Decoder. plain-English AI glossary

Diffusion Model

● Core

A model that learns to make images by starting from pure static and cleaning it up step by step.

Think of it like

Like watching a photo develop in a darkroom tray, except it runs backwards from noise toward a picture.

Example

When you type a prompt into an image generator and watch a fuzzy blob sharpen into a cat over a few seconds, that’s a diffusion model at work.

How it actually works

Training teaches it to predict the noise that was added to a real image; run that predictor in reverse many times and you turn random noise into a coherent picture. Each step removes a little noise, guided by your text prompt. It’s slower than a single forward pass, which is why image generation takes seconds, not milliseconds.

For product teams

The reason text-to-image tools feel magical and also why they cost real compute per picture.

For engineers

A generative model trained to reverse a fixed noising process, sampling iteratively from noise to data.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome