Image Generation Basics

Image generation turns a text prompt (and sometimes a reference image) into new pixels. Modern systems are usually diffusion or related generative models, wrapped in product UIs with safety filters and style controls.

Intuition

You describe “a watercolor fox under streetlights.” The model starts from noise (or a latent) and iteratively denoise toward an image that matches the prompt embedding. Editing tools (“inpaint this region”) constrain which pixels may change.

Practical knobs (product view)

What goes wrong

One-line summary

Text-to-image models synthesize pictures from prompts; treat them as creative tools that still need safety and rights review.

Key terms