It's based off a model - not an actual person but a trained AI model that responds to certain trigger words. So you're not taking an existing picture and adding to it, but telling AI what to do with a trained model and it creates it from nothing.
Full disclosure, I did not use one of my models as I'm still figuring it all out. This was made with https:// civitai. com/models/139562?modelVersionId=789646"].

