Stable Diffusion

← Back to the article

Stable Diffusion starts with messy dots and slowly cleans them into an image that matches a prompt.

What it does

Stable Diffusion starts with messy dots and slowly cleans them into an image that matches a prompt.

Denoising: Cleaning static • Hidden code space: A small number-map the AI uses

How it works

  • Start with random noise.
  • Use the prompt as a guide.
  • Clean the noise step by step until an image appears.

Uses and applications

  • An artist can make a quick idea picture from words.
  • A designer can try many poster concepts.
  • A game maker can explore character or background ideas.

In short

It is like finding shapes in clouds until a picture appears.

Related methods

  • Diffusion: The noise-removal process that Stable Diffusion is based on.
  • VAE (Autoencoders): Compresses images into the latent space where it works.
  • CLIP: Connects the text prompt to the image.

Back to Image and content generation · All topics

The interactive demonstration and the Python walkthrough for this topic need JavaScript.