Latent couple guide

Basics

Think of it like slicing up a pizza.

Divisions: First decide what kind of slice you want. If you wanted to slice out the middle piece for example, you would need to divide the pizza into thirds. 1:3 means 1 division parallel to x-axis and 3 divisions parallel to y-axis.

Positions: Positioning is like a coordinates system starting with zero (0, 1, 2). If you want the middle slice, you pick 1. So 0:1 means "the 0 x division and the 1 y division." Remember we only had 1 division on x and 3 on y, so 0 means the one and only x division and 1 means the middle y division.

Weights are assigned in order of positioning which determine how strong your prompt toppings are on that slice.

The visualizer is useful:

The following jsfiddle demonstrates this https://jsfiddle.net/usw9jfmh/

Full example

Combine with ControlNet

The first image in this rentry was made with ControlNet canny on the following image with default settings.

"End at step"

The "end at this step" works like prompt editing.
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#prompt-editing

You're deciding at what sampling step the latent couple ends.