There are really beautiful images here!
I actually used Midjourney in my latest drawing! This was the AI generated image I based my art on:
The trippy, fever-dream weirdness of AI is a fantastic source of inspiration, imo
That's really great art !
I tried stable diffusion lately, it's different but it allows other ways to create art (and also I can use it on my own computer, and it's free). I still have to work on it, but here is one of the pictures it produced that I really liked, as you said it has a "trippy fever dream weirdness" vibe.
Don't know if this counts or not, I am working on a campaign in Hebrew, using Ars magica as a base for a Harry Pooter-based game, set in a minor school in Israel.
I have created the crests for the five houses in the school.
House of Adam
House of Eve
House of Lot
House of Sheba
House of Solomon
Love the Hoopoe and Peacock!
Next step in my "ars magica art by AI" journey, img2img...
You feed the AI with a picture and a prompt, and it tries to produces something using both.
I used stable diffusion v1, and gave it some very crude drawings done in paint (it's quite ugly but also quite fast to produce, and I don't own photoshop at the moment)
It's supposed to be a witch in blue robes wielding a wooden scepter. Yeah I know, it's really awful
That's what the AI produced first :
Not bad, but I had to describe it next as a "female" witch in the next prompt because this was not what I expected
This was better, but I guess it produced weird stuff with the staves because I also tried to make it generate a forest background, which does not exist in my original picture.
Notice how it struggles with hands : this is an issue it has in common with midjourney AI.
It also produced several other versions of old women in blue bathrobes wielding various wooden sticks in one or both hands...
Next I decided to try the AI on a scenery, drawing this as the interior of a roman temple that could be of use for my next game :
Once again I apologize for the very crude level of my drawings
Once again stable diffusion produced various versions, this time it was overall really nice in my opinion
Final tests, as I need some art for a future meeting with the mages of Atsingani : Vivianna as a fortune teller, and her faerie wagon. I still need to do something for Jacques of Flambeau (from the Lion and the Lily)
I hope this is not too boring, and sorry for the long post, but I really find this new technology quite fascinating. It's still very early, but it means that soon anyone could be able to produce nice art with minimal efforts, the potential is just huge !
That is great, i am quite impressed by the img2img stuff.
Also, the "temples" look like some cool themed sauna...
Damn! The img2img stuff is really cool and may be the thing that finally gets me to set up an account and start paying for one of these AI art services. Even without that I think it significantly beats what I’ve done in the past with just trying to find a “close enough” image through google.
I don't think I'll be able to use it (I'm nearly near as good with drawing, with or without paint, as you are), but I really appreciate you testing these and letting us know of the results.
Stable diffusion is a free install : a friend of mine that is also testing for his own purposes (probably also for roleplaying materials) paid for the first level of midjourney subscription (about 10$/month). We compared some of our results. As an example, he tried to fed it with the same wagon drawing and prompt, midjourney produced inferior results more in the style of child drawings than in the style of medieval illumination that is specified in the prompt. But overall I think midjourney seems to be able to produce bigger and nicer pictures with more detailed backgrounds. Maybe my computer is just not powerful enough for stable diffusion to do better at the moment : if I try to ask too much in the settings, it just crashes.
So my advice would be to try midjourney trial account first to have a go at the whole thing, then if you find it useful and don't want or can't install a stable diffusion instance on your computer, then you can consider if it's worth a 10$/month fee for you (that's about 200 prompts/month if I remember correctly).
You're right, but it can be used as a kind of roman aquatic sanctum could look like that as well ! I think it comes from the blue/teal colour I used in my drawing : I made some more tests that I will share soon where I used other dominant colours, and it seems to influence the end results a lot.
This was in no way a criticism. I'm a sauna enthusiasts, and since the romans brought the "bathing culture" here along the Rhine, many nice saunas and other "wellness" focused businesses in my region take on the trappings of Roman deco.
"Notice how it struggles with hands : this is an issue it has in common with midjourney AI."
A universal problem, even for artists
On that theme, I read that this morning.
Went to my best friend's place, played some Solasta (In short, DnD dungeon crawler).
Which reminded me about something I've been wondering for years.
There are all kind of 3d video games in which you can customize your character's appearance, sometimes to an amazing degree, and put equipment on it.
Yet, as far as I know, not only to they lack an export feature, no one has ever done "just" a character generator. Which feels weird: Since they do it in the first place, there must a a public for this.
I remember spending lots of time on SoulCalibur character creator, just designing people and imagining backgrounds. I can't have been the only one
Why, oh why, don't the gods smile on us, drawing impaired folks?
There's Hero Forge, which can also provide nice coloured 3D images of heroes, from which you can take snapshots.
Here's an example: