is
16:36the more efficient one
16:38at resolving issues that understands the language, but I wouldn't say, and this is maybe a tough statement,
16:45I wouldn't say Owen is creative, though.
16:48Interesting. So in the category parrot versus aliens, it stays in parrot, possibly, which is okay.
16:56I think it's what works for a lot of brands that need to replicate larger collections.
17:01That's what you need. You need a model that really understands work jacket,
17:06and then be able to expand that work jacket in multiple collections, slight changes on the style, maybe,
17:13that probably work for a lot of them.
17:15And I think surely with agentic AI, soon we'll have, I mean, I just hope that sometimes we don't have
17:21to be waiting for models to finish their work,
17:24and we can just be like, okay, I know that I need to work on this and that dress, and
17:27hopefully we can get it done by tomorrow.
17:31Like, you can just leave it, like, you can just leave it, like, working overnight.
17:33True, and especially for when you talk about large collections and big retailers, that it's not really like couture,
17:39that you really need to go beyond in terms of creation.
17:42And I think that takes me to another interesting point.
17:45When we talk about the sort of homogenization of aesthetics, right, I think that's a very interesting subject,
17:53because the models, as you said, they are trained in a particular data set and kind of become a little
17:59repetitive.
18:00They are not really adding any extra layer of creativity.
18:03So when you are creating your pieces, after you created your AI-generated image, what makes that yours?
18:12What differentiates what is yours from what is the machine's input?
18:16Like, do you feel there is a difference, a separation?
18:19I mean, I think I want to, well, these are some other designs, but I'm just going to skip through
18:25the slides.
18:26Take us through all the images you want.
18:28Well, this, maybe, just to refer back to the previous subject, this is when we were using them for more,
18:32like, 3D experimentation,
18:34and this is when the models were getting good, and I could, I was working with a Korean company called
18:38CVSM,
18:39and they produce this, like, inflatable clothing and also bags,
18:44and they were kind enough to give me some of their dead stock so I could play with it,
18:49and then these were some of the, at that point, actually, the AI models were getting good enough
18:53that I could take pictures of the inflatables and then tell it to play around with shapes,
18:58but they were still quite symmetrical, which is interesting.
19:01Like, I would be like, okay, put them on the chest, but I would still not get them to be,
19:06like, in different angles,
19:08or, like, just, like, it would, and also this maybe refers back to what you were saying about physics
19:13and knowing the real world.
19:15It doesn't know what's back forward or, like, at that point, but this ties back to what we were saying.
19:21This was actually my first generation with Comfy UI.
19:25Okay, so it's a node.
19:27So I was, this is a node, but this was, like, the whole, like, workflow,
19:32and I don't know if this is when, when I was doing Comfy UI, I was like, I'm going to
19:37get so into it.
19:38This is, like, the future of the, how networks are going to be built and whatever,
19:43but obviously, like, models got better, and then I never had to look back at this kind of,
19:46back at this kind of interfaces, but what I find interesting about this is that
19:54I was prompting it to create an emo, like, an emo, like, I was just, like,
19:58I want, like, an archetype of, like, a style of person,
20:02and the fact that it got it wrong or was strangely right, that's what I found interesting.
20:10And that's what I find interesting, and maybe I was trying to look for, like, okay,
20:15can I try to have AI have its own personality, and I think something that I like to think of
20:21is that part of our humanity comes from being flawed and having mistakes,
20:26and I'm intrigued on, like, can we make AI be also flawed?
20:31And, like, because I think when you have, like, a designer, it's, like, they're interesting because of their flaws,
20:38they're interesting because of the way they see the world, but it's, like, can we have AI see the world
20:44differently?
20:45Not get, like, the best, but maybe just have, like, some difficulties along the way.
20:50Which I think it's the conversation of bias, right?
20:53Because when we think about what is beautiful, and I think part of the mission of creative designer,
20:58I think, is to really look into things that may not be, as per the standard, beautiful,
21:03and find the beauty in everything in some way.
21:07And I feel that, depending on the AI model, it will be really hard for you to push the boundaries
21:11on that end
21:12because you will prompt beauty, and that is associated to a particular data set that represents that.
21:19And, you know, I think that the moment we get through that loop,
21:22maybe we can start considering if AI can be creative,
21:24but, like, so far, it's creative in the same way on textbooks is, you know,
21:30it's, as you're saying, it comes with a lot of bias of what we think of society today.
21:36Absolutely.
21:36And now, how do you actually avoid your work from looking like everybody's work?
21:41Because I think when you look into image creation,
21:44which is a different set of work than what you're necessarily doing,
21:48we see there is a certain AI statics, a certain repetition of the same.
21:53How do you avoid that happening with your designs?
21:56And do you think that's also going to happen with other designers,
21:58that some would just fall into this trap of doing more of what is already happening with models?
22:03I think for sure, and this is me speculating,
22:06we'll see a lot of fast fashion just looking very similar or hopping into the same trends,
22:12which already happens.
22:13And because surely, like, if we are doing it at houses,
22:16I'm sure they're doing it in fast fashion.
