Multiple Subjects Using Midjourney V6: All You Need To Know

In this newsletter, read about:

• 🕵️‍♀️ Multiple Subjects with Midjourney V6

• 🗞 News and Top Reads

• 📌 AI Art Tutorial: Midjourney in 2024

• 🎨 Featured Artist: Hana Katoba

• 🖼 AI-Assisted Artwork of the Week

• 🤓 A Comprehensive Midjourney Guide

🕵️‍♀️ Multiple Subjects with Midjourney V6

The Midjourney team has finally released V6. It's currently in the Alpha stage, indicating that its performance is not yet reliable, and many features from previous versions are not available in V6. However, you can already observe the significant differences in this new version.

Let's briefly explore how the prompting of V6 differs from that of earlier versions. And then, we'll delve into the main topic of this issue – how depicting multiple subjects in one image is elevated to a whole new level with V6.

Writing Prompts in Midjourney V6

The announcement introducing V6 is saying: “Prompting with V6 is significantly different than V5. You will need to 'relearn' how to prompt.” Yet, it doesn’t mean that all you know about Midjourney is wrong now. A lot of general principles remained the same, but there are some changes that you should be aware of.

• V6 can process much longer prompts, but it’s not that much about the number of words (like it was in earlier versions), but rather about the semantic complexity of the prompt.

  • If you have too many details packed into a short prompt, Midjourney will not be able to process all the requests.

  • On the other hand, you may describe the scene and subjects in multiple simple sentences, and even if the prompt looks long, Midjourney V6 might be able to process all the details.

• V6 is much more sensitive to the prompts. As a result, “junk” words like “8K, photorealistic, award-winning, hyper-detailed” are not just unnecessary, but can actually be detrimental to the prompt.

• Prompts written in simple sentences with proper spelling and punctuation tend to work better.

• V6 can incorporate text into the image. Introduce the text in quotation marks and make sure that there is a place for it to show up.

• Finally, in V6, you can describe a simple scene with several different subjects and Midjourney would be able to reflect most of the details (if not, all).

Depicting Multiple Subjects in V6

Having multiple different subjects in one image is not that much of a challenge now as V6 is so much better at following the instructions. If you are prompting for a reasonable amount of details, you’ll probably get what you want without any specific prompting hacks. But if you have issues, here are some recommendations based on Clarinet’s guide on Multiple Subjects and my own experiments:

• It’s usually helpful to start with a sentence, describing the entire scene, before going into the details, for example, “Three men sitting in the bar” or “Four different cats sitting on a couch.”

• Then, describe each subject by referring to them similar to how you introduced these subjects in the first sentence, e.g., “a man in the middle”, “a cat on the left.”

• It’s better to use “positional” words like “on the left”, “in the middle”, and “behind”, than just “first”, “second”, or “third”.

• After describing the subjects, feel free to add details about the background and environment, if you have something specific in mind. Describing the setting in detail usually works better.

• Finally, you may add some phrases, describing the image aesthetics. The more specifically you describe your desired aesthetics, the more likely you’ll get what you are looking for.

Despite the significant progress V6 made in depicting multiple different subjects in one image, it’s still not perfect. Quite often, you’ll get the subjects blended, meaning that the details described for one person, will be assigned to another person. For example, if you describe a couple that includes a man and an Asian woman, you are likely to get both person Asian. To minimize blending, add the corresponding details for the second person as well. In our example, you may specify that you want an image of a Caucasian man and an Asian woman.

Now let’s see a few examples of how V6 handles multiple distinct subjects.

Experiments

I’ll start with a scene depicting just two different subjects.

Amazingly, we could get all the desired specifics without any blending between the subjects. This was impossible in earlier versions.

Now, let’s try to challenge Midjourney with more subjects: a family, consisting of a mother, a father, a daughter, a son, and a cat 😄 

I needed to do a few re-rolls to get to this outcome and you may notice a few examples of “blending” – the son has glasses that were requested for the father only, and the mother’s eyes look slightly strange, as Midjourney was struggling to keep her Caucasian in the middle of the Asian family members 😄 

But otherwise, it’s a pretty amazing result, considering the number of distinct subjects in the image.

Now, let’s see how V6 incorporates text in images.

It’s not exactly a mugshot but very close to our request. There is still quite a lot of blending and to get the text right, I needed to request variations of the image generated with slightly wrong spelling.

As with other AI image generators, it’s much easier to get some basic words in Midjourney V6, like “Thank you”, “Welcome”, “Hello”, etc. These can be generated from the first attempt, without any trouble. If you want something less common, it can be a challenge, but you may still get what you want after a few re-rolls.

I hope this brief guide will help get the most from Midjourney V6. Happy prompting!

🗞 News and Top Reads

As the year 2024 is predicted to be all about AI video generation, it’s not a surprise that we have lots of video-related news this week:

• Leonardo AI has released a new feature to create dynamic motion videos from images.

• Pika 1.0, the state-of-the-art text-to-video AI platform, is now available for everyone to use.

• Midjourney will start training video models this month, as announced by the company CEO David Holz during the latest “Office Hour” session.

In other news:

• The New York Times is suing OpenAI and Microsoft for copyright infringement.

• Developers of generative AI models might be soon required to disclose sources of training data so copyright holders know their information was taken (if the corresponding bill passes in the US).

• Microsoft Copilot is now available as an app on Android devices.

📌 AI Art Tutorial: Midjourney in 2024

In this video, Nolan explains in detail how prompting V6 differs from prompting earlier Midjourney versions. He also covers most of the Midjourney features, providing a great overview of everything you need to know to use Midjourney in 2024.

🎨 Featured Artist: Hana Katoba

Hana Katoba is a digital artist, who has managed to captivate the attention of critics and art lovers with a unique blend of artificial intelligence and the symbolism of flowers. She starts her creative process by meticulously choosing various flowers, each carrying a distinct meaning. Through the application of artificial intelligence, these flowers are digitally merged and manipulated, resulting in visually stunning compositions.

🖼 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!