To resolve the issues presented in the introduction, we will use the new features of Panning & Zooming (also called out-painting) and Region varying (also called in-painting), along side the classic image prompting.

Crucially, we will leverage a technique called “Character Expression Sheets” to allow Midjourney to essentially store a working-memory of our Character while we do these edits.

This technique works because when we are in-painting, or panning an image, Midjourney (at least for now) looks at the current image and the prompt we’ve supplied (if using remix) to try and match the existing image’s style.

So this means that if the image is only a single character but with multiple versions and poses, Midjourney infers that it should try to re-create this character in the new region that we are in/out-painting!

Now this doesn’t just magically work perfectly every time. And although this image above is a decent demonstration of the technique in practise; it won’t always work so well unless we take steps to guide it in the right direction. That’s what I will explain.

Let's Get Started!