Diffusion Model

Generative architecture behind DALL·E, Stable Diffusion, and Midjourney.

Diffusion models generate images by starting from random noise and gradually refining it into a coherent image, guided by a text prompt. They represent the state of the art for high-quality visual generation.