How to Create Stylized Selfies in Midjourney

In this newsletter, read about:

• 🕵️‍♀️ Creating Stylized Selfies in Midjourney

• 🗞 News and Top Reads

• 📌 AI Art Tutorial: G-Prompter

• 🎨 Featured Artist: Linda Dounia Rebeiz

• 🖼 AI-Assisted Artwork of the Week

• 🤓 A Comprehensive Midjourney Guide

🕵️‍♀️ Creating Stylized Selfies in Midjourney

Midjourney is not a photo editing tool, so it is not designed to stylize selfies. However, you can use a few tricks to generate a stylized image of a person who resembles you, although it will never be a close resemblance.

Preparing reference images

To start with, you’ll need to shoot 3-5 selfies of yourself. Here are a couple of recommendations to get better results:

• Photos should be close enough to see details.

• Your face should be front-on facing the camera with good lighting and clear focus.

• Make sure the whole face is visible and key elements like eyes, nose, and mouth are not covered with hair, sunglasses, or anything else.

• The proportions should be about the same for all reference images – i.e., don’t blend full-body images with face images.

• Your age should be about the same on all reference photos – don’t mix old and new selfies.

• Backgrounds and lighting can be different.

Here are my selfies that I used as a reference.

Getting image description

To find the best prompt to describe yourself, I would recommend using the /describe command with one of your selfies. For example, here are the descriptions that I got for my photo.

After testing all of these prompts, I found that prompt #1 produced the best results, so I used it as the basis for my future experiments.

The images look awesome, but they don't quite capture the resemblance or have any cool stylization yet. Let's keep pushing forward!

Blending reference images with a text prompt

First, I want to create an anime version of myself. So, I start with one of my selfies as a reference and use the following text prompt.

Note that I’m using the Niji 5 version in this prompt as it works best for anime and illustrative styles.

I like the image, but let’s try to get an even closer resemblance. This time I use all three of my selfies as a reference and also add the image above because I’d love to achieve something similar. Here’s the prompt:

Great! Now let’s do some zooming out to get a more complete image.

Ok, here’s an anime version of myself. Let’s now experiment with some other styles.

Experimenting with styles

To get my image as a futuristically styled illustration, I use all of my selfies as references and combine them in the following prompt.

Next, I want to experiment with watercolor portraits. My initial experiments showed that with three photographs as references, the results are too photorealistic even if I prompt for a “watercolor portrait”. So, I created a watercolor portrait first to add it to the reference images along with my selfies.

Next, I run the following prompt to get my image in the style of a watercolor portrait.

Note that I put a higher weight (::1.3) for the image of a watercolor portrait above. Otherwise, the output images still look too photorealistic. Here’s the closest I could get in terms of resemblance.

That’s it! Here I showed just a few examples of how you can play with your selfies as references to get images in a variety of styles. Feel free to experiment with other styles. You may even go further by experimenting with different backgrounds and environments that you put yourself in. You may need to use panning and zooming for this, as it would be better to start with a close-up face image for better resemblance. Remember that reference images in Midjourney are just a guide, and it is usually not possible to get an exact likeness of yourself.

🗞 News and Top Reads

• Adobe's Firefly generative AI models, previously in beta, are now generally available in Creative Cloud, Adobe Express, and Adobe Experience Cloud.

  • Adobe plans to charge for Firefly using credits known as "generative credits." All users of existing paid Adobe plans will receive a sizable allocation of these generative credits. Additionally, once you run out of credits, you will not lose access to Firefly, but it will run significantly slower.

  • The company is also launching Firefly as a standalone web app.

  • Note that Adobe has also started distributing bonuses to the Adobe Stock contributors, whose images were used for training the commercial version of Adobe Firefly.

• OpenAI is racing to launch GPT-Vision, a multimodal large language model that can process and understand images in addition to text.

• TikTok introduced a toolset for creators to label AI-generated content.

  • The company requires that creators label AI-generated content with realistic images, audio, or video to help viewers understand it and prevent misinformation.

  • However, TikTok users are not obliged to use the newly introduced label. Instead, they can use stickers or captions to distinguish content without violating community guidelines.

• Andreessen Horowitz, a prominent venture capital firm that invests in AI, has published the results of its research on how consumers are using generative AI. After narrowing down the top 50 generative AI web products by monthly visits, a16z found that:

  • ChatGPT accounts for 60% of monthly traffic to the entire top 50 list, with an estimated 1.6 billion monthly visits and 200 million monthly users (as of June 2023).

  • In addition to LLM chatbots, two other categories have begun to drive significant usage in recent months: AI companions (such as CharacterAI) and content generation tools (such as Midjourney and ElevenLabs).

📌 AI Art Tutorial: G-Prompter

In this video tutorial, Matt Wolfe demonstrates a new tool called G-Prompter. This tool enables you to train your own prompt generator using prompts that have previously yielded excellent results for you or other users. For instance, you can train it to generate detailed and lifelike images by providing 5-10 prompts that have produced such images in the past. Once trained, you simply input a desired subject (e.g., "a firefighter"), and the tool will provide a prompt that will generate a photorealistic image of a corresponding subject. The tool works for different styles and image generators.

🎨 Featured Artist: Linda Dounia Rebeiz

Linda Dounia Rebeiz is an artist and designer who investigates the philosophical implications of technocapitalism and its role in furthering systems of inequity. She also writes about and curates art exploring these ideas. Her work mediates her memories as alternative truths and evidence of excluded ways of being and doing. It is formed in the liminal space between the immaterial and the material through the combination of analog and digital mediums – acrylic, ink, pastels, markers, scanners, vector, video, GANs, generative AI, code, and a range of materials not intended for art making.

🖼 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!