The highly anticipated Midjourney V6 dropped as an early Christmas present last month, and just like that, it felt like the entire AI image generating community came alive overnight. Myself included, if I’m being honest. But now that the holidays are over, I’ve had some time to play around with V6 and relearn many of my prompting habits and explore more of what Midjourney has to offer.
If you’ve spent some time on Midjourney V5, you must have already experienced the frustrating process it was to consistently or accurately prompt two or more subjects in one image. Sometimes it was downright impossible. V6 has completely changed that. Now you can prompt two or more distinct subjects without too much of an issue, going as far as to assign them specific details, clothing, appearance, etc. This is a game changer.
In my previous article, I introduced some of the new features and capabilities of V6 Alpha at launch. Now I will go a little deeper into the multiple subject prompting techniques and examples.
To get started, be sure you have turned on V6 from your --settings dropdown, or simply add the --v 6.0 parameter at the end of your prompts.
A New Way of Prompting
There are a few notable changes in the way the V6 model understands prompts vs what we’ve all grown accustomed to in V5. I think the most important being that V6 uses natural language and each prompt can now be long and detailed in a way it wasn’t possible before.
Key Points taken from clarinet’s post on Midjourney’s Discord server:
- Strive to write in simple sentences of English that have good spelling and punctuation
- V6 is capable of understanding nuances of punctuation and grammar
- V6 occasionally understands some natural language negatives like “no”
- You can specify colors, positions and other details, but do not relay too much on pronouns.
- You can prompt for more than one subject with details
- You don’t have to worry about the number of words, as much as the amount of detail in your prompt
- You can add text to your image by using quotation marks
So with all of this in mind, let’s experiment.
Generating Multiple Subjects
While V6 doesn’t require special prompting, Midjourney has suggested a starter template that should help you figure out the ins and outs of achieving multiple subjects in one image. Essentially, you will want your prompts to look like this:
Prompt: [archetypal scene] [call-back details] [setting or background details] [vibe or aesthetic]
Here are the four main parts of your prompt:
1. Archetypal Scene
Start your prompt by using generic or archetypal terms in the first sentence. Best practice is to keep it to one straight forward sentence with just enough detail to make sense in the context of your intended result. For example:
Prompt: Two lovers posing for a photographer. --ar 3:2 --v 6.0 --style raw
As you can see, a generic prompt leaves the interpretation completely up to Midjourney. However, it’s worth noting that by describing the subjects as “lovers,” Midjourney was able to understand the context of their relationship vs simply saying “two people” which may have placed some distance between them and resulted in a completely different vibe.
2. Call-Back Details
Next, let’s start adding the details. The key here is to repeat the nouns you used to set up your scene. In this example, my subject are “lovers,” so I will repeat that to clarify each subject making note of their placement in the image, their ethnicity, appearance, etc. Again, use simple, straight forward language for the best results.
Prompt: Two lovers posing for a photographer. The lover on the left is an alternative caucasian man with cropped red hair and light eyes wearing a black crew shirt. The lover on the right is a latina woman with long silver hair wearing a black dress. --ar 3:2 --v 6.0 --style raw
It missed specified eye color on the man, but I could have rerolled or tried a Creative Upscale to fix that. Otherwise, all the other details specified are there without Midjourney getting confused and blending details across subjects, which was fairly common with V5.
3. Setting or Backdrop Details
The next part of your prompt is the setting, or any additional background details you’d like to see in your image. Be as specific as you can with your vision. If the prompt breaks and Midjourney gets confused, then roll back some of the detail until you’re happy with the result.
Prompt: Two lovers posing for a photographer. The lover on the left is an alternative caucasian man with cropped red hair and light eyes wearing a black crew shirt. The lover on the right is a latina woman with long silver hair wearing a black dress. They are on a Williamsburg Bridge with New York City in the background. --ar 3:2 --v 6.0 --style raw
Ok, so there might be some questionable geography here, but that’s to be expected. But otherwise, I’m super happy with this result, right down to the overall color and aesthetic of the image. It reminds me of how I would edit one of my own shots, particularly for a fall aesthetic.
4. Vibe or Aesthetic
Finally, add anything else you’d like in terms of the style, vibe or aesthetic of your image. This is where you can reference a photography or art style, particular artist or photographer, time period, movie title if you’re going for a specific look or overall vibe. This can be as long as you’d like, and again you can always roll it back if the prompt breaks and it isn’t working.
Prompt: Two lovers posing for a photographer. The lover on the left is an alternative caucasian man with cropped red hair and light eyes wearing a black crew shirt. The lover on the right is a latina woman with long silver hair wearing a black dress. They are on a Williamsburg Bridge with New York City in the background. Vintage 1980’s photography. --ar 3:2 --v 6.0 --style raw
Let us take a moment to appreciate the contextual understanding Midjourney has. Admittedly, I tested several tail end vibes, references, photographic styles etc. I almost used a black and white but figured that wasn’t good enough an example to show how that last bit of the prompt can really shake things up. Vintage 80’s does it. Not only is the photographic quality reminiscent of 80’s film cameras, the subject’s appearance changed completely to best represent the time period.
Additional Examples
Because one example is not nearly enough, and I honestly could prompt for days once I get into the groove of it, let’s put it all together with several more examples using the above template. While this guide is specifically for multiple subjects, you can definitely use the template as a general prompting guide for pretty much anything.
Images like this make me wish V6 already had the Zoom and Pan features, but they are coming, potentially as early as next week. I for one, am super excited for that.
Three legs is clearly V6’s stylistic choice for this image. Vary Region, we need you.
This isn’t quite what I had in mind, but a pretty interesting result all the same.
This is actually a pretty good example of how a long prompt works out pretty nicely. It’s definitely longer than any prompt I’ve used in V5.2, and while I wish more of it was in focus, I gotta say I’m impressed with the adherence to the prompt.
I really like how this one came out, from colors to composition. The details got blended a little with the vase color and pattern being similar but I should have been more specific with its description.
Things got a little weird with the birds being inside the coffee shop at first, and I realized my original mistake: I did not specify that the pigeons were flying outside. Her hand is also a little janky, which Creative Upscale didn’t quite fix but I thought the original was more in line with what I had set out to do.
It’s interesting to note that V6 very much understood the assignment. Although the likeness isn’t exactly on the nose for Harrison Ford, it’s pretty damn close without the need to input a reference image.
Midjourney V6 didn’t quite nail the illustrative prompt here, but that’s in part my fault. V6 leans heavily toward more realistic and photographic styles, so it might be beneficial (at least for now) to reference a particular style or artist.
I hope this guide has been helpful! Stay tuned for new guides as Midjourney V6 updates continue to drop.