Diffusion Model

A class of generative AI model that learns to create images, audio, or video by reversing a noise-adding process. During training the model learns to denoise progressively; during generation it starts from pure noise and iteratively refines it into a coherent output. Stable Diffusion and DALL·E 3 are prominent examples.

“The diffusion model generated photorealistic product photos from text descriptions in seconds.”