get stable diffusion running in <10 minutes in colab:
this notebook contains:
- the absolute minimum code needed to generate images with stable diffusion
- tips for writing good prompts
currently this notebook only works on colab and not when run locally; a pr changing this would be appreciated
stable diffusion is a text-to-image model similar to dall·e 2; that is, it inputs a text description and uses ai to output a matching image
for instance, the following image was generated with stable diffusion using the prompt gallant thoroughbred, a surrealist painting by Andy Warhol, mystical, ominous:
it is generally considered to be of similar quality to dall·e, but is:
- more computationally efficient
- available to the public (ie can be run locally, not just through openai's playground)
a more in-depth explanation of stable diffusion's architecture can be found in this notebook or here
