Haunted House

Oct 9, 2022

I decided to dive into the image to image feature of stable diffusion, which allows you to start with a simple input image, and then use a text description to have stable diffusion hallucinate your image into one that more closely matches the text description.

The theme this week was an appropriately Halloween theme of a haunted house on a hill with some jack-o-lanterns. For this exercise I took the final picture that my wife painted and then fed that as the input into stable diffusion to see what we could get as a result.

Overall I think the image to image (img2img) feature is pretty impressive, I was able to get some pretty awesome results. At this point I’d venture that the img2img feature might be the most impressive aspect of SD so far.