Seems reasonable. I assume the future version of this is generating a model based on the user - either client side with a local model or maybe you have a vector that your browser shares that represents what you want to look like and websites or ads can incorporate your (preferred) appearance into their image generations.