22:19But I think what makes it distinct is very much, like, the person,
22:23like, their own experiences, their own imageries.
22:25I remember when I was in my bachelor's, they would talk about,
22:28they would call it, like, primary research in a very, like, official way.
22:32But primary research was anything that you were creating.
22:36So that was, like, a sketch or a picture or writing.
22:39And I still feel like that's what gives it identity to things.
22:44So as long as you can provide that kind of thing,
22:47it will always make it more authentic to you or to your life experience.
22:51Once we did a workshop that was about creative writing,
22:55it was not necessarily around fashion design.
22:57But I remember that the instructor was saying that for writing,
23:01your personal experience, especially when you co-write with AI,
23:05as a journalist, for example,
23:07your personal memories and experiences are really important.
23:10And I feel that that's a bit of that,
23:12that when you think about what is your, what is only yours, right?
23:16It's really your take in life, how you used to go to school
23:19or your memories of childhood or even what you like to eat today.
23:22Those things only you know, right?
23:23It's funny that you use the word memories
23:25because it reminded me to the movie Blade Runner,
23:29which I'm sure a lot of people love too.
23:31But Blade Runner, like, the whole point,
23:33it's that the robots, what they need is, like, memories
23:38and they need, like, this subjectivity of, like,
23:40who they are in this world,
23:41even if they're almost perfect humans.
23:44And I think that's always, that it's taught at art schools
23:48because my, obviously, IFM is a fashion school,
23:52but my bachelor's was at an art school at St. Martin's.
23:56And that's one of the things that you get taught in an art school.
23:59It's, like, what is really that subjectivity that makes you, you
24:02and makes you, like, you know, drives you in a way,
24:06like, what desires there are.
24:08Also, maybe this is another word that I like to highlight,
24:11like, desire is something that you don't get from AI.
24:14An AI doesn't desire.
24:15And I think that's, it's an interesting question
24:19because it ties back to, like, is AI conscious?
24:22It's very true, of course, that there are some theories
24:25that maybe it will be,
24:26but I think right now it's really hard for AI to understand desire.
24:31And maybe even just as a phenomenon culturally,
24:33I think we are always looking for the authenticity,
24:37but in many ways I think the pop culture
24:39feeds so much into memories that already existed
24:42when you look into all the nostalgia effect.
24:44Like, it's not that you're creating new culture,
24:46we're remixing culture that somebody else created generations ago, right?
24:50So we already live in a crisis of creativity, in a way.
24:53For sure.
24:54And maybe AI is just gonna enhance that problem, I do.
24:57Make it faster, maybe.
24:59Yeah, exactly, make it faster.
25:01So now I think that this takes us to another question
25:04that I think is very important,
25:05that before I do, I want to see how many people here
25:07actually is a fashion designer?
25:10Okay, here.
25:15And how many of you want to work in fashion?
25:19Okay.
25:19How many of you work in retailers?
25:24Retailers, magazines, infrastructure for the industry of fashion?
25:30Anyone here?
25:31Some.
25:32Anyone here for marketing?
25:34For fashion?
25:37Cool.
25:38So the reason I'm asking this is because I think
25:40one of the critical questions that we see is also
25:42will AI replace creatives?
25:45Of course, that's a question that we have been asking
25:47for the past maybe five years for the ones
25:49who are working with machine learning for longer.
25:51But now, you being a creative designer
25:53in the fashion space,
25:54do you think that's possible?
25:57Do you think it's gonna happen in some level?
25:59I do think, I mean, I think the next slides, yeah,
26:03this is what I wanted to show for,
26:05and this probably responds to that question.
26:08This was like an output for something,
26:11for a project within the house that this is not like
26:14any work that you'll see coming out from Balenciaga for sure.
26:17Just a disclaimer.
26:19Sure.
26:19Because I did sign an NDA, so.
26:23And this is like already the final product
26:25of like what a prompt would be.
26:28And what I find interesting is just like,
26:31you're already, this looks like so much more realistic
26:34and achievable,
26:35and something that you could even give to an atelier.
26:38So when you can hop so fast from like,
26:41this was the final product,
26:42which seems like realistic,
26:45and you start with like,
26:46okay, but this was maybe like the fitting.
26:49This was like a random image online.
26:52This was a dress from like Binted.
26:56And this was like an archive piece
26:58from like a different collection,
27:00a different time, a different designer.
27:02And then to hop so quickly onto like something believable,
27:06I do feel like it's gonna speed up the way we work
27:09because otherwise we would have to do this with Photoshop
27:13and it would take at least 30 minutes.
27:15And I think, I don't think it'll replace us.
27:20I think that's, it's, maybe it's,
27:23that's an alarming feeling that,
27:25and you know, obviously we're human,
27:27we don't want to be replaced.
27:29Yes.
27:30But, but I do feel like it's gonna speed up
27:33the process really aggressively.
27:34And do you think teams will be optimized?
27:36Because I think that we see this wave now
27:39of a lot of corp,
27:40not just in the fashion space,
27:42actually more in the tech space,
27:43firing people because they are optimizing their workflows.
