Creating Consistent Characters in Midjourney Version 5

Having trouble getting consistent characters in Midjourney V5? You're not alone. Version 5 requires a different approach than version 4, which did a lot of the heavy lifting for us. This new version pushes us to improve our prompting skills. Let's look at some methods that still work for creating reliable characters in Midjourney V5.

Starting Points: Why Old Prompts Don't Work

If you try using your old V4 prompts directly in V5, you might get unexpected results. While the images might be realistic, the style might be completely off. Midjourney V5 seems to interpret prompts differently, sometimes ignoring style cues that worked before. This can be frustrating when you have a specific look in mind.

Adjusting Your Prompt Structure for V5

One key change in V5 is to put the style description earlier in your prompt. Instead of describing the character first and then adding the style at the end, try leading with the style you want.

For example, instead of: `Imagine a beautiful woman... in the style of Marvel comic book --v 5`

Try:

[LI]`Imagine Marvel comic illustration style, headshot of a beautiful woman... --v 5`[/LI]

This simple change can significantly improve how Midjourney interprets the style you want. You might find the results much closer to a specific illustration or art style you're aiming for.

Blending Styles for a Specific Look

Sometimes, a single style isn't enough to replicate a previous look, especially if your old V4 images had a unique finish. V4 often added its own layer of stylization. In V5, you might need to be more explicit about combining influences.

Consider blending different art styles together. For instance, if you want a polished, slightly stylized look, you could try combining elements like:

• Comics style

• 3D animation style

• Anime style

By listing multiple styles, you can tell Midjourney to create something that falls in between them. This requires some experimenting, but it allows you to cook up almost any look you can imagine if you use the right words.

Adding Detail and Parameters

To get even closer to a consistent look, add specific details to your prompt. Think about:

• Background color (e.g., on white background)

• Shot type (e.g., close-up, headshot)

• General appearance cues (e.g., attractive European woman, though using simple descriptive terms is often better than demographic labels if not essential to the concept).

• The `--stylize` parameter: In V4, images were heavily stylized by default. In V5, adding `--stylize 1000` (a typical high value) can sometimes help bring back that polished look, but this should be used carefully and is specific to certain desired results.

Adding these descriptive elements helps guide Midjourney towards the specific output you need for a consistent character.

Ready to take your Midjourney creations to the next level? Explore ways to automate and manage your images more effectively. The TitanXT Midjourney Automation Suite can streamline your workflow and help you manage the consistency you're building.

Using Image References for Consistency (The Core Method)

Okay, now for the main technique to create truly consistent characters: using image references. In V5, the best results often come from generating a few high-quality, slightly varied reference images *using* V5 first, and then using those as inputs for future generations.

Creating Your V5 Reference Images

Start by generating a few base images of your character using the refined text prompts discussed above. Get a few different poses or angles if possible. Pick the ones you like best. These images will become your core visual anchors.

Generating Consistent Shots Using References

Once you have your V5 reference images, include them at the beginning of your prompt. You do this by adding the image URL(s) before your text description.

`Imagine [URL_of_Image_1] [URL_of_Image_2] [URL_of_Image_3] [URL_of_Image_4] refined text prompt goes here --v 5 --aspect X:Y`

Using multiple references helps Midjourney understand the character's look. You can also use the `--iw` parameter (image weight) to control how much influence the images have compared to your text prompt. A value like `--iw 0.75` might give the text prompt more control while still keeping the character consistent.

Consistency in Different Situations

Adding Emotions

With the image references providing character consistency, you can easily add emotions using text prompts. Simply add phrases like "smiling with a half open mouth," "laughing," or "slightly upset" to your prompt. The character's expression should change while their core appearance stays the same.

Placing Characters in Scenes

Midjourney V5 allows you to place your consistent character in different locations. Add descriptive text about the scene, like "standing in a small Parisian street" or "in New York." Be aware that lighting, background, and even clothing style might change subtly to fit the new environment, but the character's face and core features should remain consistent thanks to the image references.

Need help managing your Midjourney images, prompts, and variations? The TitanXT Midjourney Automation Suite is designed to help you keep track of your creative process and maintain consistency across many generations.

Controlling Pose and Framing

You can also guide the pose and framing by adding terms like "frontal headshot," "looking straight into the camera," "headshot from the side," or describing the shot type (wide angle, full body). Changing the aspect ratio (`--aspect 16:9`, `--aspect 4:5`, `--aspect 9:16`) will also affect how much of the character is shown and the overall composition.

Trying to Change Clothing Color

One challenge in Midjourney V5 is consistently assigning specific colors to clothing items, unless they are the primary focus. You can try phrases like "wearing blue jacket" or "wearing white pants," but Midjourney might apply the color to a different part of the image.

Using Old V4 References

Can you use your old V4 reference images directly? You can try. If your V5 text prompt already generates images with a similar style, using V4 references might help bring some of the previous look back. However, it's not as simple as plugging them in. If your V5 text prompt is far off in style, the V4 references may not have enough influence to pull it to the desired look. Creating new V5 references is often more reliable.

Beyond Portraits

The methods shown here work well for portrait shots and images where the character is the main focus. If you want to place your character in a larger scene with more action or where the scene is as important as the character, you might need to generate full-body character references first using the same principles, then use those references for the wider shots.

Your knowledge from character design in previous versions is still valuable! The core concepts of using references and refining prompts carry over, even if the specific prompting style in V5 is a bit different.

Ready to make character consistency easier? The TitanXT Midjourney Automation Suite provides tools that can help you manage your image references and generated variations, making consistent results more attainable.

Learning V5 takes practice. But by adjusting your prompt structure and using the right image reference techniques, you can reliably create consistent characters for your projects.