How AI makes virtual try-on more realistic
Google - Ira Kemelmacher-Shlizerman
Google has developed a new AI model that enhances virtual try-on (VTO) experiences for apparel. The model uses diffusion, a process that gradually adds and removes pixels from an image, to generate high-quality, realistic depictions of clothing on different body shapes and sizes. Unlike previous techniques that produced imperfect results, the new model creates images that realistically adapt to the body, accounting for details like draping, folding, stretching, and wrinkling. By training the model with data from Google's Shopping Graph, the VTO feature can render tops on real models, offering users a more accurate representation of how the clothing will look on them.