What happens when Midjourney v6 chooses all the details of an image?

I recently read a Medium post by hejrene titled “I let Midjourney v6 go wild — here’s what happened”. I was absolutely fascinated by the concept presented of letting Midjourney unleash it’s creative abilities by letting it fill in details that you don’t specify.
hejrene presented a very generic prompt used to demonstrate how Midjourney can come up with its own details:
/imagine prompt: the most [fill in the blank] photo ever taken
This prompt is very basic, but importantly it leaves out all of the details that we’re so used to giving Midjourney to direct its generation of the images we prompt for.
In my last Medium post, I used the following prompt to demonstrate the point I was trying to make in the post:
a carefully executed ink and colored pencil drawing of a surreal scene of a barnacle-covered whale with a cluster of mushrooms growing from its back, mind-bending forms, surreal shapes, made of all of the above, clever use of negative space — ar 3:2 — v 6.0 — stylize 0
Unlike the prompt hejrene used, this prompt is all details telling Midjourney exactly what to create.
The beauty of hejrene’s prompt is that it leaves all of those details open-ended for Midjourney to decide on its own.
In the post, hejrene filled in the blank in the prompt with adjectives like “fun”, “epic”, and “adventurous”, all of which produced awesome results!
For the 48 hours after I saw hejrene’s post, I was completely entrenched in imagining every adjective I could think of, over 800 and counting! The collage you saw at the very start of this post includes the first 192 prompt results I imagined (since making the collage I prompted with another ~600 adjectives to make an even larger collage that I’ll share another time)
I’ve included the top 162 images below that Midjourney produced from my list of adjectives. These 162 images are ones that I feel most closely aligned to the adjective in the prompt. Many of the adjectives I used produced images that were totally nonsensical and not related to the adjective at all, but many of them were adjectives that would be really difficult to visualize, like “calculated” or “unheard”.
One of the things I really enjoyed about this exercise is that it demonstrated how AI like Midjourney “understands” the human experience. When the AI gets to choose the subject and the details of an image based on a simple adjective it’s given, it really showcases what that adjective means to the AI. For example, when I think about the word “compassionate”, I think about humans tending to and caring for other humans. All 4 of the images that Midjourney generated for the word “compassionate” involve humans caring for animals. This is obviously a completely valid interpretation of compassion, but not the one my mind immediately went to.
Have a fun scroll through the images below, and make sure to leave comments about any trends or details you notice, or any other things that stand out to you! For each image, I used hejrene’s prompt, “the most [fill in the blank] photo ever taken” where I filled in the blank with the word in each image caption.
This was a super fun exercise, and I highly suggest trying it out for yourself!
Thanks so much for reading, make sure to leave a clap and follow me for more fun content like this!
Enjoy the images!




































































































































































This story is published on Generative AI. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories. Let’s shape the future of AI together!
