Online shopping is big business: 71 percent of Dutch people order online. In almost half of the cases, it involves clothes or shoes. Ordering clothes online, however, is not without its challenges. For consumers, it is difficult to judge whether that particular blouse will turn out as beautiful as the model in the photos – who often does not look anything like them. Consequently, in 2021, 34 percent of online clothing orders were returned. That’s 7.8 million returns. Startup Lalaland, armed with their AI software, aims to do something about this.
The mission of Lalaland co-founders Ugnius Rimsa (24) and Michael Musandu (23) is clear: improve online shopping and help companies become more diverse, inclusive and sustainable. And they do so with lifelike virtual mannequins of all ages, sizes and ethnicities, generated by Artificial Intelligence. Ugnius and Michael’s mission is being picked up – Wehkamp and Stieglitz are already customers – and rewarded. The men were named best young entrepreneurs in the Netherlands by MT/Sprout.
AI as an art detective
The short description that Michael uses to explain to his grandmother what he is so busy with is that at Lalaland they “create people.” For a slightly more sophisticated explanation, Ugnius compares Lalaland’s AI software to an art detective. “Imagine a painter and a forger who both send art out into the world. Everything goes past an art detective who evaluates the work for authenticity. The only information the forger gets is feedback (fake/original) from the art detective. After a few million times, the forger becomes so good that original and fake are indistinguishable even by the most skilled art detective,” says Ugnius.
This is exactly what Lalaland wants to do with its AI software: generate photorealistic models that are indistinguishable from real ones. Developing an algorithm that creates models that are just as diverse as the world’s population also immediately presented the biggest obstacle. “Collecting a dataset that represents the full range of ages, body types and ethnicities was difficult. Data for minorities was not available or simply did not exist.”
Ugnius compares the AI algorithm to an engine that you have to supply with the right type and amount of fuel. “If you have a million cat pictures, AI can create the perfect, realistic cat. But if you don’t have enough images, gaps appear.” To eventually develop a representative algorithm and fill in the gaps, they used a technique by which they superimposed certain minority characteristics over the images of standard models. “That style transfer technique can be compared to a subtle snapchat filter. It allowed us to balance the datasets,” Ugnius explains.
Models without diva behavior
Lalaland solves several problems. For example, the model landscape is not nearly diverse enough now, Ugnius says: “In 99 percent of the cases, it’s young, slim, white women who are modeling clothes.” Lalaland wants to put an end to this one-sided representation with their software. “It doesn’t make sense that someone with size XL buys clothes based on a model with size XS. That doesn’t say anything about how a product would look on the customer, does it?” In the online shop of a company that collaborates with Lalaland, you tick your own size, after which a model with the same size appears on your screen.
It doesn’t make sense that someone with size XL buys clothes based on a model with size XS.Ugnius Rimsa
It is now too costly for companies to create more varied campaigns themselves. “A campaign with one model is already expensive, let alone hiring six. Our software can show all the ethnicities, ages and sizes you want with the click of a button. And our models don’t exhibit any diva behavior, haha.” Thus, Lalaland ensures that customers can better assess how the garment will look on them in real life. Say goodbye to those few eight million returns. Ugnius: “In addition, 80 percent of all returns are not sold again, but destroyed. Together with all the extra trips made by the postal workers, this has an enormous impact on the environment. We want to put an end to that, too.”
With this vision, Michael and Ugnius have been declared the best young entrepreneurs of the Netherlands. They haven’t really had the time to let the realization sink in. “We are so busy that everything that is not directly related to our end product fades into the background somewhere. But after winning this prize, I am really well integrated into the Netherlands,” laughs Ugnius, who is from Lithuania.
Lalaland does not lack plans for the future: the online shopping experience must become even more dynamic and personalized. For example, by uploading a selfie and submitting your measurements, after which you see yourself walking around the screen as a kind of avatar wearing the garment. “We notice that companies often don’t really know what their customers want. Our ideas are new and we are still discovering what the market wants. Some people would rather not see themselves online, but from a technical perspective it’s a challenge we’re willing to take on. So we have to find a balance between what the market wants, what customers can get used to and our groundbreaking ideas.”
There is no way Ugnius and Michael can make those plans a reality just by themselves. In 2020 they hired their first employee; by now the team consists of eighteen people. “In the beginning, managing a team seemed especially difficult to me, but I have since learned that they are there to help. It’s not just Michael’s and my product anymore, it’s theirs, too. We work as a team towards one goal: making Lalaland a better service,” Ugnius says. At Lalaland, anything is possible – hence the name. “Lalaland is a country where anything goes, and we are a startup where anything goes. Our software and algorithm can make anything. And we contribute to a more inclusive and sustainable world.”
Also interesting: Key to XR’s success is rich and eventful content and experiences