Motivation

  1. Image editing. Diffusion model can only generate a good picture directly from a prompt. It cannot edit some very detailed part of the prompt. If I want to change the color of the T-Shirt