How To Create Consistent Characters In Midjourney

Midjourney Character Prompts produce random and inconsistent images by design. After all, the whole point is creating random art. But what if you wanted to make consistent characters with different poses and expressions? This could be useful for creating original characters for a graphic novel, comic books, video game personas or illustrated books for children.

Here's a quick answer on our own StackOverflow-like platform.

I found this great Midjourney 5 Consistent Character Tutorial on YouTube (it even explains how to use image seed numbers and additional generic "action" references to improve quality of actions applied to your character, like running, sleeping, reading, and so on!)

I recently started this other How to use Midjourney it's a guide for complete beginners.

Oh and one more thing guys...I'm an independent author, I would be grateful if you shared this article around. I created this button below, just click and share the link.

It took hours to put this material together. Please help me out by sharing it 🙂

Click button below to share link on WhatsApp or Discord or with friends:

This is really helping me write more free Midjourney tutorials!

First, decide what type of character image you want to create

• Is it a face or shoulder-level avatar image?
• Is it a full body character you want to capture in different poses?

These two types require a different approach. Step by step instructions for creating the full body character are described below. Face and avatar will be added soon.

Why would you want to do this?

One strategy for becoming successful at selling books on Amazon is to produce a large number of no-content or low-content products. But this can be very time consuming.

By using Midjourney you can significantly reduce amount of time it takes to create your illustrated book characters, backgrounds and even book covers, without having to hire a professional graphic designer or illustrator.

Now imagine if you could speed up the process of creating medium-length content books for Amazon KDP program? In this how to make consistent characters in Midjourney tutorial, I will share several techniques you can start using to make mid-size illustrated books faster.

Before you get started...

First step to creating consistent characters is to not expect perfect results, or for it to even work properly. Art generated with AI tools was never meant to make uniform characters, so generally it is not reliable for the task. This means there will be madness to the method!

Having said all this...we can still wrestle with the AI to make consistent characters, even when they appear in different poses or have different expressions on their faces.

Creating consistent characters is useful if you're working on a graphic novel, concept art for a video game, children's illustration book or similar project.

Important: Midjourney can help you create different images of your character in different poses, with different expresisons. But don't expect Midjourney to perfectly generate this first image for you. You'll have to do some work on your own, and that's perfectly ok.

If you are an artist, you can use an existing character you created before. Just draw it in multiple poses and split each pose into a separate image. Then use these images as a starting point for your midjourney image prompt.

If you're not an artist, you can use Midjourney to generate your first character.

Below we'll discover several techniques on how to hone your skill of creating these initial images. And while results won't always be perfect, Midjourney prompting is something you will just have to get better at with time and practice.

When making characters with Midjourney for a book it's important to make them consistent.

This is like going against the grain of what AI art is all about.

After all, the whole idea is to create improvisations?

Midjourney generates random images and that's a problem if you want to create characters that remain consistent throughout different prompts.

Here's a YouTube video on the subject:

In this tutorial we will explore a workflow for creating consistent personas.

What we want to do is to generate a character whose face, and unique characteristics, can be recognized across multiple prompts, regardless of the action they are performing (walking, swimming, pouring a cup of tea, driving a car, riding a bicycle, etc.)

How To Create Consistent Style Characters In Midjourney

To make consistent characters follow instructions below:

• Start prompt with a URL image describing character's face.
• Use reference photo to generate your prompt.
• Use same image as a starting point to generate all future prompts.
• Try using the same --seed parameter.
• Use clear, consistent descriptions of facial features.
• Describe the character in as much detail as possible.
• Avoid conflicting keywords.

For example, you could prompt the AI to generate an image of a character with brown eyes, a narrow nose, and curly hair. Avoid ambiguous or conflicting keywords as they can lead to inconsistent results.

When writing the prompt be clear and consistent in your descriptions of the characters' facial features, such as the shape of their nose, the color of their eyes, and the style of their hair.

Also be consistent in the type of clothes the character is wearing. You have to retype all of those keywords into each prompt. So make sure to make a solid base prompt first.

How To Create Consistent Character In Midjourney

Follow step by step instructions below to start creating your own original character, consistent in looks, regardless of what it's doing, its pose or face expression.

To create a grid of multiple images of the same character, also known as character sheet, use this Midjourney prompt:

Notice --no outline parameter.

In Midjourney using --no followed by a keyword, will remove that keyword from generated images. Or at least try to avoid it to its best ability. When making an illustrated children's book, this might be a good choice.

The prompt above generated the following image:

There are 4 variations here. Each character will look similar but not exactly the same. Pick and upscale your favorite version by clicking on any of the U1, U2, U3 or U4 buttons.

I chose first image and clicked on U1 button.

The image generated was:

(Of course you'll want to create it from your own description, which could be completely different than this little witch!)

Now open this image in Photoshop or similar image editing software and separate each character by extracting individual frames from the prompted image above.

Use selection tool to draw a box around one of the characters that closely matches your favorite version. Copy it and paste it into Discord channel.

Repeat this step for several character images. The idea is to make Discord generate unique URLs we can use to pass into our prompt in the next step.

Right click on each of the individual character images in Discord, and click on the "Copy Image Address" option.

Start /imagine prompt and paste the URLs into the prompt.

Repeat for all other images, until you have 4 or more URLs separated by space in your prompt. This will give Midjourney different versions of your character to work with. From now on, future prompts will be consistently tied to this look and feel.

Once you added all URLs into the prompt, copy and paste the original prompt with a space after the images.

Now change the part of the prompt that used to say multiple poses and expressions with just the action you want the character to perform. For example: running.

Your new prompt looks something like this:

Here imgurl1.png, imgurl2.png, etc... are the URL addresses of individual images we pasted into Discord earlier. They can be located anywhere on the internet - you can even upload it to your own web server. But it's just easier to extract them from the same Discord channel without being distracted from the workflow.

Hit enter and the prompt will produce the same character running:

Repeating the same prompt yields even more variations:

Final Thoughts

Did we achieve perfect consistency? Not really. But it's not bad considering we only spent about 5 minutes creating a completely original character.

You can always fix nuances in post-production and make your character consistent with all of its manifestations. Still this is definitely a time-saver, compared to creating characters from scratch.

Oh, and another thing...I am an independent writer, so if you help share this article on Discord or WhatsApp or whatever app you're active on, I would be very grateful! Thank you 🙂

It took hours to put this material together. Please help me out by sharing it 🙂

Click button below to share link on WhatsApp or Discord or with friends: