Skip to content

Control‐LoRAs

Jump to bottom
MoonRide303 edited this page Sep 12, 2023 · 5 revisions

Control-LoRAs (available in CN tab, as they work as ControlNets) are additional models that can guide image generation process with information acquired from the input. Canny relies on edge detection, Depth uses depth information. Control-LoRAs can work in both normal generation (Text-2-Image), and in Image-2-Image mode (using pixels from input image as starting point for generation).

UI for Control-LoRAs:

image

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 also enabled Canny):

image

Sample output #3 (same as #2, but reduced Canny Strength to 0.2):

image

Sample output #4 (prompt: "perfect supercar", seed 0, resolution 1248x832, disabled Image-2-Image, enabled Canny, Canny Stop 0.5, Canny Strength 1.05): image

Sample output #5 (prompt: "perfect supercar", seed 0, resolution 1248x832, disabled Image-2-Image, disabled Canny, enabled Depth, Depth Stop 0.4, Depth Strength 0.8): image

Sample output #6 (prompt: "perfect supercar", seed 0, resolution 1248x832, disabled Image-2-Image, enabled Canny, Canny Stop 0.25, Canny Strength 0.5, enabled Depth, Depth Stop 0.25, Depth Strength 0.5): image

Clone this wiki locally