Latest / The Interline Podcast / Is There Such A Thing As An Authentic Synthetic Image?

Transcript

Is There Such A Thing As An Authentic Synthetic Image? · The Interline Podcast

• speaker-0: Hey, welcome back to the Interline Podcast. You're seeing a lot of generated images these days. And I don't have to know anything about you or where you shop to make that kind of proclamation because everyone is seeing a lot of generated images right now. That's because generative AI has very quickly become a massive part of fashion's imaging pipelines. From the world's biggest retailers working to automate photography that previously came saddled with super high overheads. To small brands who suddenly find themselves able to produce PDP content at a scale they couldn't have before, or to generate campaigns they never would have had the budget to shoot for real, everyone's reckoning with what it means to be able to create pre-real lush looking fashion photography without cameras, without models, without lighting, without locations, and so on. And in a lot of cases, that reckoning has turned into AI-generated images being just everywhere, labeled and unlabeled. If you park the ethical concerns for a few minutes, and I promise it will just be a few minutes because we're coming back to them in this episode, this all makes pretty unavoidable commercial sense. Unlike 3D slash CG, which have a high skill floor and require a relatively heavy time investment, anyone with the right tools can generate an on-model image of a garment that's potentially fit for purpose, depending on what that purpose is, in minutes. I've had a few conversations recently with people who've called this all essentially a solved problem or something close to it. I don't know if I fully agree with that, but it is hard to ignore the sheer speed at which AI has taken root in the journey from a finished sample to a believable biofacing image. But if that's all table stakes now, then what are all the brands who've taken generative imaging and embedded it into their content pipelines doing next? Technology in the big lab open market sense isn't standing still. GPT image 2.5 came out a couple of weeks before we recorded this episode. And that model contains some pretty pronounced advances in multitone editing and coherence, and video generation too is rushing forwards. But what is a fashion brand actually going after now? The universal answer seems to be control and precision. It's no secret that a lot of AI generations still don't hit the mark and end up being discarded. Depending on who you talk to, the actual percentage figure they place on the state of the art there can be wildly different. But the essence is that there's still plenty of ground to gain when it comes to making image generation work more accurately, more consistently, and more reliably. But what you then build on top of that pursuit for accuracy, or as today's guest is going to call it, authenticity. It's starting to look very different company by company. Some want full lateral coverage for their online storefronts. Some want to push the frontiers of generative campaigns. Some want to get into programmatic content refinement and personalization. Some want to do all of that. And before we start, let's take the ethical and practical concerns out of park now and put them back in gear, because the way people feel about AI culturally is Also very different person by person, and isn't getting any less polarized over time. For all its commercial impact, generative photography is still far from universally well received and well-liked on the consumption end. And that's without mentioning how human models, photographers, stylists, makeup artists, and so on feel about their livelihoods at the moment. So there's a lot going on in generative pipelines, in other words. And I could have brought plenty of different people on to talk about that, but A fair few of them would have been able to give me answers about adoption and interface design, which wasn't really what I wanted today. We've done that relatively recently. Instead, I wanted to bring in someone who's been involved in actually training frontier models for image generation, specifically for fashion, and someone who has a candid insider's perspective on how far industry uptake has really progressed, how much further it's likely to go. what technology vendors are prioritizing and building and why. So I've invited on Erin Daggan. Erin is the CEO of Botica, which, and I quote, offers AI fashion models trusted by world-class fashion brands. Now, you know by this point that I like talking to CEOs who know their products and who know their market, and Eren is definitely one of those. I don't agree with him on everything, and you'll hear some pushback from me as we get into this, but I do think Erin has a well-justified perspective on what it means to create authenticity out of tools that build synthetic content and about what that's going to do to the industry over time. So let's hear that perspective. Okay, Erin Daggan, welcome to the Interline Podcast. speaker-1: Thanks, Ben. speaker-0: Glad to have you here. we start every one of these shows with a couple of things. We take a look at the guest's day-to-day role. And we asked them to define something. I'm interested in what your day looks like because from my outside observer's perspective, the rollout of generative AI in the imaging pipeline for fashion just seems to have blown the doors off, not to put too fine a point on it. It's happened so quickly and so pervasively that I'm actually not really sure what it looks like to be the CEO of a company selling that kind of technology to brands. In my head at least, it's half brands. biting your hand off and half people sort of slamming the door in your face for ethical reasons. I I imagine there was a short window where maybe you spent a lot of time justifying and evangelizing image generation. And I do want to press you on some of the economics and the ethics and things later, but I think it's fair to say that basically any company that does product photography now is already thinking about experimenting with and confronting the idea that Generative AI might be a more straightforward, low overhead, maybe potentially more creative way to do it. No doubt the technology is still improving, but the business case seems to have written itself, essentially. and that's resulted in a very competitive