How to create multiple consistent characters in Midjourney?

In my last post, I showed you how I create consistent characters in Midjourney.

A reader then asked me if it was also possible to create images featuring two different people, for instance, in a conversation together, based on two distinct character references (Crefs).

This is a valid question, especially when I want to use the feature to create visual stories.

So, what happens when I use two —crefs in my prompt?

Unfortunately, I get an error message as soon as I press Enter.

Does this mean that I can only create one consistent character per image with Midjourney?


There’s actually a workaround that allows me to achieve this. Let’s take a look at how this works in detail!

Let’s create two or more consistent characters in Midjourney

For my example, I want Albert Einstein and Steve Jobs to appear together in one image. For this purpose, I downloaded two reference images from the internet:

Next, I created the first generated image with the following prompt:

cinestill 800T film, Steve Jobs walking through Apple Campus, close up cw 20 –cref –v 6.0 –ar 2:1

And here is one of the four results (I only present the best one here, which I continue to work with):

You are already familiar with these steps from my last post. But what can I do now to incorporate Albert Einstein into this image?

As far as I’ve discovered, there are two ways to do this:

1. Vary (Region)

You’re probably already familiar with the Vary (Subtle) and Vary (Strong) commands, which allow you to regenerate upscaled images. Now, there’s also the Vary (Region) command, which enables you to specifically alter areas of an upscaled Midjourney image.

What happens if I use this option to integrate Einstein’s image into the picture using the –cref command?

Let’s give it a try:

Voila: Here is the result:

That worked exactly as I had hoped.

Just note that you need to modify the prompt when using Vary (Region). Had I used the original prompt, the variation likely wouldn’t have had any effect (at least that was the case in my tests).

So, I recommend altering the prompt in such a way that it’s clear to Midjourney what you want to add when using the Vary (Region) option.

In the given example, I added “and Albert Einstein”, so the new prompt text read:

cinestill 800T film, Steve Jobs and Albert Einstein walking through Apple Campus, close up cw 20 –cref –v 6.0 –ar 2:1

2. Expand (left, right, top, bottom)

You can also achieve similar results with the expand command.

The Expand commands allow you to expand the contents of an upscaled image beyond its original boundaries in all 4 directions without changing the contents of the original image

Of course, you can also add new prompt elements to the expanded image area.

For this article, I expanded the existing Steve Jobs image to the right and inserted the modified prompt as before.

What happens?

Well, unfortunately not what I would have imagined. Einstein is nowhere to be seen.

Let’s expand the image again, this time with a simplified prompt text that simply says “Albert Einstein.”

But even now nothing happens.

Does the trick with Expand not work?

Interestingly enough, when I run the same scenario with an image of Einstein as the starting image:

As before, I expand the image to the right with the following prompt:

cinestill 800T film, Albert EInstein and Steve Jobs walking through Apple Campus, close up cw 20 –cref

And as if by magic, Steve Jobs actually appears in the newly generated image area.

Conclusion: Multiple consistent characters in Midjourney

We see that it is possible to create multiple consistent characters within an image using Midjourney with little effort.

But we also see that the expansion of the image does not always work as desired. In my experiments, it was easier to add the second character into the image using Vary (Region) than using the Expand method. When working with Midjourney, it is important not to bury your head in the sand too early and give up if something doesn’t work as desired. As a rule, the desired image result is only a few variations away.

So don’t give up, just keep on prompting!