How does AI create images?
Generative image models create pictures from text, images or other inputs.
What is it?
Image models create a new picture from a text description or another image. Most of them gradually “draw” an image out of noise to match the description. The result can be photorealistic, painterly or schematic.
An everyday example
From the description “an illustration of a fox reading a book, in the style of a children’s picture book”, an app produces several versions in seconds.
Why does it matter?
Image models are changing design, advertising and illustration, and raise questions about copyright and fake photos.
A common misconception
That a photorealistic image shows a real scene. Looks are not proof that anything happened.
In more detail
How did the model learn about images?
From a huge number of examples and mathematical procedures that link visual patterns to descriptions.
What is a diffusion model?
An important family of models that learns to produce a meaningful image from data that has been gradually scrambled with noise.
Why do images sometimes have strange details?
The model follows learned patterns and instructions, not necessarily strict physical or anatomical accuracy.
Is it a photograph?
A generated image is not a documentary record of an event, even if it looks like a photo.
Remember: Photorealism is not proof that the scene happened.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.