Code Clinic | Finding your Seed in Stable Diffusion

A short python script that runs a locally hosted model on a MacOS Macbook Air (M1)

Stable Diffusion is a really powerful image generator. Yet, one of the problems I encountered time and time again was finding the right seed.

Most UIs randomize it and I found this really annoying. Just for the record.

Here is an example from stable diffusion how you can work with a prompt if you have found a good seed.

Besides the prompt and the model, finding the right seed is really powerful because it helps to ensure reproducible results. And if you want to do actual work with the images you have created, then finding the right seed should be a core component of your workflow.

If you are really in a hurry, here is my final code on GitHub, the software is provided as is.

So how can you find the right seed?