landscape, which I'll also get to shortly. But aside from the commercial side of things, there's also a lot of feelings about generated images. I think the you know, the the market I described, where there's a lot of people excited about it. lot of people very pessimistic about it. That feels accurate to me still. How accurate is that mental picture? Like how are you seeing the the industry respond to what you're putting out day by day? And when you do sit down at the start of the week, what are you paying the most attention to? speaker-1: Yeah. you know, I think that picture was definitely maybe very accurate two or three years ago when you know AI was very early and people were still still maybe a bit more on the fence in terms of are we going to use that or not? Is it any good or not? you know, i is it just honestly a trend that will pass or not? And I think now the conversation has changed. my day in my day to day that's conversation is mostly gone. I think brands kind of accept that AI is going to be a meaningful part of their workflow and they're now trying to understand, you know, just how much of their workflow is going to be done with with AI. so I think there's there's less resistance and less evangelizing as you as you've kind of said it that we have to do and we had to do a lot of it starting up. yeah I remember talking to brands and they were like Look, we've tried, you know, computer generated com photography which you know came before AI, and it it was no good. It never lived up to our expectations and and all of that. And we had to do so much work just showing them you know that that the technology has changed and that the results that we'll be getting are substantially better. and and that I think shift has mostly completed. There like you said, there there are still a lot of open questions, but I think the general sense is the brands kind of accept that okay, AI will be there. And then it's kind of a a matter of how deep the integration will be and how much we'll we'll use. the current questions in my mind are a bit more they're blessed about you know, does the image has six fingers or not? Which you know used to be kind of the main the main test our customers would put us through. Right now it's more about how controllable is it? does it fit my brand? How like can I am I, you know, legally covered when I use it? can I scale it to thousands of products? and how does it impact the rest of my workflow? So you know, images are shot in the studio, but there's so much work that happens later. And I think brands right now are kind of navigating how how the entire process is gonna change. So Back to my day-to-day, maybe that's kind of my focus. I try to imagine a world where AI has been adopted. so we kind of were past the conviction point where really think about what happens once like you assume that AI is there. How does it impact the entire brand? And I would say the the greater creative processes and tasks that brand can brands kind of work through. speaker-0: And when you mentioned that, you know, the the question being more about how deep are we going to embed AI in our imaging pipelines, our marketing and so on as a brand, what would you say is the average depth right now? How far have people gone? speaker-1: Yeah. I think here is one of those places where the the the scale is is it really, really depends on the brand. We we have brands that produce thousands of images a day right now on Bodica. And we have ones that are just dipping toes, trying out, seeing what works, what doesn't, trying to think scope it out a bit better and understand which products are better like a good fit for AI, which ones they still want shot physically. So it really, really depends. We see, you know, the digital natives that go kind of all in on AI and everything is AI powered. And we have more traditional brands that are just trying out and seeing if it can help them, you know, work out some kind of surprise issues and stuff like that. so they if you know if they have a deadline that is coming up and They see they won't be able to pull it out, pull through with traditional photography. So it really, really varies. speaker-0: And if you think about the ones that are doing the thousands of images a day, are they primarily e-com? you know, are they product detail page images? Are they campaign photography? Are they kind of social channel assets, or is it just a pretty broad mix of it all? speaker-1: Yeah, I think everybody the pattern that I've I've seen is that most of them start with the PDP. because PDPs are you know, th there's just the sheer scale of it that you have to that you have to process and kind of push through to get, you know, every product out in time and to make sure it's well represented. So we typically see customers start with the PDPs and then move on to socials, marketing, campaigns, all of that. but PDPs are I think an an an e an easier starting point for most brands. speaker-0: But I think that that mirrors what I've seen as well. And I think I think the value is is easier to quantify if you're going for reducing the cost involved in that. You're going for like gallery completeness and and all of these other metrics that you would use that are they're harder, more rigid metrics than campaign photography, which is which is quite subjective. For the definition side of things, we don't do day to day. I wanna ask you to define authenticity. Now, don't take this the wrong way. I'm not purposefully being a pain in the ass here, but in a very literal sense, you trade in synthetic content, or at least you trade in the tools that allow companies to create their own synthetic content. But your product promise right at the top of the website is authenticity. So tell me what it means, in your opinion, to create an authentic synthetic image. speaker-1: excellent question. When we started, I think the company, we spent a ton of time kind of working on what makes an image a good image. Because there's for every product page, you could shoot it in an endless, endless variations. And the one thing you always have to get right is the product itself, because it is a product page. So to me, the Authenticity is really is the product represented faithfully. As a customer that has seen this image and will purchase this product and w when they get the product will be exactly what they've seen and what they've imagined and how they thought it's gonna, you know, look and