Diffusion model — Generative AI glossary — LiliDi
A generative model that produces images or video by iteratively denoising random noise.
By LiliDi Editorial
A generative model that produces images or video by iteratively denoising random noise. Diffusion models start from random Gaussian noise and iteratively predict a slightly less noisy version, conditioned on a text or image embedding. After 20–50 steps the noise has resolved into a coherent image or video. Stable Diffusion, Flux, Nano Banana and Veo are all diffusion based.