The Prompting Revolution: Adapting Your Approach to Midjourney V6’s Creative Evolution

Beginners in AI art start here.

If you are not a beginner, be prepared to let go of everything you know about prompting. It no longer works. You need a beginner's mind to get the most out of Midjourney V6.

The list below comes from the announcement of V6 by DavidH:

Style and prompting for V6

Prompting with V6 is significantly different than V5. You will need to ‘relearn’ how to prompt.

V6 is MUCH more sensitive to your prompt. Avoid ‘junk’ like “award winning, photorealistic, 4k, 8k”

Be explicit about what you want. It may be less vibey but if you are explicit it’s now MUCH better at understanding you.

If you want something more photographic / less opinionated / more literal you should probably default to using --style raw

Lower values of --stylize (default 100) may have better prompt understanding while higher values (up to 1000) may have better aesthetic.

Total beginners start with the documentation, but beware that as of 9th Jan 24 it only goes up to version 5.2 and has not yet been updated for V6

Settings

/settings will get you to this page in Discord

I begin all of my projects with the settings shown and occasionally alter the stylize setting once I am getting the composition I am looking for. As stated by DavidH above, this gives you the best chance of having your prompt followed accurately.

Avoiding Junk

You will find a lot of junk if you copy other people's prompts. This creates extra noise in the images and distracts MJ from giving an accurate interpretation of your prompt.

I just went to Explore on the Midjourney website and found this prompt used recently.

A hyper-realistic photo of a beautiful young ginger woman whiteviolet eyes, high-detailend, 35mm, 8K,

This is the old way. It's noisy, and a lot of it is unnecessary. The noise is heightened by typing mistakes.

Be explicit

A beautiful young woman with ginger hair and violet eyes

The default in V6 is a realistic, detailed photo style, so you no longer have to prompt for that. Just say what you want to see. I liked this image best, even though the eye colour was incorrect. So, I ran the same prompt again but added the image above as part of the prompt.

Out popped a similar young woman but with violet irises. This is a simple use of a reference image, which is the standard way of getting a consistent character — recycle a character image with its prompt.

For this version, I added a violet top.

A beautiful young woman with ginger hair and violet eyes, violet top

Creating a full-body shot from a portrait

To get full body shots, you need an aspect ratio to suit, which in the case below is ar 1:2

The first image was created from text alone. Various things are helpful for getting full body shots in the text, like saying barefoot or describing shoes.

The second was created from a headshot. This is a good method if you want a character to be fairly consistent.

The process for the second full-length image was to use the same prompt several times but use an image from one generation as a prompt for the next. This gradually draws the camera back.

You can also use Zoom and pan, but results are not so consistent yet.

Keeping scene and character separate

It is very tempting to pile everything into your first prompt, but the danger is that it all gets too complicated, and you can’t work out where you are going wrong or what you need to change.

So rather than saying my young person was on a beach, I got the character correct first, then created a scene and brought them together in a prompt. I created a tropical beach.

I used this image and the first full length image and combined them in a prompt

photoshot on a tropical beach, a beautiful barefoot young dark skinned woman with ginger hair and violet eyes, wearing a light violet linen dress

giving me lots of possibilities

Creating scenes and characters separately also has the advantage of changing the scene with similar characters. Use any scene image and your character and include the prompt for both

A full length image of a beautiful barefoot young dark skinned woman with ginger hair and violet eyes, wearing a light violet linen dress, Hyde Park London on a Sunny day, many people walking and relaxing

Fantasy elements

Midjourney is great at fantasy-type images.

I ran a series of prompts for doorways into space using a variety of wood carving styles.

A door opening into Nasa deep space, surrounded by a massive [carving style] door frame

I googled carving styles and tried out the Scandinavian chip, Hosogiri, Celtic, Chinese relief and Gothic. I also cranked up stylize to 800 for these because I am looking for something fantastic and aesthetically pleasing rather than accurate.

Feel free to grab the images and the prompt and create your own mashups of these. I couldn’t resist putting my lovely young lady in the doorway. The first with image prompts, the second without.

A full length image of a beautiful barefoot young dark-skinned woman with ginger hair and violet eyes, wearing a light violet linen dress, A door opening into Nasa deep space, surrounded by a massive Hosogiri carved door frame [ stylize 1000 ]

Cute animals

The prompt here is general, and MJ will make its own choice of which animal to create. You can, of course, specify an animal if you wish.

I focused on giraffes and changed the style to needle felted.

Using styles

A style tuner is being developed, which should soon work with V6. This will bring in a long-awaited facility to create your own characteristic style and have Midjourney create images which are unique to you and specific to your purposes. If you currently make physical art, you will be able to feed in images of your sketches and paintings and create examples of new work following your own stylistic traits.

In the meantime

My advice is to begin creating your own style. A lot of AI art output references specific artists, and it can be a lot of fun to do that, but already, platforms that sell AI art are refusing to accept images where the prompt references living or deceased artists.

It is a great way of creating inspirational raw material. For example, I have used the prompt for the woman in the space doorway above and inserted an artist’s name at the beginning. Lets make it a quiz! See how many artists you can name. Answers in the comments.

One approach to creating your own style is to research the stylistic elements of a particular artist and incorporate some of that into your prompt. I asked ChatGPT to analyse Peter Mohrbacher’s work:

I took those headings and integrated them into my prompt to produce this:

An Etheral and otherworldly painting of a beautiful barefoot young dark skinned woman with ginger hair and violet eyes, wearing a light violet linen dress, A door opening into Nasa deep space, surrounded by a massive Hosogiri carved door frame, intricate and elaborate details, distinctive character design, rich and vibrant colour palette, surreal and symbolic imagery

The problem with a prompt with many stylistic directions is that I no longer know the effect of each separate element. I might take each element on its own to try out the effect.

distinctive character design, intricate and elaborate details, rich and vibrant colours

SUMMARY

Prompting in V6 is all about specifics. Midjourney is in the process of creating a remarkably controllable tool that has limitless possibilities. I advise that if you want to create distinctive original images, learn from scratch how to prompt specific images that you visualise with the stylistic elements only you can imagine.

Because of the limitless possibilities, there are also limitless dead ends and rabbit holes. You enter that limitless wilderness as soon as you emulate or copy other prompters.

Creating a text prompt is only the first part of the process. Refining happens when you make variations and remix images, such as when you add an “almost there” image to the prompt. For me, a final image never comes from a single text prompt. It is part of a process of combining text and images and moving towards the final article in several steps.

This final image took 14 steps, and within that, I recycled 3 images to achieve my goal.

Thanks for reading to the end. Use the comments box to ask any questions. Often, responses can inspire a whole new article.

Please show your appreciation by clapping, responding, following and subscribing. Thanks.