I’ve been continuing my experiments and tests with Midjourney for image generation. For one of my current client projects, I needed a number of images with a unified look and feel. In the past, I have spent a significant amount of time searching stock image sites for a series of images with that consistent style and the right subjects. Sometimes I have found images that were close and then edited them. However, just like Midjourney can create consistent character images, it can also create images with consistent style. This is especially effective for images in different illustration styles.

In this post, I’ll share the process for generating images with a consistent style in Midjourney. You can use this to create a set of images for courses and other projects. It’s definitely not perfect, but I recommend experimenting with it yourself so you can see how it works.

Prompt for consistent style with –sref

The first step is to create the original image that will be the source for the remaining images. For this set of images, I’m using “2D illustration” as the general style prompt. Use the command “/imagine” to start Midjourney prompts.

2D illustration, woman, long brown hair, blue shirt, working at a desk in a modern office

I picked my favorite one out of the choices and upscaled that image. This will be my reference image.

To create images in the same style, use the style reference command at the end of the prompt. Start with –sref, followed by a link to the reference image. In the example below, I used –sref and included the description of the style (e.g., 2D illustration) in the prompt.

2D illustration, a meeting in a modern office conference room, --sref https://s.mj.run/kgCthfgMz88

This is a little more stylized than the original reference, but it maintained the color scheme and general style. There are some weird polygons on the chairs and walls though. Count the number of feet in relation to the number of people too; Midjourney struggled to stay accurate with this many people in the image.

I followed the same process to generate additional images. Each time, I included –sref in the prompt with a link to the original image. The realism and level of detail noticeably decreases when the images include multiple people or objects. You can see that the colors are pretty consistent, but not always the style.

Create icons

I experimented with creating a set of icons in Midjourney as well. The phrase “modern line icon” seemed to be the right one to get the style, but I used several phrases to generate the initial image prompt.

Computer mouse, modern line icon, vector line art, bold outline, solid color, white background

After that, I generated the rest with the –sref attribute. There are some inconsistencies, so I’d probably want to clean them up a bit if I was using these in an actual elearning course.

• The source image has a more saturated red color, but the later image generations used a less saturated version (especially the trash can).
• Some of the images had light blue circles or ovals behind them (which would be easy enough to edit out).
• Some of the images had shadows, even though the original didn’t.

Even with the issues, if I wanted to generate unique icons, Midjourney seems very usable. The minor editing on these images would probably take less time than finding and editing icons from a stock image site. I would certainly use this approach for a course, especially if I needed some specialized icons that were hard to find.

Read more

Read more about style reference in Midjourney’s documentation.

Check out my previous posts on Midjourney.

• Create Consistent Character Images in Midjourney
• Experiments with Midjourney Character Images

Share this:

• Twitter
• Facebook
• LinkedIn
• Email
• More

• Reddit
• Pinterest
• Pocket

Related