How AI Will Revolutionize Fashion Retail And Virtual Try-On

It’s impossible to switch on your laptop or even glance at your phone without hearing about how AI is going to change everything. How we work, how we make things, how we can meet a virtual partner, even how we will treat cancers in the years to come. But let’s tackle something more manageable, everyday and — dare I say it — fun. How will it change the way we shop for clothes? By using data to drive retail interactions, the future could involve more targeted, compelling, and personalised interactions and better-priced clothes to match our personal style, mood and wallet. There is also the potential for more efficiencies around sustainability.

With e-commerce becoming the fastest-growing segment in the retail market in Europe (up by over 116% in Western Europe), online retail is an area that it is primed for an AI tune-up. An early experiment in this field in 2020 allowed me to work alongside a number of collaborators to develop a method of scanning your own face and body at home, using a mobile device, to create a virtual version of yourself for online shopping. This virtual “stylist” would then make relevant outfit suggestions for each day based on your calendar, location and even the weather forecast.

Interaction with the stylist took place via your mobile device, tablet or a mixed-reality headset. It was very much a proof-of-concept project with a few edges that needed smoothing out, not least the “conversational” interaction with the AI model making the suggestions. Despite not being fully ready for the general public, the project piqued the interest of brands, retailers and technology companies alike, all eager to see technologies like this unleashed on the world.

So what has changed in the last few years to make interactions like this more possible, but also, more believable and natural? How can we move away from the “you might also like these items” suggestions and the often-frustrating chatbot interactions on retail websites?

This is where artificial intelligence steps in — not the scary, take-over-the-world sort, or the will-my-job-be-here-in-five-years kind, but the integrated, useful and empowering kind. According to research by Accenture, AI is predicted to lead to a 59% increase in profitability for the retail sector by 2035.

Let’s start with the idea of speaking as a method of communicating with an online or virtual stylist. Many of us have used ChatGPT to fill in an application or create a presentation, but large language models like this have developed dramatically over the last few years. Generated responses are now difficult to distinguish from human writing or conversation and models retain information for longer periods and across multiple interactions. These advances make communication much more seamless, personalized and “natural” feeling for the customer.

Better still, these new LLM models can receive multimodal inputs, meaning interactions with your virtual stylist are no longer limited to verbal communication. You could show your “stylist” images of a venue, a new item in your wardrobe, a TikTok video, or the tone of your skin after a recent tanning session — anything that could provide additional context for its styling recommendations.

Moving on to scanning yourself, or someone else, using a mobile phone to create your stylist, or scanning garments from your wardrobe, advances in AI have also created a step-change in the generation of digital assets. New scanning software benefits from real-time AI feedback during the scanning process (based on factors such as distance, lighting and textures), allowing the customer to make any adjustments to ensure that the output is as accurate and successful as possible. Using AI to clean up these digital items will drastically increase the speed and the accuracy of photogrammetry scans (the technique of combining multiple photos into one photorealistic digital model), as well as improving the appearance of scanned textures and color consistency across scans.

Other techniques, such as the amusingly named “Gaussian splatting,” let users scan people and items from mobile devices at much higher fidelity, incorporating complex textures and finer details (such as different hair types). They can also scan reflective objects (mirrored shades, shiny dresses) and semi-transparent surfaces (like tuille or gauze), as well as traditionally matte materials and completely transparent materials. These are issues that have baffled the scanning industry for years, but new possibilities are now upon us.

Recent research from Shopify has shown that presenting accurate digital models on a retail website can significantly increase online sales conversion (by a staggering 94%) and customers can make more informed purchase decisions, leading to a reduction in online returns.

U.K. department store Selfridges recently announced aims for half of its customer activity to be centered around resale, rental, repair or refilling by 2030. For those of us looking to indulge in resale or rental of items from our wardrobes, new capabilities of AI models allow for “visual recognition”, evaluation and classification of scanned items, so that we no longer have to spend hours trying to create an accurate description for resale or rental marketplaces of items bought long ago, with a style name or silhouette that we can no longer remember.

Now back to the data. One of the biggest problems facing fashion today is overproduction — mountains of unsold goods being sold at hugely discounted prices, or worse still ending up in landfill (57% of fashion waste ends up in landfill). With more information available than ever before about our online shopping habits, and years of purchase history alongside that, AI models can use this data to respond to consumer demand, increase profitability and reduce waste.

Sharing sales data with brands and manufacturers can help AI models make better predictions so that we can create products at the right time with the correct pricing to meet public demand and in appropriate volumes, thereby reducing overstock. No need to put unwanted garments on sale or destroy them or put them in landfill.

There are a number of companies working on these solutions right now with aims of more personalized interactions, more accurate images or models of garments, and more ways to simplify the circular economy and tackle fashion’s overproduction problem. Fashion is ready for this AI revolution and the industry will be in much better shape for it.