Muscle Bears with Chinese Floral
From time to time, I go on a “stylism” hunt. I look for interesting styles and design motifs, and use /describe to find the core keywords MJ use. This is how you can learn how Midjourney thinks, and here’s how I approach it. For example, you give it an image, and it says:
A man in the style of A, B, C, D, E.
I would then generate that exact prompt to look at the result of that combination, but also split them up individually and run:
A man in the style of {A, B, C, D, E}
This is equivalent to running five separate commands:
A man in the style of A
A man in the style of B
A man in the style of C
A man in the style of D
A man in the style of E
Why would you do this? To IDENTIFY the actual keyword that influence the final look of the image. If you don’t do this, you would never actually know what is influencing what, and then it would be impossible for you to then go create your own look.
Note that using {A, B, C} like this does require fast mode. Relaxed run quite fast anyway so I usually use my fast mode to run these stylism runs or if I’m on a deadline. You can do it manually also.
After looking at the results, which I don’t usually upscale, I would then also run a standard prompt combined with a new style that I’m experimenting, e.g.:
muscular 40yo italian man in the style of A, C, D
Doing standard prompts with the same controlled subject is the best way to see what the styles are doing. Use age and race to limit variance from MJ trying to be culturally diverse.
Then you can use those styles to create final works that involve actions, compositions, scenes, and always get a reproducible results. Building up on Midjourney’s keyword library will help you go really far besides being really fun.
These three sets are from those stylism hunts. Can’t you tell?
Text prompts in Midjourney v5.1
