It's been nearly eight years since Zalando first began experimenting with virtual try-ons.
In 2018, the German e-tailer began developing its own software that can generate photos of customers wearing products from its catalog. However, the advent of large-scale open source language models has changed the game, allowing Zalando to improve its image generation capabilities and begin rolling out virtual try-ons to customers. It began testing in small groups in 2022 and gradually added new features, such as a measurement tool that can create images of a person's body shape from photos or 3D rotations. The company is also seeing benefits during the testing phase, with a test in April 2023 showing a 40% drop in profits, said Reza Shirvany, Zalando's director of applied sciences.
Zalando is now ready to launch virtual try-ons for all customers in 2026.
Generative AI has “reduced many of the barriers for companies, including ours, to innovate for customers,” Silvany says. “As the world invests in LLM, there is less and less scientific reason to build this kind of technology in-house.”
Online retailers have long hoped that virtual try-ons will help reduce return rates, but the technology often fails to accurately represent how clothing will look on a real person. But generative AI is bringing that possibility closer to reality, making try-on images more photo-realistic and showing how a particular fabric or cut might sit on different bodies.
These advances have given way to a new cadre of AI-powered virtual try-on startups, including Doji, Zelig, and Stiled. Software giants are also betting big on the technology. Starting in 2023, Google will expand the number of products in its shopping catalog, allowing users to see how they actually fit on digital models in sizes from XXS to 4XL. In May, the company released a feature that lets users upload photos of themselves to see how they would look wearing the apparel products on offer, and this month it introduced a new feature that lets users generate try-on images from selfies. Virtual try-on is now the number one Google Shopping tool shared by users on social media, said Lilian Rincon, Google's vice president of product management.
“Generative AI is mainstream. Everyone is using it and starting to trust the technology,” said Jen Jones, chief marketing officer at e-commerce software company Commerce Tools. “People who would never have uploaded their selfies to a site like this… [Google’s] A Gemini two years ago would have done it without a second thought today. ”
However, there is still a way to go before mass adoption. An October survey of nearly 1,000 adults between the ages of 18 and 65 by eMarketer found that only 1.4% regularly use some form of virtual try-on.
To make this technology an essential part of e-commerce, software startups are increasing the speed at which images are generated, developing advanced features such as 3D images and videos, making images shareable on social media, and convincing more consumers that the technology is fun and easy to use.
“Virtual try-ons are going to get better and better in terms of image fidelity and personalization,” said Amy Wu Martin, an investor at Menlo Ventures. “What matters is how well it translates into the user's desire to buy clothes.”
How generative AI is changing virtual try-ons
Until now, virtual try-ons have often required users to scan their entire body and render an image or create an avatar that closely resembles their real-life appearance. Generative AI models, on the other hand, allow users to upload images of themselves and instantly see how the clothes will look when they are “trying on,” while also taking into account important features such as “fabric nuances,” Google's Rincon said.
These improvements allow companies to accurately display garment details such as T-shirt graphics or the shape of a blazer collar. This means retailers can use try-on software even in categories that are difficult to render, such as denim, which can accommodate up to 100 sizes.
“We are no longer in a world where this is done through augmented reality. Gen AI is creating something that feels just like you in the real world,” said Juan Pererano Rendon, chief marketing officer at software company Swap. “That's going to create more demand.”
Gen AI is also helping the virtual try-on startup reduce internal development costs. In March, Zelig, a four-year-old company that develops try-on technology for retailers like Revolve, began combining open-source LLM with proprietary software that captures the natural movement of fabrics. I want you to imagine the knot at the top of your pants. Sandy Sholl, Zelig's founder and CEO, says the company was able to create these images by fusing existing models with generative AI, saving the company millions of dollars.
“The way we did it wasn't scalable,” Scholl said.
Software providers also use LLM to experiment with complex features such as image and video rotation. When Swap launched in September to help brands create their own AI agents, it included a try-on tool that generated both still images and videos of users turning around in clothes, Pererano-Rendon said.
“We make sure that the fit of the clothes and natural body movements when turning or walking are the same as when trying on clothes,” he added.
The future of virtual try-on
However, all these improvements won't mean much if people don't actually use virtual try-ons in real life.
To attract consumers, some startups are reinventing it as a styling tool rather than a return solution. For example, on Doji, which launched in 2024, the app's most used features allow users to shuffle through outfits and receive style and product recommendations, said Dorian Dargan, Doji's co-founder and chief executive officer.
“Virtual try-on allows people to dip their toe in the water and really build an emotional connection with the brand,” Dargan said. “If you focus primarily on revenue and conversions, you can miss out on opportunities for inspiration and exploration.”
Instead of waiting to partner directly with brands, some virtual try-on providers are extending their reach through browser extensions. Doji released an extension in Apple's Safari in November to allow users to generate images featuring more of the brand's products on the internet. Stiled, an AI-powered shopping app, overlays products from any brand or retailer's site onto your images. Will Hanson, the company's founder and CEO, said it is hoped Styled will be able to use these results to persuade brands to partner with them.
“You're basically thinking about how to make [virtual try-on] Wu Martin said, “It's a daily habit. In the end, it's about seeing more of it.”
The next challenge is to prove to brands that virtual try-ons are a clear revenue stream. Zelig's Sholl says it will be important for software providers to partner with brands and organize data to directly attribute sales increases to software trials.
“The real change will happen when everyone realizes that this technology is going to make a lot of money for everyone,” Scholl said.
