The stable diffusion image to image robbed me of 70 sats lol I did prompt "smiling" and this image link: https://primal.b-cdn.net/media-cache?s=o&a=1&u=https%3A%2F%2Fnostr.build%2Fi%2Fnostr.build_921103a346a15dd528b2f2e56db0d30bac05dfd6604d7a2873d474b718621a58.png maybe it just needs to validate the image before payment.
It can't actually do image manipulation in the sense of giving it commands what to manipulate. I can add something for that (that works okayish). This DVM is more for style conversion. You can say "in style of Picasso oil painting" (or whatever) and it will adapt the style of the image