American Internet users who are looking for a female top on Google Shopping can now view it on a model with their morphology and their hair and skin color. This tool based on generative image AI will be extended to other wardrobe pieces in the near future.
E-commerce players are beginning to take advantage of the tools offered by generative artificial intelligence, capable of producing text, images or sounds on demand. After Amazon, which uses AI to summarize the comments of its customers on products, Google in turn presented on June 14, 2023 a feature called Try on: the Internet giant wants to allow Internet users to view clothes on hundreds models with different sizes, weights (from XXS to 4XL), skin and hair colors. The objective: to find a model that looks like them to find out if these burgundy velvet pants would fall well on their hips or if the color of this fuchsia jacket would suit their complexion. Several images are available for each garment. The models take different poses, which allows the user to imagine how his crop top would behave if he raised his arms in the air.
Solve the issue of product returns
This feature is already active in the USA for women’s tops from a few brands, such as H&M, Anthropologie, Everlane, and even LOFT. It will gradually be offered to other companies and extended to other pieces of the wardrobe, starting with shirts and other men’s T-shirts, which can be scrutinized by e-consumers from every angle thanks to Try on by the end of 2023. This tool will not fail to interest e-merchants who promote their clothes on Google Shopping, because it could allow them to limit the costly returns of their products (in France, 30% of clothes purchased online are on average returned). Google does not give details on its cost for the moment.
This issue had been identified by start-ups such as the Israeli Zeekit, acquired by Walmart, or the French Veesual AI. Over the past few years, they have developed virtual fitting tools, based on AI or augmented reality, but which sometimes lacked realism. Generative artificial intelligence comes to solve this problem: the Try on models are very real, but the images of them wearing the clothes are generated by algorithms and seem more real than life, as shown in the screenshot below . They were trained internally by Google from two tools for automatically generating images from text, DALL-E 2 (created by the Open AI company, which is also behind the text generative AI ChatGPT) and Stable Diffusion (developed by the company Stability AI). Next step ? Allow the customer to post a photo of themselves to Google so that the product is projected directly onto their body.
