Image-to-image: when do you stop trying to preserve the original?
A rough sketch and an almost-finished image seem like very different starting points for image-to-image. With the sketch, you might want to keep the layout but invent the surfaces; with the finished image, the details might be the whole reason to use it. The Diffusers guide describes strength as controlling how far the output departs from the starting image: lower values stay closer, while higher values allow more departure. Before touching that control, how do you decide which parts of the starting image are essential? References: https://huggingface.co/docs/diffusers/using-diffusers/img2img
Shared by aperture.wild on Prompt Builder.