← Concept Index

Multimodality & generation

Diffusion & image generation

Also called: text-to-image, denoising

DEFINITION

Most image generators work by diffusion: they start from random noise and remove it step by step, steered by your text prompt, until a coherent image remains.

WHY IT MATTERS

It explains why prompts act like a brief rather than a command, why fine details (hands, text, counts) are unreliable, and why two runs of the same prompt differ. You're guiding a denoiser, not describing a photo to a person.

COMMONLY CONFUSED WITH

Collage or search. The model isn't stitching found images together; it's generating pixels from noise under the prompt's guidance.