feel and drape and and and all. So to me, yeah, an image can be synthetic and still still be authentic if it portrays the product authentically. So there is a a little bit of a an interesting mix here of taking something from the real world and trying to reproduce it digitally, but in a way that is reproduces, I have to say synthetically, but in a way that is is faithful to the source. So it's all about that, I think. so if you are representing the color, fabric, texture, draping, fit proportions. if everything's accurate, then the customer will be happy. The customer will see something that was rendered synthetically but is faithful to kind of the the product they've received. So I think it's it's all about that gap. It's all about making sure that everything you see is is as accurate as it can be. And yeah and and making sure that customers eventually get what they've what they've purchased. speaker-0: When you say as accurate as it can be, what is the state of the art when it comes to that kind of authenticity if we scope it like that? You could be percentage of generations that you consider to be accurate, it could be outcome. Just just walk me through what you see as the frontier of authenticity. How far are we to that goal of what makes a great image, but it's created synthetically? speaker-1: Yeah. Yeah. Look, I think we can look at it in kind of a a number of different ways. the you know, there's you could look at kind of negative signals like a return rate, right? if a product was returned because it was unfeefully rendered. Or you can look at just how many images that a brand produces on vodka are actually used under a website. So past their internal kind of QC. we we typically find that this is a a good enough measure. it's not perfect, I have to say. you could everyone can make mistakes. So even if you are the brand and you know the product, you may have missed something and talk about how we try to help brands work through that. but in general we see that it's about 90% of the images that we render in Bodica don't get. kind of sent back. and when I say sent back is we we have this feature where if you do find a mi find a mistake, you can kind of mark it up and and we'll fix it for you. it's kind of that guarantee we we give you for for getting you from, you know, the AI state of the art to a hundred percent. And yeah, like I said, the state of the art right now is around ninety percent and you know, you still have ten percent, which is a significant part of the business that you need to kind of fix. So we chose to fix this with a dedicated retouching team that we have. And it's one of those decisions we've made very, very kind of early on, not understanding how big of an impact it will have on kind of the trajectory of a of the business. but it's been one of those services that just really helps you kind of trust the system and the fact that if the system doesn't get it right, then somebody will get it right for you. so yeah. speaker-0: When when you're talking about clients who are generating thousands of images a day, that's a hundred or so images a day per client that are potentially need that additional work if the if the error rate is in the ten percent range. speaker-1: Yeah, exactly. speaker-0: Fine. So you mentioned some of the work that you've put into kind of getting to the state of the art. And I think that's a good jumping-off point for us to establish some technical basics before we get too deep into some of the economics and the ethics and things. There are, as I mentioned earlier, this is a competitive space. There's lot of companies doing image generation for fashion, either for initial creative design exploration or for the spectrum of downstream use cases that we've talked about, PDP campaign and so on. A lot of those solutions are also built on the same commodity models from the same pool of AI labs. And their points of differentiation seem to come from the interface, the experience of the team, and so on. Explain to me what is going on under the hood with Botica. So I did some reading and your infrastructure partner modal published a case study of Botica's operations that talks about, and I'll quote here, proprietary foundation models with tens of billions of parameters. And roughly 15 models serving production inference, end quote. So it's clear that Botica isn't just an aggregator of other people's models, but it's also not immediately obvious from that quote to me at least what the smallish proprietary model does or what those 15 runtime models are. So walk me through the technology infrastructure behind the images and the tools. Are you selling your own models, existing models in a better harness, both or something else? speaker-1: Mm-hmm. Yeah. so I think the the important differentiation here is the there's the generation model that is our own foundation model. And there's like you said, about a dozen other models that participate in the generation process. you know, for tagging metadata, extracting all sorts of features, so pretty much acting as part of the kind of generation pipeline. But the generation itself runs on a financial models that we've been researching here. internally, kind of in-house. My co-founder is actually leading this. it's kind of an interesting team of really, really smart people, to say the least. So all of them did their bachelors when they were around 15 year olds, and kind of have had a ton of experience in general AI. So yeah, we we've we have been building our own financial model. It's I have to say it's pretty is it's a it's a pretty big undertaking, resource wise and time wise, but it does I think it it is worth it. so we're definitely not taking, you know, one of those open or commodity models and kind of putting a fashion interface around it and calling it Podtica. we train models that generate damages that our customers need and want. So to me, I think that the big the w when we started working on the company, this was kind of a big part. I told my co-founder, look, I want us to be able to take customer feedback and very literally translate it into model improvements. we want to be able to control the kind of the entire cycle. We don't want to be attached to whatever roadmap that, you know, Nato Banana or Jipity Image has. so we we chose to kind of take the I think the the long way and to build ourselves. to add that there's a lot of you know, stuff that happened in in that model