What is Diffusion Model? Plain-English Definition

In more detail

A diffusion model is the technology behind most AI images. It starts with a canvas of pure random noise — television static — and removes the noise in steps, each pass nudging the chaos a little closer to a picture matching your description. Order sculpted out of static.

It matters because it made images conversational: type a sentence, get a picture. The same start-from-noise-and-refine idea now powers AI video and audio too. Whenever you hear about AI art, a diffusion model is almost always the engine underneath.

📌 See it in action

You type ‘a corgi surfing a wave at sunset, oil-painting style.’ The process begins as pure static; a shape emerges, becomes dog-like, gains a surfboard, then golden light. A few dozen refinement steps later there’s a painting that has never existed before — denoised into being from your one sentence.

Goes with

Text-to-Video →Multimodal →

Now put the vocabulary to work.

Meet the models →Which AI is mine?