CompVis/stable-diffusion ? reverse-engineered prompt

I want this repo turned into an easy local text to image generator I can actually use. Please wire it up so I can install it, add the model checkpoint, and run one simple command to generate 512 by 512 images from a prompt like “an astronaut riding a horse.” Use the provided Stable Diffusion v1 code and keep the sensible defaults from the project, including the normal guidance setting, around 50 sampling steps, the safety checker, and the invisible watermark on saved images.

Also make it easy to save results into an output folder, set a seed for repeatable results, choose how many images to make, and optionally read prompts from a text file. If the machine has a supported GPU, use it, otherwise fail with a clear message about the requirements. Please include a short README with exact setup and run steps, and look up current docs online if you need to.