around taking a lot of references in, you know, being able to take about twenty reference images of the model, the human model, I mean, background, makeup, lighting, the products themselves. we go into very, very intricate details and and to kind of take them all in and as at once and generate one very accurate and still very visually compelling image. So yeah. But I think that the the distinction here is that we don't really need to own every single model in the stack, but we we own the generation part. speaker-0: So I'm going to split my next question up into two pieces. And the first one is a bit of a follow-on from that, which is you don't want to be beholden to the roadmap for GPT Image or Nano Banana or any of the other labs. But as a company creating your own foundation model, you are competing with them to some extent. So I'm curious to see: do you do you worry or do you keep tabs on? How frontier development and the state of the art in those from those big labs with those big image generation models. Do you do you worry that as they progress, they start to erode some of the advantages of having built your own? speaker-1: So we we definitely track everything that's coming out of the big labs. and we try to use that progress ourselves where it makes sense. I think particularly recently there's been a lot of progress around kind of image editing. so that has been kind of useful for us. We didn't have to, you know, reinvent the wheel here and to build another very big model just for editing. I think the core problem that we try to fix is is a bit different. Then the big labs. So the big labs are building generalist models that you know try to be good enough in everything. the problem with fashion is that you know nothing is really good enough. you're chasing perfection. And like it's like we kind of previously said, it's all about that authenticity. And if the product is beautifully portrayed but slightly off, then the the image is not usable. So I think we're Just the the way we optimize is we say, okay, how can we be more accurate and not how can we be more creative per se. So I think a lot of kind of the big image generation models try to be very, very creative. speaker-0: I I I would agree. And I and it did seem like GPT image two point five, I think it is in particular, seemed to have a big focus on like coherence across multiple turns or ed editing and and and and I understand that. So y you can take advantage of some of those open market features without having to give over your entire stack to that single generalist model. You're able to take advantage of the best characteristics of what's out there. But to apply it to your own foundation model which is narrowly f focused and scoped on fashion. speaker-1: Yeah, exactly. speaker-0: So the the the other part of my of of that question was how do you So when we talk about kind of the state of the art and the authenticity side of things, and we are talking about representing real physical products, what standards do you set at the point of input? Because the the vision, as I understand it, is to go from any kind of photo, so lay flat, ghost mannequin, like any any kind of photo of an existing final approved sample and to then be able to generate the sort of PDP assets in particular that we've talked about already. What sort of standards do you need to enforce at the point of input? Is it literally any photo? Is it is there is there a certain level of quality and different angles and sort of representation that you need at point of input because I there's inference from kind of known quantities and then there's inference from like unknown quantities. And like a single photo of the front of a garment is not maybe necessary information to get an accurate or authentic depiction of the back of the garment. So to get a good image from Bottiker as you define it, what do I need to be putting in about my product? speaker-1: Yeah. Honestly the you know there's a lot of things that you can work around. but what we found is the one thing you can't really compromise on is it's just natural light, natural lighting. if you you know take a a product and let's say your light is very, very yellow, we could try to adjust, you know, the white balance and to play around that, but the truth is Once you've captured the that photo, nobody really knows what the true color of the product is. So using natural light, I think is is important. Again, we you can try to course correct after you've shot it, but it's one of those things you don't really want to do. color is extremely important for consumers if you're selling something and the color is off. I think that's I think it's the number two reason for returns, if I'm not mistaken, after fit. so you definitely want kind of that part done right. but yeah, we we typically ask for just wallet photos and front and back view of the garment. And when we you know when we started Bodica, the technology to go from kind of flat lays to on model was still very, very early. so we used to only support existing on model photos or mannequin photos and go from these to kind of professional. on mobile photos. And over time, as we saw the technology progress, we kind of I would say just r released a lot of the constraints. And now the main thing we care about is can we see the products well and is it well lit? you know you you always want kind of decent resolution. So anything that comes from a phone is actually good enough because you want to see, you know, the zipper and the logo and like all the fine details. that you don't want the model to miss. But I would say at the current state of things with AI, decent lighting is pretty much all you need. speaker-0: That's a useful answer. So we've just talked about representing products. I'd now like to turn our attention to representing people. We've we've had a couple of conversations on the show in the last couple of months about likeness and persona and things in the age of AI. We recently ran an op-ed from Benjamin Hori, which is about ownership of faces and things. It's safe to say there's a lot of uncertainty around the legal status and the ramifications of. Showing people and faces in generated images, especially when the provenance of the original image, the original photo, whatever was used as an input is kind of hard to ascertain because in a lot of kind of node-based generative workspaces, the initial image could have come from pre-existing