How to Create Diverse People in Midjourney

PLUS: latest updates from Google, Stability AI, and Instagram

In this newsletter, read about:

  • 🕵️‍♀️ Creating Diversity in Midjourney

  • 🗞 News and Top Reads

  • 📌 AI Art Tutorial: Best AI Upscaling Tools

  • 🎨 Featured Artist: Amber

  • 🖼 AI-Assisted Artwork of the Week

  • 🤓 A Comprehensive Midjourney Guide

🕵️‍♀️ Creating Diversity in Midjourney

The Midjourney team is on the verge of releasing V6 any day now. However, before they do, let's have some fun exploring further the latest available version, V5.2.

Today, I want to experiment with creating diverse representations of people in Midjourney. We all know that AI image generators, including Midjourney, often tend to produce a particular type of 'beautiful' people, with the generated images frequently reflecting societal stereotypes. Recently, I faced a significant challenge when attempting to generate an illustration of a female scientist presenting at a research conference. Simply specifying 'a researcher' would yield images of men, and adding 'woman' or 'female' would result in highly sexualized versions of 'researchers.'

So, let’s explore how we can generate portraits of individuals from various ethnicities, with diverse body types, and perhaps even with specific disabilities.

Ethnicities

To depict various ethnicities, it's usually sufficient to add the corresponding description to the prompt, such as 'Black,' 'Native American,' 'German,' 'Japanese,' 'Mexican,' etc. However, please note that in many cases, Midjourney will attempt to represent the requested ethnicity with the corresponding national costumes and possibly even surroundings. If this is not the desired outcome, you can add a few words specifying clothing and other image aspects that would better suit your particular case.

Here are a few examples.

a photo of a beautiful Native American woman, fashion photography --ar 4:5

As you can see, in these images, Midjourney tries to communicate ethnicity mainly through clothing. So let’s try to add “casually dressed” to reduce this effect.

a photo of a beautiful Native American woman, casually dressed, fashion photography --ar 4:5

Here you can still see the elements of the Native American clothing style, but it looks much closer to what I wanted to get.

Below, I do the same trick with a portrait of a Japanese woman.

a photo of a beautiful Japanese woman, fashion photography, golden-hour lighting --ar 4:5

a photo of a beautiful Japanese woman, casually dressed, fashion photography, golden-hour lighting --ar 4:5

Body Types

To depict diverse body types, we again start with the vocabulary. Pick up one or a few words that describe what you are looking for. Some examples include slender, muscular, curvy, plus-sized, slim, chubby, toned, petite, etc.

You may just add these words to the prompt and see how it works for you. However, it’s often the case that you get “too much” or “too little” of the requested body type. In this case, you can use the “slider” method (i.e., text weights) to adjust the image.

For example, let’s start with a photo of a muscular man playing beach volleyball.

a photo of a muscular man playing beach volleyball --ar 4:5

If we want the man to look less muscular, we can reduce the weight of the “muscular” word with the following prompt.

a photo of a muscular man playing beach volleyball:: muscular::-0.3 --ar 4:5

Alternatively, we can generate an image of a really BIG guy by increasing the weight of the “muscular” word.

a photo of a muscular man playing beach volleyball:: muscular::1.5 --ar 4:5

Now, suppose we aim to generate a realistic image of a black woman shopping in the mall. By default, Midjourney tends to create slim figures, unless specified otherwise. To achieve a more average body type, let's include the word 'plus-sized' in our prompt.

a photo of a beautiful plus-sized black woman shopping in the mall --ar 4:5

Not bad, but the women are probably too ‘plus-sized’ for the average body type. So let’s try to adjust this by reducing the weight of the ‘plus-sized’ part.

a photo of a beautiful plus-sized black woman shopping in the mall:: plus-sized::-0.5 --ar 4:5

Looks great!

Persons with Disabilities

In most cases, you can use appropriate words to describe specific types of disabilities based on your intended image.

For instance, if you seek an image of a person in a wheelchair, you can simply specify that in your prompt. Similarly, using terms like 'disabled' or 'accessibility' typically generates images featuring individuals in wheelchairs.

a photo of a professor in a wheelchair in a classroom --ar 4:5 --s 50 --style raw

a photo of an accessibility female professor in the classroom --ar 4:5

To generate images depicting individuals with Down Syndrome, you can simply use the corresponding phrase.

a photo of a girl with Down Syndrome reading a book --ar 4:5 --style raw

Remember, similar to what we did with body types, we can always use a slider method with various text weights to emphasize or de-emphasize certain traits.

I hope this guide helps you create diverse images, celebrating the richness of our world.

Happy prompting, and of course, Happy Holidays!

🗞 News and Top Reads

  • Google introduced Imagen 2, its AI model that can create and edit images given a text prompt.

    • Compared to v1, Imagen 2 significantly improved in terms of image quality and can also render text and logos.

    • Currently, the model is only available to Google Cloud customers using Vertex AI.

  • Stability AI introduced Stable Zero123 for generating 3D objects from single images.

    • Stable Zero123 can understand the object’s appearance from various angles with notably improved quality over the previous Stability AI’s 3D models.

    • The model is being released for non-commercial and research use.

  • Midjourney has launched image generation through its website, but now only those who have generated 10,000 or more images have access to this feature.

  • Instagram introduced its generative AI-powered background editing tool

    • When users tap on the background editor icon on an image they get ready prompts like ‘On a red carpet,’ or ‘Surrounded by puppies,’ or they can write their own prompts to change the background.Currently, the tool is available to US-based customers only.

📌 AI Art Tutorial: Best AI Upscaling Tools

In this video, Matt experiments with two new AI upscalers, Magnific AI (paid) and Krea AI (free). They are not simply increasing the resolution of images but can create a realistic image from an old video game screenshot, truly reimagining the provided images. It finally looks like magic 😄 

Amber, an AI artist utilizing Midjourney, creates distinctive masterpieces. Formerly a portrait and fashion photographer, her images showcase how her photography skills guide her in selecting prompts to generate unique styles that resonate so well with her audience.

🖼 AI-Assisted Artwork of the Week

🤓 A Comprehensive Midjourney Guide

To get a link to a comprehensive Midjourney guide, please subscribe to this newsletter. The guide is a dynamic document, which I intend to keep up-to-date with the latest Midjourney updates.

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!

Join the conversation

or to participate.