Playing around with prompts: creating a man casting magic.

One of the more difficult processes of working with Midjourney — and I assume, other AI Art generators — is crafting the perfect prompt in order to get what you actually want.

Being someone who is more into fantasy than anything else, I've struggled quite a bit trying to get my thoughts organised in order to get the images generated just as I want them. And I'm often not entirely pleased, because I'm a perfectionist.

In this post I'm trying to set this basic scene:

Tomias the mage is casting a spell, protecting both himself and Kelsey from an incoming attack. He's supposed to be standing tall and brave, casting the spell, and she's supposed to be hiding behind him, cowering.

image.png

No matter what I did, all I could achieve were the two people just standing there, or walking down a path. Yay, two people! Doing exactly what I don't want them to do.

So I thought that I would focus on one aspect at a time and try to figure out how to do what I want to do.

image.png

I started with Tomias, casting a spell. And once again, all I got was the man just standing there, doing nothing.

image.png

To get rid of the silhouette in the distance, I decided to implement a "close up" in my prompt. There appears to be magic swirling around this very unpleasant-looking man, but he's not casting the magic.

image.png

This time I put my scene modifiers first — full body, close up, cinematic lighting — and then stipulated the man to "cast a spell from his hands."

The bottom right image was perfect! Kind of. I wasn't after fire magic, just a spell, preferably purple because my brain imagines "arcane" to be purple, and I want him to be casting a protective arcane barrier shield thing.

image.png

So I upscaled that bottom image and tried to work with it some more. This time trying to turn the magic into something purple. It wasn't working too well though. Either one hand held purple magic and the other fire, or the purple magic was just relegated to the background.

image.png

So I tried to do it again with the command: --no fire

And it didn't work, he was still summoning fire.

image.png

So I tried blending the image with one that was holding at least one purple magic with another randomly floating around. xD

image.png

Which just resulted in more one purple/one fire.

So I went back to the start and recreated the first prompt that had delivered me the first man wielding magic properly, this time specifying the purple magic.

image.png

full body, wide-angle, cinematic lighting, fantasy:: a man with black wavy hair and orange eyes casts a purple spell from both hands --c 40 --aspect 3:2 --no fire

A bit more of a success this time! I rerolled it a few times to see what else I got. I started getting more and more guys with angel wings? So I had to change my prompt to --no angel --no wings as well as --no fire.

image.png

I got this dude who looked pretty cool, so I thought that I might merge him with one of the other guys, and this time get my forest background back.

For my forest, I find that "the background is a green forest that glows green" works quite well for what I'm after, so I popped that in.

image.png

Anndddddd, his magic is mangled again. xD

But at least he's casting it. In a forest. Okay. We are partially there!

image.png

After a few re-rolls, I got this! Getting better! Unfortunately the art is starting to become a bit distorted now, though. Anything I do now to further this particular image results in severe grain and jank.

So I thought that I'd go back and blend the above image with a couple of previous ones.

image.png

Cool, cool. The forest is starting to fade away though, and now I would like the woman to be included in the prompt, too.

So I insert the image and input the following:

full body, wide-angle, cinematic lighting, fantasy:: a man with black wavy hair and orange eyes wearing a silver robe casts a purple spell from both hands, he protects a woman with long red hair who cowers behind him, scared:: the background is a green forest that glows green --c 40 --aspect 3:2 --no lantern --no torch --no fire

image.png
image.png

...and get some "interesting" results, haha.

At this point, I just keep re-rolling and hoping that something will just randomly pop up. Unfortunately, faces and such are becoming mangled and distorted again, so mostly I'm just hoping that I'll get something where the two people are in the correct positions and then I can work off that image.

image.png

I did get this, which looks interesting. She's not exactly cowering, but she is standing at his back and he kind of has a protective air around him.

I figure that this image is decent enough to work off of, and reintegrate it into the prompt, this time trying to add some "scared" and "afraid" modifiers to the woman. I'm not too happy with the man's face and hair, but I'll get what I can get at the moment.

image.png

I like this one!! His hands aren't what I intended, but he's still clearly casting a purple spell. I save this image and reintegrate it into the same prompt, again. I have no idea if this is what you're supposed to do, but it seems to work at clarifying images for me, anyway.

image.png

This one's a bit dark and overly contrasted, and he's not wielding his magic, and there's a rando in the background. But their positions are much improved. I decide to blend all three of the above images together, and reiterate my same prompt, again.