photography from a previous campaign. It could have come from somebody importing an image from web search. It's unclear in a lot of cases. Every botticum model, human model, is generated from scratch with no real person as the starting point. You've said that there is, and I quote, no risk of misusing someone's likeness through this pipeline. Tell me how that works. Because my understanding is there's two ways a likeness can find its way into a generative output. One is the one we've just described, because the user used it as an input. They took a celebrity, they took a previous model engagement that they'd done, and they used it as an input. The other is because that person's likeness is sufficiently represented in the original model training data, i.e., the person is famous or just gets photographed a lot for some other reason, for some combination of their features to then exert enough influence over the model weights that you get something kind of recognizable, even though the model is not specifically reproducing a a single image or a single face. The same modal case study I referenced before talks about you having a hundred terabyte image data set. Which is you know, it's a lot of images, and I can't imagine those are all synthetic training data. just walk me through this, because surely there can't be a scenario here where you've rebuilt all of the primitives of fashion shooting with no reference to real photography of real people at some point. There must be some licensing or something else going on. Explain to me how that promise works. speaker-1: Yeah. so I think it's a good point. There are two slightly different things here. So yes, our data set is large. It's like I said, it's about a hundred terabytes. We've built it over years. but that data set is really more about kind of garments, products, garments, outfits, fit, all the metadata around them. actually it doesn't contain human faces. So when we get new data, we just crop out the face to ensure that nothing gets stored. as part of this data set. So it's it's a lot of data that teaches the model, you know, what carvings look like, how term fabrics behave, what happens to fabrics in different poses, how fit changes across different bodies. so it's not about how do we like you said how do we sort of memorize faces, right? because that is one of those things that tend to happen with with large models that have trained in kind of public data. the the distinction I think here is that garments have no structure. So you need to have a lot of data in order to understand them, while faces actually have a lot of structure. And so you know eyes go here, nose goes there. so there there are some degrees of freedom, but it's a lot easier to just play with within that structure and to get new, entirely kind of unseen faces. and to sort of imagine those almost as you know if you were a painter and you were to paint a face, you have a certain a number of elements that you kind of play around with. so this is kind of what we try to do with our our creative director. this is what they do. so yeah it's it's one of those areas where you don't really need those enormous amounts of of data. So when we create vodka models, we just kind of start from scratch, like a blank canvas and not as you've mentioned, we you know, some some platforms allow you to upload your own model to create a model that is that resembles somebody that is kind of real. so we just don't allow that as a general sort of rule. there is no way to kind of prompt us into creating a new model or to upload your own reference image. We just kinda don't allow you to do that. so yeah, that's Okay. It's obviously very conceptual, but this is kind of how we treat the problem. speaker-0: It is conceptual, and I'm gonna give you a conceptual philosophical reply to it, which is a you know, your painter example. If I was a painter and I sat down at an easel and I painted a face, I would subconsciously be painting some aggregate of all the faces I've seen in my life. like I'm not I'm not doing it completely suigenaris. it was it would it would be coming from things that I have seen. and the same would be true if you gave me If I sat down with like an an Elder Scrolls game and I was like, okay, I'm gonna I'm gonna build a character face here. Somewhere along the line, I am instinctively pulling that from prior knowledge and and prior experience. I don't think there's any way that you can completely extricate the idea that somebody might end up recreating something that looks superficially like somebody else. It's no different to musicians sitting down and coming up with a melody and then going, crap, I've just reinvented something that already exists. the it's it's a that's a very real thing that happens. But the fact that it's not trained on a big public data set of celebrities and previous fashion models I think is the key. speaker-1: Yeah, yeah. I I think that's exactly right. Because you can always you can always accidentally create something that is similar to existing work. So by the way, when we upload a model to the system, we try to do you know some kind of reverse image searches to make sure that it doesn't look like somebody that we can at least via kind of public means find. so we we try to kind of cover this, but you know it's always kind of the the work of our creative director and, you know, their ideas and how they play with, you know, the different elements that really make up a face. but yeah, I I definitely agree with with your point. Cool. speaker-0: So let me pull on this thread just a little bit more because I've spent enough time talking to brands over the last year or so to get feel for how people want to use generative workspaces and image generation tools. And they seem quite keen on taking existing references as inputs for these kinds of workflows. That's behind what everybody listening to this will be familiar with some of the legal test cases and stuff for retailers who've said we've previously run Campaign, we've previously done a campaign with this human model, and they're like we we believe that their model release or contract is sufficiently loose in its language that we can use AI to remix it. That's the extreme example. But then you will also have a brand who, for instance, has done a bunch of shoots in the past, and they might go, I just want to extract the pose from this one. I you know, the the the this person is