By using Stable Diffusion, the company is able to train the model based on the user's profile picture, enabling high-quality visual outputs.