Image to Image
Generate from a reference image instead of a blank canvas — results follow your upload's look and layout.
Image to Image lets you start a generation from an existing picture instead of from scratch. Upload a reference (or click Use as Base Image on any of your results), and the AI creates new images that closely follow its composition, colors, and subject — steered by your Prompt.
What it's great for
Remixing your own results — keep a composition you love and change the style or details.
Style transfer — like an XL model's accuracy but not its aesthetic? Generate a base with XL, then re-generate it with your favorite regular model.
Guided edits with Paint — brush over just one part of the reference (say, the eyes), type what should change ("blue eyes"), and get a nearly identical image with only that part different.
The key setting
Denoising-Strength controls how faithful the result is: low values stay close to the reference, high values treat it as loose inspiration.
See also
Inpainting — surgical edits to a finished image
ControlNet — borrow only the structure of a reference (pose, outlines)
Image-to-Prompt — extract a text prompt from an image instead