standing exactly the way that I want to have this generative model stand. I just want to use that as a reference. Now you mentioned, you know, you're not gonna let people upload any face references based on what we've just talked about. But what if they wanna upload pose references, environments, materials, you know, something else that they want to riff on? How do you manage that side of things? speaker-1: Yeah. so yeah, we we never let them, like you said, upload kind of the the the the human faces, or I I have to say other elements that make up a human model. So you can't really copy somebody's I have to say the term likeness is is pretty broad, right? it's not just the face, it's the body, it's a lot of other things. So when if you want to give us existing references, we'll normally kind of give you our own internal tools that you can sort of play around with. So for example, for posing, we just have this nice interface where you can pose a virtual model and then you have that pose and it was created digitally. It doesn't copy a real photo. environments are I think a bit more tricky here. as you know in general environments are you know they can be totally imaginative, they could be real places and then you know you you don't really know where that data of those real places came from. So we typically work with the customer to understand what exactly is that that they need and then we try to find a way that I would say is the safest. but yeah I I have to say, you know, there's there are still with AI and with all those different elements to make up an AI picture. It's hard to kind of control all of them. So we try to really de-risk the you know, the the biggest factors which are the the model and their likeness. the other parts are I have to say, it's uncharted territory. Hello. speaker-0: Yeah, I think I think that's right. And I think the fact that it's so uncharted is the reason that people fall back on existing behaviors like let me take a reference and let me just Yeah. Let me just riff on it slightly. Like I think I think that I think those two things go hand in hand, to be honest. Brass tax question. Who owns a bottica image? Who owns the output? Because what you've described so far is is, as you said, your own tooling. It's a fairly closed loop that you're taking when it comes to image generation. If everything is happening in that kind of proprietary closed loop, I think people might sort of ask themselves how free to travel the outputs are. Do you provide the images to users under license? Does the customer get full perpetual rights at the time they generate an image? Explain to me how the output works. speaker-1: Yeah. so the customer can use the outputs freely, I have to say, as long as it doesn't conflict with our terms of use against, you know, certain types of uses. but those terms are, let's say, w we built them around the use case of fashion brands that can use models across you know, their website marketplaces, advertising and social, so wherever they need them. so even though the the generation happens within the bod within kind of the Bodika platform and it uses proprietary tools, the the image is very much the customers to use. So we don't really want the value of vodka to come from kind of trapping the asset and you are on kind of a a harsh license. it should come from I think the work that Bodica can kind of keep doing for you. speaker-0: I want to get your take on some of the economics. I mentioned it up top. now I as I said in my intro question, I I'm pretty clear that I understand the business case for replacing traditional product photography. We'll we'll we'll zero in on PDP photography, but I can you can reference this to campaigns and everything else as well. I fully understand the business case. If I was running a content studio for a big fashion brand, bringing a lot of styles to market for a lot of different markets potentially each week. And I'm photographing hundreds of gamets a day with human talent. I'd find it very hard to look past just the sheer potential cost savings of doing things this way. Models, photographers, stylists, locations, and so on, they're expensive, they're hard to procure, they're hard to organize, there's they're messy, even if that mess sometimes leads to some creative sparks. I get it. Like the the pure commercial business case I get. Implicit in that. business case though is the the understanding that if you're saving on production overheads, which is what we're talking about here, that means reducing the amount of work sent to those people, to those human models, to those stylists, and so on. When I had Glaciani Rocha from Zalando, who's created director at Lounge by Zalando on the show a couple of months ago, he was pretty open about this. He sort of said, look, the balance of roles that are needed to execute on a campaign vision is shifting. Maybe you don't end up with fewer people overall, but the type of work those people do is changing. And depending on the final deliverable, that does mean some roles may eventually not be needed at all. Now, I think the obvious quick answer that people would give to this would be okay, so we eliminate a bunch of stylists and photographers for basic e-commerce stuff, but maybe we reassign them to take on image work that we wouldn't have done before. Maybe we do more campaigns. How much of your Customer base's workload, honestly, is content that wouldn't have been made before versus how much of it is content that's now faster and cheaper to make without the people who used to be necessary to make it. speaker-1: Yeah. I think good question. so the pattern that I've been seeing, I think so there's been there there has been a shift in this. as we kind of as we spoke earlier, you know customers kind of start by replacing, right? Something that they're they're they're doing. So they take the PDP imagery, and they kind of you know start doing that digitally. But we now see, I think, the sort of the first wave of customers who've been using Modica for you know two years now, who are kind of I would say taking the next step in terms of using AI. So that is when they really start leaning into it in terms of creating stuff they couldn't create before. So think about kind of different imagery for different markets, a lot more variations per SQ, doing operational stuff like replacing out of stock. products that are you know in existing looks within other products or