The result isn't too great. So instead, I mix the above image with a previous one of the man wielding the purple hands, and see what happens. I'm feeling quite optimistic now, like I'm actually getting somewhere with this!

image.png

full body, wide-angle, cinematic lighting, fantasy, less contrast:: a man with black wavy hair and orange eyes wearing a silver robe casts a purple spell from both hands, he protects a woman with long red hair, she cowers behind him scared afraid:: the background is a bright glowing green forest that glows green --c 40 --aspect 3:2 --no lantern --no torch --no fire

I tried to add "less contrast" in, hoping it would lessen the amount of black in the image. Didn't notice much of a difference. What I am noticing, however, is my woman is now starting to cast magic. Not my intent. And things are getting distorted again.

image.png

I did reroll the previous one though, and thought that their faces and bodies in general were much improved!

After I got the above image, I also got this message in the Discord:

You have been awarded 1 free fast hour of GPU time for making it to the Top 1000 users who've rated images in the gallery today!

WOO! More hours for me! xD

image.png

I'm mostly happy with the above image, except she doesn't seem scared and I'd prefer more magic in the man's hands. So I tried one more time to blend a couple of things together.

aaannddddd........

image.png

Damn it. My girl is wielding magic again. And she's becoming distorted now. Cries in distortion.

In an annoyed fit, I inserted my nice image again and stipulated:

the woman looks scared, the man is holding a large ball of purple magic:: background is a green forest that glows bright green --aspect 3:2 --no fire --no lantern

image.png

I seem to have returned to square one. xD

While I didn't get exactly what I wanted, I did end up with a few images I rather liked.

image.png

Most notably, this one.

If I had more experience with Photoshop and things, I could probably manually change it a bit. Add more purple flare to his hands, and change her face to look a bit more frightened. Unfortunately, I don't have Photoshop and neither do I know how to do such things in Paintshop Pro, my preferred program — I only really use it to lightly perfect photos, not make massive changes.

It would be something to dabble with though, for sure! I really like that above image. Everything is on point except for a couple of details. And they both look reasonably how I imagine too.

It's pretty cool.

 

Until next time! 😊✨

 


 

All images in this post are courtesy of me, @kaelci and were created with Midjourney.



0
0
0.000
14 comments
avatar

These are excellent creations. The last one could be the cover of a magical romance novel. He's on a mission to rid the world of evil. She's on a quest for... love.

image.png

0
0
0.000
avatar

Haha! That's basically it, too. XD

Well, she's on a quest to make a new life for herself in a new world; he's on a quest to end the world-ruining rituals of a demented Magister.

So basically love finding and evil thwarting, haha!

0
0
0.000
avatar

Heh. That's awesome. The images nicely capture it.

0
0
0.000
avatar

Need to try out this Midjourney. Do we have to pay to use it or is it free? I'd love to use this and create some etheral images.

0
0
0.000
avatar

If you sign up to their Discord, you'll have about an hour to play with it.

It's not 60 minutes of our time, but rather 60 minutes of "GPU computing time" -- you should be able to play around with 25 images or so during that time. 😊

After that though, you have to subscribe to keep playing with it or spend a bit of time in the gallery voting on other people's images. If you vote on enough images, you'll get a free hour of GPU time again.

0
0
0.000
avatar

I need to delve into all the new features Midjourney has added in the last few months. I have some designs that I'm gonna get printed and framed for my new nerd room.

0
0
0.000
avatar

Apparently v5 of Midjourney is coming soon! 😁 I had a look at a few of the examples and the images look really cool. Looking forward to playing around with that when it's released.

0
0
0.000
avatar

It seems like you've got a tad bit more patience than I do! I don't fancy redoing the images over and over and over 😂 Its a bit of a mood killer to me, laughs I want shit done now! But I think it's best to start fiddling around with it.. The more you fiddle the sooner you will be able to easily create images

0
0
0.000
avatar

It's part of the process! xD Part of the puzzle that needs to be worked out to accomplish one's goal.

0
0
0.000
avatar

Midjourney's AI softer is impressive. The pictures you presented are amazing. It's difficult to believe that you provided only a small amount of information to get it developed. I'm learning to use Nite Cafe's AI site myself. I've got some hits, but they're mostly misses as I'm trying to learn which words produce the images I need.

Anyway, thanks for sharing your work. Have a great day.

Posted using Proof of Brain

0
0
0.000
avatar

It's definitely an endeavour trying to utilise the correct words, phrasing, and idea separating in order to paint the picture you're after. 😅 But it's a learning process and I find the more you play around with your idea, the more the AI learns what you're trying to accomplish, too!

0
0
0.000
avatar

It's definitely a learning process. And it's also amazing how adaptable the AI platforms are today. I can't only try and imagine what these platforms will be like in the future.

0
0
0.000
avatar

Great images,

its so tricky even IF you get the perfect prompt sometimes you don't see that it is correct until a few generations or images. So, as painful as it is i always ask for a few rerolls everytime. You never know where the perfect image is hiding

happy creating

Sarah

0
0
0.000