1) Get a prompt to use in Stable Diffusion and a model from civitai.com
https://lexica.art/ is an excellent resource for finding great prompts.
2) Use Text2Image and generate 12 images.
3) Select one of the 12 images and use Img2Img (inpaint) to refine the image by adding or removing elements.
4) Generate an additional image or two using Img2Img which will clean up the chosen image.
5) One the extras tab upscale the image the the desired final size.