testing different PDP images and refreshing content more frequently. So there's there's definitely been, I think the the the first step has been let's replace what we have. But once AI sort of compressed the the existing workflow, then the amount of content actually starts expanding. So you can get a lot more a lot more work done, a lot more imagery done. than you could that you that you could before. as you said, the the the economics of it previously simply w weren't really possible. So it starts as how do we do something you know faster, cheaper, but then it really I think now is the step of okay, how do we get just more value and more photography and you know, in a sense, how does the generation never stop? speaker-0: Okay. So but do you see those new opportunities, th those expanded use cases as a bigger market for AI, generative AI tools, or do you think they open up a reskilling market for some of the people who might have fallen by the wayside in that replacement phase? speaker-1: Yeah. so I think I think it's a little bit of both, to be honest. i it it it's hard to claim that AI, you know, doesn't replace some of the people in in the process. but at the same time it does create a lot more opportunities and not more work downstream for other people to kind of say work on. So I do think there is, there there are people, and we are seeing this by the way with brands already, that are sort of shifting to more kind of digital roles from physical roles. but I can't say, you know, that everybody has a new position in this in this world. It's I think it's it's unfair to say that. speaker-0: I think yeah, and think that's been true of a lot of technology revolutions as they have happened. So I I don't think AI is exclusively to blame there. And I will say that same Zelando conversation also included quite a lot of explanation of the fact that AI is not necessarily a shortcut when you're using it for creative purposes, you still end up with the same amount of people having input on creative direction, you still end up working with agencies and things. It doesn't fundamentally change it, but it does It does shuffle roles around and there are roles that will be, I guess, progressively deprioritized if AI becomes a bigger part of the imaging pipeline for fashion overall over time. the other half of the business case, beyond what we've just talked about, which is that images are easier and cheaper to create at scale, is that those images then go on to sell more clothes. That's the promise here behind this, or to well, reduce returns long term but sell more clothes. One of your published retail case studies, you talk about fourfold conversion result. And I want to get into where that comes from. Is that four X conversions compared to products that had what you would call like incomplete galleries before, where you only had certain angles of certain colorways, and maybe you had videography of one model and were not of another? Is it a case that actually it was poor quality traditional photography before and the AI image quality is better? and drives more behavioural change. Basically, are people buying more products because AI completes a picture that was only partially painted before, or is AI doing something else to get transactions over the line? speaker-1: I think that I think they're both really. I think first of all, you know, each brand that I work with has this this story about a product that you know that wouldn't sell for whatever reason. And then they shot it again and again and again until until it started selling. we worked recently with a very large kind of a billion dollar brand that kind of insisted on shooting one specific product I think like at least five or eight different times through the studio and that became kind of a bestseller of you know, it it took a while, but it became a bestseller. once they've kind of tweaked a pose, I have to say. So it was just something about the the way the fabric sort of drapes that made it sell. So I think AI gives you the the ability to kind of iterate and experiment. the problem with photo studios is that you shoot something, it's static, right? That's just the way it's going to be portrayed. And you don't really have the kind of time or resources or even attention, I have to say, to reshoot it. So with AI, it makes it a lot easier to just try something and try something else and and kind of keep doing that until you find something that works. but going back to kind of what you initially said, yeah, for a lot of brands, it's even just you know getting their products on models, which they only had if they only had, for example, flat lays before. or just upping the the level of photography that they can sort of get to. it's it might be easy to forget now that everybody has access to AI, but you know, not too long ago, smaller brands didn't really have the ability to produce images that looked like a billion dollars images. So yeah, I think it's buff of these. speaker-0: And so what there's one other element I think to make conversion and customer experience, and that is the idea that people like seeing garments on models. They also like seeing garments on models that represent them. And one of the I think something that's it's a big promise of generative photography, but it's also something that bothers people about generative photography, is this idea that it can be, to borrow a phrase from your website, inclusive by design. Pro-argument for using AI to represent products on a more diverse set of human models is, if I understand it correctly, is that it allows companies to build in more representation and diversity into campaigns so that a wider range of shoppers can see products on people who look like them. The counter-argument would be you didn't need AI to do that. You could have done that pre-AI. And doing it now represents a sort of fake or sets representation that's maybe worse than the real thing, not because it's synthetic necessarily, but because it closes doors for modeling talent that had, in many cases, only just had them opened. I think everyone's conversant with the idea that fashion had a very fixed idea of what constituted beauty and acceptable body types and things for a long time. Like a lot of people, I'm pretty conflicted about this. Again, I understand the both arguments there. And I'm also part of a demographic that just gets implicitly catered to in all marketing. So maybe I have a