27:47And some people are rising the AI washing problem
27:51that there is this layoffs justified by AI
27:54when maybe it's really not AI caused,
27:56but just economic caused.
27:58Do you think that the teams will be shrinked
28:00a little bit maybe or, or not?
28:02Or just optimizing the work that is already there
28:04with the team that is already there?
28:05I think creative teams tend to be actually really small
28:08because you don't want to too much,
28:10too many cooks in the kitchen.
28:12Okay.
28:12Which is a reality.
28:13Like I think Balenciaga at its biggest size
28:18used to be actually 40 designers,
28:21which is quite huge.
28:22Yep.
28:23If I'm not mistaken, I could be wrong.
28:25But it was like quite, no, it was between 40 or 20.
28:29I'm not really sure,
28:29but that's already quite a big team.
28:32And I do think that today you can get like a lot more done
28:37with current AI tools.
28:39And especially for like sketching,
28:41like today we do, it used to be that you always had
28:43to sketch from scratch.
28:45But today we know that something
28:47that will be relatively basic
28:48and we'll just change the proportions.
28:50Like let's say it's a tailored jacket
28:52and you just want to make it extra long
28:54and make these leaves shorter.
28:56Then for those things,
28:58it's like maybe you needed before two designers
29:00to work on the sketches.
29:02Now you do only need one.
29:03So I think, but also maybe it can be seen
29:06from the perspective of like it'll give more power
29:09to individuals, you'll have then someone
29:12who won't need a massive team
29:13to start creating something.
29:15Yeah, I think that on the atelier,
29:17definitely I think there is an opportunity
29:18to optimize processes.
29:20But I think if you look into the emerging designers,
29:23definitely it will give an opportunity
29:25for the younger generation, including you,
29:28if ever you want to do that,
29:29to really streamline the whole business
29:32through tools that probably will make
29:34the process more accessible, less expensive.
29:36Yeah, I think also what is interesting
29:39is that I was lucky enough to have that scholarship.
29:43But what I do find worrying is the fact
29:46that we shouldn't get to a point
29:49where these models are not accessible
29:52by everyday students or people
29:54that might not have the means to access them.
29:57So I do think that's very important.
29:59And, you know, it's a bit sad
30:01that these organizations coming from being non-profits
30:04to like becoming for-profit companies.
30:08It's a bit worrying, you know?
30:10It is, definitely.
30:11There is a lot of political conversation
30:12that comes behind all of those.
30:14So now our last question before we wrap.
30:17My question, I have actually two in one.
30:19So the first one is like, what is next for you?
30:22Right now you're working,
30:23you're doing your internship at Balenciaga.
30:25I'm sure that you have great dreams as a designer.
30:27But then also, what is the message you leave
30:29for the younger generations
30:31or anyone that wants to start using more AI
30:34in their own creative process as a fashion designer?
30:37I would say let yourself surprise
30:40by how capable these tools can be.
30:43Like when I was finishing my master's actually,
30:46and this is what this image has come from,
30:49it was I didn't want to spend any more money
30:51on the collection
30:52because it's already like really expensive
30:54to do a collection.
30:55And then I still had access to these AI models.
30:58So these were all images that I generated
31:00with just the runway pictures
31:06because the collection gets shot professionally.
31:09And I was just like, you know what?
31:10I'm just going to throw this to the AI
31:11and that's how I'm going to present my collection.
31:15And the jury was pleased with them.
31:17I kind of got away with it.
31:20And even like the details,
31:21like what I was mentioning before,
31:23like these inflatable panels,
31:25he was able to like completely understand
31:26what they were and put them at the back of a jacket.
31:30And even like incredibly like very cinematic stuff,
31:33I was very impressed by it.
31:36So I think it's let yourself surprise
31:38with what you can achieve.
31:39Even if it's not necessarily for fashion design,
31:42there's also so much more you can create with it.
31:44And like the fact that I was able to give creative direction,
31:48give it just really good images from a starting point
31:52and then still have like creative control of it.
31:55You know, AI, it's not just a tool for sketching.
31:58It's not just for visualizing designs.
32:01It's also for many other things.
32:02And it can,
32:04I think it's just let yourself surprise by it in a way.
32:08Amazing.
32:09So if anyone here wants to find you,
32:11follow and talk to you beside right now,
32:13after a stage where they can be in touch with you.
32:15I am on Instagram.
32:17My handle is a bit tricky,
32:18but like,
32:21but so it's four and then another four
32:24and then Angeles,
32:25but then you get rid of the A.
32:27Okay.
32:28So it's four, four.
32:30English.
32:31English, basically.
32:32Yeah.
32:32Okay.
32:32Amazing.
32:33Because I'm sure you're going to hear a lot about you
32:35in the coming years.
32:35And thank you for being part of today.
32:38Do you want to show any extra images
32:40that you didn't show to us?
32:41No, I think that's it.
32:43Thank you so much.
32:43Amazing.
32:44So I hope you enjoyed this session.
32:45I'll welcome you to talk to us
32:47after the session is done now.
32:49And thank you so much.
32:50And I hope you enjoy the rest of your day at Viva Tech.
Comments