Matthieu Rouif Podcast Transcript
Matthieu Rouif joins host Brian Thomas on The Digital Executive Podcast.
Brian Thomas: Welcome to Coruzant Technologies, Home of The Digital Executive podcast.
Do you work in emerging tech, working on something innovative? Maybe an entrepreneur? Apply to be a guest at www.coruzant.com/brand.
Welcome to The Digital Executive. Today’s guest is Matthieu Rouif. Matthieu (“Matt”) Rouif is the Co-Founder and CEO of PhotoRoom, the AI powered photo editor and listing studio used by entrepreneurs, brands, and teams who need standout visuals without the time, cost, or complexity of traditional production.
Best known for capabilities like fast, high quality background removal and product focused image editing. Photo room helps businesses create content for product listings, ads, and social channels more quickly and at a lower cost. Born in France and shaped professionally across both France and the us.
Matthieu is a Stanford alum and a Y Combinator backed founder. Before founding photo room, he worked in product at GoPro, where he learned what it takes to turn sophisticated imaging technology into tools that are simple, reliable, and loved by mainstream users.
Well, good afternoon, Matt. Welcome to the show.
Matthieu Rouif: Thank you. Good afternoon.
Brian Thomas: Absolutely my friend. I appreciate it. You’re hailing out of Paris, France. Today I’m in Kansas City, so we’re about a seven hour, six, seven hour difference in time and I just know it’s hard to make calendars, especially when it’s international, so thank you. Matt, if we could, I’m gonna jump right into your first question.
Photoroom is often described as a listing studio for modern commerce. How do you think about the role of visuals today in driving conversion trust and brand consistency across marketplaces and social platforms?
Matthieu Rouif: Yes, it’s a great question. Visuals are the shop front of commerce today.
And what we usually say at first from is like, people buy pictures online and then they receive products that most of what you do when you are buying is looking at a picture. You can’t touch the product. So. It is the most important touch point, the most important proxy for what you do. And so the work you put in the visual, the quality of the visual represent the best estimate of what the brand, the product, the work, the craft behind what you’re selling is.
So really making this. We see photo as well as making like this high quality studio quality unbrand imagery to make you look professional. So when someone buys it, they think the product is professional. And we want that to be fast. So, you don’t need a studio to, or a large budget to do it anymore. So this is very key and.
Like what’s more is like, it’s not even sure in the future or today, like you, that many purchase are happening, on a shopfront like on a website. A lot of that happens in marketplaces, even in chat now. So like, really the, what stays is the visual. So, it’s really, really key. You want it to look professional and brand and really high product to the real product.
Brian Thomas: Thank you. I appreciate that. Visuals are, as you mentioned, the storefront and e-commerce in shopping today and making those visuals the best they can look is really important. But I like how you take all that production time out of it, having an expert there. And we’ll get into that here in a little bit, but just amazing to see that type of.
Professional visual for a product that it’s representing will make all the difference in the world. So, thank you, Matt. Photoroom now processes billions of images annually and supports teams in more than 180 countries. What technical or organizational challenges emerge when AI creativity needs to operate reliably at that scale?
Matthieu Rouif: Yes. At scale what’s difficult is not creating or generating the photo, the, it’s kind of a scaling problem. So, what becomes difficult is maintaining speed, and that’s something that is very important, like where we work with hundreds of millions of users. But also, the biggest marketplace is like, Amazon DoorDash.
So, speed, like making sure that when you have millions of concurrent users, you maintain the speed and it’s the best speed on the market. On creativity, what’s really important is find a balance of, then the second thing is trust. So trust of your brand and trust of product. It’s quite easy to get one image to generate one image that is kind of high, productivity.
You go through a few tweaks and you get there. What becomes difficulty do when you have like tens of photos per ku and then you are selling 1000 SKU or millions of skews like a marketplace of products. And so here you need to put in place the right kind of ground rails for your AI to maintain the productivity, the trust.
Of the imagery. So, for that, we train the AI to give back how much trust they have. We build our own ai, our own generative AI algorithm. So, we don’t touch the product. We build a way to put the logos and be faithful. So, you need that special engineering solution and a machine learning solution for that.
And the last part is like, when you have scale, it’s about like having the best AI for all the distribution of like a kind of visual you can get there. You maybe. Some use case might use a, like a one foundation model, some other might need another technique. And really learning this, like for which for each use case, each country, each context, what is the best AI that is going to bring the, like the best solution.
