The professionals way of creating images.

How to use regional guidance

Author

Nejc Susec
June 11, 2024 • Estimated Reading Time: 5 minutes

Hello Friend

Welcome to this week’s lesson on AI workflows, techniques, tools and everything else that will help you use AI to enhance and elevate your work as a creative.

Today we will be looking at images. More specifically, how to use regional guidance to influence the generation of specific elements in specific areas in the image.

Let’s get into it!

Invoke has been my go-to Image generator for probably a year now. One of their recent updates confirmed why - they are building their tool with the creatives in mind. The people who are experienced and will be using it in their day to day work.

The update I’m talking about introduced Regional Guidance layers.

This might not tell you much, so let me show you instead. 😁

This is an image I made recently:

prompt: sunrise above the clouds at the top of a rocky mountain

The prompt by itself might be enough. But for me, the goal is often to go from the idea I have to the image I want in as little iterations as possible.

If you’ve tried generating an image before, you have inevitably come across writing a prompt and pressing the button over and over again. Hoping it will give you the shot that you have in mind.

Well, regional guidance can help us shorten this process.

We can draw a mask over an area of our image and use part of our original prompt to guide AI what to generate in that area.

Three different masks for three different elements in my image.

On an empty canvas, I can choose the areas and I can choose the elements that will be generated inside those areas.

When you have a concept and vision in mind, this is an incredible tool - being able to communicate what should be and where it should be.

More than that, regional guidance is useful for adjusting existing images as well.

As an example, I wanted to modify this image:

I uploaded it into controlNET with canny and added a regional guidance to the face area. My goal was to add a specific face to this character.

ControlNET with canny

Regional guidance with IP adapter. Using the image as reference.

Canny + regional guidance

Final image

Regional guidance unlocks plenty of new ways to control the AI generation of images.

If you see visual ideas in your mind, give regional guidance in Invoke a try. I’m sure you will be amazed at how quickly you can go from idea to image.

If you enjoyed this post here are a few other lessons on image generation:

Whenever you’re ready, here are three ways I can help you: