3 Ways to Blend Images in Midjourney

In this newsletter, read about:

• 🕵️‍♀️ Blending Images in Midjourney

• 🗞 News and Top Reads

• 📌 AI Art Tutorial: Assigning Weights to Reference Images

• 🎨 Featured Artist: Adrian Pocobelli

• 🖼 AI-Assisted Artwork of the Week

• 🤓 A Comprehensive Midjourney Guide

🕵️‍♀️ Blending Images in Midjourney

Combining multiple images can frequently yield surprising outcomes with captivating images in unique styles. Many Midjourney users opt for the /blend command to combine two or more images, but this command was actually introduced as a user-friendly image prompting tool for mobile users, while Midjourney allows several more methods to blend images. These alternative solutions offer a greater degree of control and flexibility together with more aesthetically pleasing results.

In this guide, I want to introduce you to all these alternative solutions, enabling you to choose the one that best fits your specific use case.

The /blend Command

For quick results with minimum effort, you can still just use the /blend command. All you need to do is type /blend, upload the images you want to combine, and optionally, select the dimensions parameter from the three available options.

As an example, let’s experiment with these two images, representing famous paintings by Salvador Dali and Gustav Klimt.

If we simply blend these two images with the dimensions parameter set to “Landscape”, we’ll get the following results.

We can clearly see two styles combined, but the resulting images look a little bit chaotic and not so much coherent. We’ll address these issues with alternative solutions to combining images in Midjourney, but for now, here are a few things to remember when using the /blend command:

• It works with up to 5 images.

• It doesn’t accept a text prompt.

• When blending images you can set the optional dimensions parameter to “Square”, “Portrait”, or “Landscape.” If nothing is specified, square images will be generated.

• It’s recommended to blend images with the same aspect ratio for better results.

• If you have a preferred suffix set, it will be added to the end of the /blend prompt, like it’s added to the /imagine prompt.

Image Prompting

Image Prompting is probably the most convenient way to combine images in Midjourney while having a certain level of control and flexibility. You basically use the same /imagine command that you usually use with your text prompts, but you start the prompt with the links to one or more images that you want to be used as a reference for image generation.

The most significant benefit of this approach is the opportunity to direct the subject matter of the generated image through the text prompt. In the image above, I requested “a couple kissing under the tree”, which is closer to Klimt’s painting used as a reference. Still, the generated images clearly keep some style details from Dali’s painting.

Alternatively, we can ask for “a couple kissing near the clock”. We’ll get a new set of images with a different subject matter, but still keeping the styles of both artists.

Another benefit of image prompting is the possibility of adjusting the image weight (--iw) parameter. By default, this parameter is set to 1. But if we want the generated images to be guided primarily by image references and less by the text prompt, we can set this parameter to a maximum value of 2.

This image set is even closer to the reference images by Dali and Klimt.

More excitingly, Midjourney has recently added the possibility to assign different weights to different reference images. Check out the video tutorial of this newsletter for more details.

And I want to present an alternative approach to image blending, which also enables you to control the importance of specific reference images.

The /describe Command & Multi Prompts

This is a slightly out-of-the-box, yet still interesting approach to image blending, with its own unique benefits. I first encountered it in the video tutorial by Christian Heidorn.

The idea is to start with the /describe command to create prompts for our reference images. Then, we can generate images using a text prompt that combines the prompts we got from the /describe command.

For the sake of the experiment, I used the exact prompts I got from the /describe command (i.e., picked up the best), but obviously, you can adjust the prompts if you want.

Secondly, note that I used separators (::) after each of the two prompts so that Midjourney understands that we have two different concepts to combine.

And thirdly, you can use the text weights to tell Midjourney, which reference image should have a higher weight. Should output images be more of Dali or Klimt?

In the prompt above, Klimt’s aesthetics has a higher weight (1.5 vs 1). And here’s what happens when we switch the weights.

This image set is clearly more “Dali”.

Finally, you can obviously add the third part to your prompt, guiding the subject matter of the generated images. For example, let’s try to add “a couple kissing under the tree” with a weight of 2.

To Sum Up

There are several possible ways to blend images in Midjourney, and you can pick the one that best suits your request:

• If you want a quick and simple solution, use the /blend command.

• If you want to define explicitly the image subject matter and decide how much weight the reference images have on the final output, use image prompting.

• If you want to have flexibility in terms of the subject matter and the weight of each reference image, try the combination of the /describe command with multi prompts.

Good luck with your Midjourney generations!

🗞 News and Top Reads

• The AI rules that US policymakers are considering, explained

  • This Vox article provides a comprehensive overview of AI-related legislation being considered in the US.

• Apple claims they are building AI into every product.

  • Responding to the ongoing questions over the company’s plans, Apple CEO Tim Cook has said that artificial intelligence and machine learning are “virtually embedded in every product.”

• Amazon CEO Andy Jessy, in a similar claim, said that “every single” Amazon team is working on generative AI.

• Tinder explores AI to help users build better profiles.

  • The company is currently testing an AI photo selection feature that looks at a user's photo album and selects the five that best represent them.

  • Additionally, Tinder is looking into other AI features, such as using generative AI technology to help users write a good bio.

📌 AI Art Tutorial: Assigning Weights to Reference Images

In this guide, Nolan will walk you through the new Midjourney feature, allowing assigning different weights to the reference images in your /imagine prompts.

🎨 Featured Artist: Adrian Pocobelli

Adrian Pocobelli is an Italian-Canadian artist based in Berlin. As a creator, he likes to travel through different mediums. Pocobelli began experimenting a bit with AI art back when Google Deep Dream was released in 2015. In 2022, he started to explore it seriously with the help of Midjourney and Stable Diffusion. Pocobelli sees art as a tool to reveal qualitative aspects of reality that science cannot reveal because it’s limited to quantitative data only.

🖼 AI-Assisted Artwork of the Week

🤓 A Comprehensive Midjourney Guide

Share Kiki and Mozart

If you enjoy this newsletter and know someone who might also appreciate it, please feel free to share it with them. Let's spread the word about AI art and introduce more people to this fascinating field!