Starting point
Just to prove a point, I chose to do a bad scribble - to see how much guidance did MJ need to make something out of it.
I took a photo, 'fed it' into the AI and prompted it: "man in a cape wielding a sword"
The results were a little terrible, so I refined it. I re-rolled and prompted "[IMG reference] man in a cape wielding a sword, ray of sun on him, dark in the background, chiaroscuro, expressionist style".
Now for some environment, so I re-rolled: "[IMG SRC] man in a cape wielding a sword in a gothic castle, ray of sun on him, dark in the background, chiaroscuro, expressionist style"
[IMG SRC] man in a cape wielding a sword in a gothic castle, ray of sun on him, dark in the background, chiaroscuro, expressionist style --ar 4:3
Now I want to start working on the 'look': I want a painterly, loose visual.
[IMG SRC] man in a cape wielding a sword in an intricate gothic castle, ray of sun on him, dark in the background, chiaroscuro, expressionist style, concept art, loose broad strokes and digital paint splatters --ar 4:3
Annoyingly, these guys started growing horns, so I added --no batman.
A few re-rolls forward and this is what the prompt looks like:
"[IMG SRC] disbehevelled man in a shredded cape wielding a sword, jagged clothing and rough armor, in a grand intricate gothic castle with tall arches, a dramatic bright ray of sun behind him, high contrast shadows, diffuse mist, golden hour atmosphere, dark in the background, chiaroscuro, sharp expressionist style, concept art, sharp strokes and digital paint splatters, variety of colors, high detail face, mixed media, 4K, epic wallpaper style --ar 2:1 --s 100 --chaos 18 --no batman, text, watermarks"
The exact same prompt, but with an aspect ratio of 1:2 instead of 2:1. Basically, switching from landscape to portrait.
I'm hoping for more texture, so I find a cojuple of images that might influence my painting.
I pick these two images to add to the IMG2IMG mix.
After a few re-rolls and tweaks, I end up here. It's good enough, for the purpose of this tutorial. I would keep refining the style, photoshopping a wand and maybe a cape, perhaps turn him into a wizard... this piece did not have a briefing, so it's really open for exploration.
This is where I stopped with the other fork of this image. Prompt:
"disbehevelled man in a shredded cape wielding a sword, jagged clothing and rough armor, in a grand intricate gothic castle with tall arches, a dramatic ray of sun on him, high contrast shadows, golden hour, dark in the background, chiaroscuro, sharp expressionist style, concept art, loose but sharp broad strokes and digital paint splatters, variety of colors, high detail, mixed media, 4K, epic wallpaper style --ar 1:2 --s 100 --chaos 18 --no batman, text, watermarks"
To take this further, I would keep working on the painting strokes; I'm not happy with how the painting looks. I would also add other elements, since right now this image has no meaning: it's just a knight-type in a castle, but you can't tell why he's there, or what he's about to do. There is no story.
Which is fine! This was an experiment on Midjourney V4 to see how to take a scribble from concept to built-up image. :)
Enjoy!