So, for that, it’s about, it’s important to be able to e test have like a really data that tells you that’s the best image you can get. So speed, trust, and really giving the best model for the right moment, right image. These are kind of challenges that are grounded in data that are complex.
Brian Thomas: Thank you. I appreciate that.
You talked about the scaling and of course the challenge, and you talked about that as maintaining that speed, especially when you have millions of connected concurrent users. The creativity, you talked about finding that balance with the trust, the brand. And then with ai, you wanna ensure that those guardrails are in place to ensure that the product and the brand is not altered in any fashion.
Really appreciate that. There’s a lot that you have to bring it all together at scale, which is sometimes very challenging. So, thank you. Matt. Generat, AI has raised concerns about quality, consistency, and brand control. How do you ensure AI generated visuals remain accurate on brand and commercially trustworthy?
Matthieu Rouif: This is one of the challenge of imagery and visual generation. So the like, the first answer is that that’s the thing we optimize for. So you have a lot of models or apps that are optimizing for creativity. What we optimize for is like trust and consistency. And we do that by set of roles.
First, we train our models to respect the product and to know if the product is respected. So if there is like a good consistency and trust. So that’s like in the ML side, we do train specific model where we layered the product on top of like a generated background and we are special algorithm that will tell you is it respected or not.
So, we we’re the only one, like to have the 100% like product consistency and product validity on that. And the second thing is for the brands, we kind of like build that as an infrastructure where you can define your brand and then our machine learning or app reads that, read that. So like if you want to have a defined color or logo or logo, we put it on top, or the color, we put it in the background and we make sure the AI doesn’t affect that.
And so. Like this combination of building the product with the primitive around product visibility and brand and training, the machine learning for that makes us able to respect that. But you can’t. I think what’s important is you can’t optimize both for creativity and like high product fidelity. I trust and we chose as we focus on commerce to optimize for trust and fidelity.
Brian Thomas: Thank you so much. I really appreciate that. I know there’s a lot that goes into that. I know AI helps a lot, especially with processing millions of images. Probably weekly at this point. The big challenge is that it’s having that consistent consistency and trust. And then as far as quality you talked about that is training, again, using machine learning.
It’s to ensure that the brand stays consistent. Colors are on par and that sort of thing. So, I do appreciate you breaking that apart for us. And Matt, the last question of the day, as we look ahead, what do you believe will define the next generation of creative tools for commerce? And how far can AI realistically go in replacing traditional photo and design workflows?
Matthieu Rouif: So we really I mean the creative work stays. I think the, what we want is to build an AI that is equipping, like giving humans really the poor at scale of creative imagery. So, we help you scale this imagery. What’s important and how it does change is the future is customer images. So, for everyone.
So that’s like happening in a few months, a few years. Basically, every image, every, every transaction you have, every communication you have will have a specific image. You have a specific to the buyer, specific to the seller. So, buyer would be, well, you, you see the chair you’re buying in the living room.
The seller would be, the story of the seller. You have maybe their logo and then the context, like maybe soon. It’s, day or there is a spring sale. And so, all of that needs to together put together into one unique image. And so then. Is, how do you empower humans to be able to manage all of that?
So we are building on like spatial tools, spatial machine learning, spatial apps that helps you manage these thousands of AI workers for you if you are creative. So it you need to make the flow happen and make a, like, the UX to be the manager of this thousand of AI creative working for you as a creative.
So that’s kind of. The, the challenge when you build that and make sure the human workforce is getting more creative than ever, it’s making, it’s becoming more accessible. And then also understanding what creative works, what doesn’t today for the buyers and maybe tomorrow is the robots are buying.
So also, what are the images that the robots are looking for when they’re buying for you? A lot of it’s going to change, but it’s exciting and needs a lot of tools for that, that we’re building at first room.
Brian Thomas: I love that. Thank you. And I know you’re really building that AI to give humans those tools for enhanced creativity making it more accessible for everyone.
And of course, you talked a little bit about the future is those customized images again for everyone. But that UI, UX, has to be on par, as you said. There’s a lot that goes into this. And again, it’s like, a really complex recipe, but you’re. Continuing to improve and innovate with your platform.
And we really appreciate, especially as creators here in the, this new marketplace we live in with AI. So, thank you.
Matthieu Rouif: Thanks.
Brian Thomas: And Matt, it was… You’re welcome. It was such a pleasure having you on today, and I look forward to speaking with you real soon,
Matthieu Rouif: Same Brian. Bye.
Brian Thomas: Bye for now.
Matthieu Rouif Podcast Transcript. Listen to the audio on the guest’s Podcast Page.
