Diffusion model
Also known as: diffusion
A diffusion model generates images (or audio) by learning to reverse a noising process. Training: gradually add noise to real images until they’re static. The model learns to undo that, step by step.
Generation is the reverse: start from pure random noise and denoise it into an image, guided by your prompt.
This is the engine behind DALL·E, Stable Diffusion, and many image tools.
The contrast with LLMs: LLMs build text one token at a time; diffusion builds images by denoising noise. Different recipes, same goal — generate something new from a prompt.