Stable Diffusion is an open text-to-image model that creates pictures from written prompts, and was light enough to run on a single graphics card with at least 10 GB of memory.
In plain words
Earlier diffusion models worked directly on every pixel, which took huge amounts of computing. Stable Diffusion instead works in a compressed version of the image, called a latent space, then decodes the result into a full picture. A text encoder turns the prompt into guidance for each denoising step. The original model was trained on images from a subset of LAION-5B and needs a graphics card with 10 GB of memory or more.
A quick example
Type "a lighthouse on a cliff at sunset, watercolour". The model starts from random noise in its compressed space and removes the noise step by step, guided by the prompt, until a matching painting-like image appears.
Where you'll meet it
- Turning text prompts into images on a single graphics card.
- Research on faster, cheaper latent diffusion models.
Related
Sources
- PaperHigh-Resolution Image Synthesis with Latent Diffusion Models, Rombach et al. (arXiv) · read 29 Sept 2026
- Docsstable-diffusion, CompVis (GitHub) · read 29 Sept 2026
- DocsStable Diffusion Public Release, Stability AI · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.