How do I create a stable diffusion dataset?

I know it’s a stupid question, but I can’t figure it out: how do I take a dataset of images that I have and create a text-to-image space based on that?