How to Speak to Midjourney V6

PLUS: latest from Midjourney, Runway, and Google

Author

Kate Koidan
January 24, 2024

In this newsletter, read about:

  • 🕵️‍♀️ Speaking to Midjourney V6

  • 🗞 News and Top Reads

  • 📌 AI Art Tutorial: Prompt Engineering with Seeds

  • 🎨 Featured Artist: David Alf

  • 🖼 AI-Assisted Artwork of the Week

  • 🤓 A Comprehensive Midjourney Guide

🕵️‍♀️ Speaking to Midjourney V6

I’ve noticed that a lot of people continue prompting V6 the same way they prompted V5.2. They attempt to convey as much information as possible using minimal words, often resorting to listing keywords separated by commas. Admittedly, this approach can still yield impressive images, especially if you don’t try to achieve something specific. Moreover, even if you have something specific in mind, but it’s not very complicated, these short, keyword-based prompts might give you what you want. However, this approach doesn't fully harness Midjourney's potential.

Honestly, I'm guilty of throwing in some V5-style prompts too. It's only when they flop that I'm like, "Alright, it looks like I am gonna need to put some more effort and write a nice, well-thought prompt.” 😄 

That’s how I could see a noticeable difference in the results when using these two different prompting approaches. So, today I want to share a few examples to show how you can significantly enhance your results by speaking to Midjourney V6 with full, natural-language sentences.

More Lively Images

When I speak to Midjourney its native language, it tends to produce less artificial and more lively images. For example, see below how a full-sentence-based prompt performed compared to a keyword-based prompt.

dancing woman, Brazilian carnival, event photography, bright colors --ar 4:5 --style raw --s 200 --v 6.0

These are perfectly fine images, but they don’t look very life-like.

The photograph features a woman participating in a dance at a Brazilian carnival. She is wearing a colorful costume that includes a bikini top decorated with beads and rhinestones, as well as a long, flowing skirt in shades of yellow and green. The woman is captured in mid-movement, with her arms raised. The photo is taken at night, and the lights of the carnival are visible in the background, along with confetti and other decorations in the air. The camera angle is slightly low. --ar 4:5 --style raw --s 200 --v 6.0

These are much more vivid images, through which you can truly feel the carnival mood.

More Artistic Images

I also think that it is easier to get more artistic and captivating images using natural language prompts with Midjourney V6.

a couple sitting on a bench in the evening, cinematic, 1960s, lanterns --ar 16:9 --v 6.0 --s 50 --style raw

The results above correspond to the prompt but don’t look very interesting. We all know that Midjourney can do much better.

35mm film still of a man and a woman sitting next to each other on a bench. It's evening and the lighting comes from street lanterns. The woman and the man are dressed in the 1960s style. The woman is wearing a red dress and the man is wearing a brown suit. --style raw --ar 16:9 --v 6.0

Here, I used a “35mm film still” keyword to get a cinematic shot. I also started by setting the main scene and then, added the details about a man and a woman. I think the results look much better. #3 is my favorite.

Better Adherence to Prompts

Finally, if you have a complicated request with several different subjects or objects to appear in the image, using natural language results in much better prompt adherence.

business photo, featuring a team of co-workers, including a middle-aged Chinese woman, an Indian man in his 50s, a blonde woman wearing glasses in her 40s, and a black man with short dark hair, standing in the office, looking at the camera, editorial photography --ar 16:9 --style raw --v 6.0

The results are not that bad, but still, none of the images follow the prompt exactly.

The photograph features a team of four coworkers, who are standing together and looking at the camera. This is an editorial photograph, taken in the middle of the office. On the left, there is a middle-aged Chinese woman. Next to her, is an Indian man in his 50s. Next to the Indian man, is a blonde woman wearing glasses in her 40s. Finally, on the right, there is a black man with short dark hair. --ar 16:9 --style raw --v 6.0

After I rephrased the prompt, I could get the desired outcome at least in one image out of four.

Conclusion

Midjourney V6 produces reliably better results when you prompt it with full, natural language sentences:

  • Don’t try to fit all the requests into the minimum amount of words. V6 doesn’t care about the number of words but about the informational loading of the prompt. If you have too many details, no matter in a long prompt or a short prompt, Midjourney will not be able to process all of them.

  • Keywords separated with commas usually produce worse results than prompts that include the same information in full, grammatically correct sentences.

  • When I’m talking about a long prompt, I don’t mean the prompts that ChatGPT and other similar bots output by default. These tools tend to produce image descriptions that include a lot of details about the feelings evoked by the image, its artistic value, etc. This is not relevant to Midjourney. The prompt should simply describe what you want to see in the image. Of course, it’s possible to also guide ChatGPT to produce such prompts.

  • Start the prompt by setting the main scene, and then go into details about each subject and object you want to see in the image.

  • Feel free to explore different styles and mediums. For more reliable results, you may include this information at the beginning of the prompt, and then, at the end, maybe with slightly different wording. For example, if I want to get a life-like photo, I can mention several times something like “an iPhone picture”, “a photo taken on iPhone,” etc.

Happy prompting!

🗞 News and Top Reads

  • Midjourney Office Hours – updates as of Jan 17, 2024:

    • Zoom, Pan, and Vary Region should appear in V6 this week.

    • A new Style Tuner will be released for V6, and it will be different from the one we use with V5.

    • Image generation through the website will be soon open to more people.

    • The Midjourney team is working on video generation, and it will come out earlier than 3D generation.

  • Runway has introduced Multi Motion Brush:

    • Now, you can request different motion patterns for various areas of your image (up to 5 independent areas of motion control!).

  • Chrome is getting 3 new generative AI features:

    • With Tab Organizer, Chrome will automatically suggest and create tab groups based on your open tabs.

    • Text-to-image diffusion model will power the generation of custom themes based on a subject, mood, visual style, and color.

    • Finally, in next month's Chrome release, Google will launch another experimental AI-powered feature to help users write about anything on the web (e.g., reviews, inquiries, responses).

  • Nightshade, the free tool that ‘poisons’ AI models, is now available for artists to use.

    • As other similar tools, it subtly alters the image at the pixel level so AI programs see something totally different than what’s actually there.

    • In contrast to the team's previous tool, Glaze, which had a defensive purpose, Nightshade is designed to function as an "offensive tool."

    • If an AI model is trained on numerous images modified with Nightshade, it might incorrectly classify objects for all users of that model, even in images not shaded with Nightshade.

📌 AI Art Tutorial: Prompt Engineering with Seeds

In this video, Nolan from Future Tech Pilot shows how to understand, which parts of the prompt contribute the most to the final result by testing various versions of the prompt with the same seed number. This technique can help you create reliably good-performing prompts without overloading them with unnecessary information.

David Alf is a licensed architect, photographer, and AI explorer. As a photographer, he likes focusing his lens on architecture, nature, and art of all types. But he also enjoys exploring design through generative artificial intelligence. David uses Midjourney and Stable Diffusion as his primary AI tools. Check out his AI-generated architectural designs at @midjourney.architect.

🖼 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!

Reply

or to participate.