Google’s AI Shopping Tool Allows Virtual Dress Try-Ons in the US

Last year, Google launched an innovative AI shopping tool in Search, enabling users to visualize how clothing items would look on individuals with various body types. This feature initially rolled out with women’s tops from a range of brands. After recognizing the advantages of this tool for both consumers and retailers, Google is now expanding its AI shopping technology to incorporate an additional clothing category: dresses.

Consumers in the United States will now notice a new “try on” badge when searching for dresses on Google. By selecting a dress, users will access the virtual try-on AI feature, showcasing models with diverse body types, from XXS to XXXL. This allows shoppers to choose a model that closely resembles their own body type and virtually try on dresses to better understand the fit before making a purchase.

Once you’ve discovered a dress that aligns with your style, you can click the link to the retailer’s site for purchase.

Google has stated that they faced numerous challenges in fine-tuning this tool and extending its functionality to dresses, given that dresses cover more of the body and often feature intricate details in their draping, silhouettes, lengths, and shapes. This includes a variety of styles from midi-length halters to mini shifts to maxi drop waists, and everything in between.

It wasn’t feasible for Google to simply adapt their existing Virtual Try-On (VTO) AI model. Although the VTO AI model effectively processed low-resolution images, this method often compromised essential dress details. Additionally, switching to high-resolution images didn’t produce the desired results.

To address these concerns, the Google research team developed a new training method known as VTO-UNet Diffusion Transformer (VTO-UDiT).

Google explains that VTO-UDiT initiates with the diffusion of lower-resolution images, progressively increasing the resolution throughout the diffusion process. This technique guarantees the precise recreation of fine details like pleats, prints, and fabric textures.

Additionally, Google has partnered with SIMKHAI, allowing users to virtually try on select pieces from their runway collection following SIMKHAI’s September show in New York. Shoppers can pre-order these items from SIMKHAI’s official website.

Source