privileged read on it. What do you think when it comes to this diversity stuff? If you increase through AI the range of faces and body types that we see in marketing and on PDP pages, but the actual range of people who get paid to be in front of or behind the camera ends up getting narrower, is that a good trade? speaker-1: Yeah. I honestly here as well, I don't think there is there's a clean answer. This is this is very clearly a very real trade-off. Yeah, yeah. AI can reduce work for models, photographers, stylists, everybody that is part of the production sort of work. yeah, there's it it's hard to argue against this. so I don't think this is there is an argument against this. I think the what we're trying to say is the scale of people who view this content is a lot bigger. So we're talking billions of people who shop online. And historically, you know, in a like it like we previously said, due to economical reasons, most of them have seen products that are on, you know, a very, I would say a relatively narrow range of of people and models. So with billions people, billions of people that are shopping, we can make them feel a bit more included with AI. And yes, I you know, there there is definitely like I said, there it it is a trade-off. but it I think it the the impact is is still meaningful and and net positive. So if you as a cons as a customer you're able to see something that like you you're able to see a product on somebody who looks a bit more like you, it could genuinely kind of change how confident you are in in buying it. So I I can't say that AI kind of solved inclusivity. It's it might be too simplistic, but I think the the cost here i is real, but there is a lot of benefit to it as well. speaker-0: That's a reasonable answer. Final question before we go, what's next for Botica? I'm not gonna name names here, but it seems like every generative workspace for fashion is now converging on the idea that they're also gonna be a PLM platform and a CAD tool and they're gonna do technical specifications and pattern development. Are you looking at extending the footprint of your platform the same way, or are you staying in the image and photography lane? speaker-1: So we're going I have to say in a pretty d different direction than these guys. I don't think the the big opportunity for us, at least is to become another kind of PLM or CAD tool. there's I have to say there is significant opportunity there. It's just that we think there's like you said, plenty of people who are playing there. we're focused further downstream. kind of how do you like I said maybe earlier the in this call It's more about what happens when AI is part of the workflow. so how do you kind of run e-commerce content where you know the there's so much content? each product comes with, you know, ten different variations for testing and different markets and personalization that is running. So for me, the the next milestone for vodka is kind of becoming this almost agentic e-commerce team that helps you manage your PDP. So you you direct pretty much in body car runs. So you define the brands, the rules, your goals, and we keep kind of the catalog on brand, we create, we replace imagery, we react when products go out of stock or sell very well, we run A B tests and personalization, we localize to different markets. So we try to kind of take a lot of the work to the a lot of the busy work. So you can kind of focus on here's what I want to do, here's the direction. And we kind of do the, you know, the actual actions behind it. So I think maybe today most AI tools are still very imperative. You ask them to do something, they make it, right? and then you kind of sort of take it from there. So to me it's all about closing that loop between here's a new product and here's my vision for it. And then vodka takes that vision and translates into photos and then runs tests and publishes the product and monitors it and pretty much helps you kind of focus on the big picture rather than the you know day to day work. speaker-0: That does seem to be a fertile space for a lot of companies as well. I think I believe YouTube are leaning more into the idea of kind of A B testing of videos, automatic generation, like programmatic generation of thumbnails, like some of those gate. It seems like we're headed towards a a web for shopping, for media consumption, for everything else that is a lot more reactive to customer need in a way that is not scalable for individual people to do. and i you need a lot of trust in the tools to be able to hand some of that off to direct rather than to manually do. But it also seems like the value of doing that is potentially going to be quite high in the near future as well. speaker-1: Yeah, yeah, definitely agree. I think it's about turning content from again something that was shot once and then sort of presented, you know, to millions of people to something that is kind of always on and is constantly sort of evolving and generating according to the viewer. so I think, you know, e commerce is a it's a pretty interesting industry to test this concept in. speaker-0: Agree. We're gonna be keeping tabs on it over the next couple of years. as that evolves, maybe we'll bring you back on and we can chat about how things have progressed at Battica and how things have progressed in the wider context. In the meantime, Aaron, thanks for letting me quiz you on a whole bunch of different stuff today. I've I've enjoyed this conversation and appreciate you being here. speaker-1: Thank you so much. It was a pleasure. speaker-0: And that's the end of my conversation with Erin. It was a long, heavy, reasonably technical one, so I'm gonna get out of your way pretty quickly now. But hopefully I fielded the kind of questions there that you feel like you would have asked in my shoes. If I didn't, I'd want you to let me know. You can email me at editor at theinterline or oneword.com. We're already canvassing for the next Mailroom episode too, so your thoughts and questions on not just this show, but anything are very welcome, especially if you feel like recording them as a voice memo for us. I'm gonna back next Tuesday with Grace for the latest edition of the edit. And next week I'll be back to quizzing another executive one-to-one about one or more of the things I think are shaping where technology and fashion are going. For now, though, thanks for listening and I'll talk to you again really soon.