AI glossary
Diffusion model
A type of generative model that creates images (or video and audio) by starting from random noise and removing it step by step.
During training, a diffusion model learns to reverse a process that gradually adds noise to images. To generate, it starts with pure noise and denoises it in many small steps, guided by your text prompt, until an image emerges.
Many well-known image and video generators are based on diffusion or closely related techniques.
Example: You type “a lighthouse at sunset in watercolour style”. The model starts from an image of random noise and, over dozens of steps, cleans it up until a watercolour lighthouse appears.
In practice
- Describe the subject, style, framing and light: the prompt guides every step.
- Generating several versions and choosing one often works better than polishing a single prompt.
- Before using a generated image commercially, check the tool’s terms; in the EU, some generated content must be marked as such.