Skip to content

Image‐2‐Image

Jump to bottom
MoonRide303 edited this page Sep 3, 2023 · 3 revisions

Image-2-Image mode allows you generating image using existing image as a base. It uses image loaded into Input tab. If more than 1 image is loaded into Input, then Image-2-Image will iterate over them when generating images.

If you want output image to look more like input, then you can increase Image-2-Image start step and/or decrease Image-2-Image denoise.

Sample input (quick MS Paint drawing):

image

Sample output #1 (prompt: "perfect supercar", seed 0, enabled Image-2-Image):

image

Sample output #2 (same as #1, but Image-2-Image start step 0.1, denoise 0.9): image

Sample output #3 (same as #1, but style: "Watercolor 2", base CLIP Skip: -4, scheduler: exponential):

image

Sample output #4 (same as #3, but Image-2-Image start step 0, and denoise 1.0): image

Clone this wiki locally