Image-to-Image (img2img)

Image-to-image ("img2img") is a diffusion workflow where you seed the generation with an existing image instead of pure noise. The model respects the composition of the input image to a tunable degree while applying the prompt's style instructions.

This is how you do things like "make this photo of my dog look like a Renaissance oil painting" or "turn this rough sketch into a polished illustration." The more you weight the denoising toward the original image, the more it preserves structure; the less, the more it reinvents.

Img2img is the key technique behind disciplined AI art production. It lets a human art director anchor the generation to a reference — a photo of the actual subject, a hand-drawn composition sketch — rather than hoping a text prompt alone will produce the right arrangement.