Welcome to my 2nd AI art post
I've been exploring Alien Art Hive to find the cool art I like, and decided to give it a go. Let me know what you think.
I saw this post and I was wondering how I would even approach something like this. The author says he did photobashing and a lot of editing. I'm only using Stable Diffusion, my freedoms are limited to prompt, negative prompt, model, size, etc.
If you see his general approach to art, it's filled with big objects that "pop" despite their abstract weirdness. I gave it a few hours to try and reproduce something similar. I failed, but I got other cool results I liked despite being somewhat different.
First generations
At first, what I tried was a naive approach. I went with a simple prompt that kinda evoked all the elements I saw in the original art + some elements I thought would be cool to add (chiefly fluffiness & bubblegum)
1. Egg in bubblegum machinery
First prompt:
intricate details, inside bubblegum machinery, mucus, alien egg, fluffy carvings, honey lightning, fantasy forest, purple sky, electric storm, blue fire fog, metal veins
First model: Cheese Daddy's Landscapes mix v3.5
Note: Even though the recommended VAE for CheeseDaddy's Landscapes Mix is kl-f8-anime, I find it to be less realistic, and less alien, because it's more saturated. It's less like a picture taken of a weird world and more like funky art. The difference is minimal but I embraced vae-ft-mse-840000-ema-pruned instead to get some more fogginess. Not using a VAE gives too gray results.
I really liked this first image, and in general the fact that I put electric storm, mucus and bubblegum together makes it so there's a lot of gray solid clouds, either surrounding things in the form of a mucous texture, or creating a solid sky as you'll see in the second result.
The following one was absolutely fantastic. I love that thing on top that gives off yellow light yet seems like a body part, yet is part of an architectural design, yet seems so soft. It's like a giant fluorescent stomach.
2. Girl in egg (MeinaMix)
When I tried the first prompt using MeinaMix, I realized it's chiefly for anime people, doesn't work that well for landscapes or abstract creations. If it finds a pattern that could perhaps be a person, it turns it into a person.
It wanted a girl inside the eggs so I allowed it to do it and helped it to do it, because why not, girl in egg, don't miss the weird chances of adventure life gives you.
Method: original prompt for 20 512x512px images, I grabbed one with a girl in an egg, and gave it the second prompt for a 1024x1024px img2img:
intricate details, inside bubblegum machinery, mucus, alien egg, fluffy carvings, honey lightning, fantasy forest, purple sky, electric storm, blue fire fog, metal veins, girl inside the eggs
Adding more machinery and H.R. Giger to the mix
(The model is the same as the first one, CheeseDaddy's landscape mix v3.5)
Giger is famous for artwork similar to that of @julesquirin in the post I shared above. In the first results, things were too fluffy and bubblegummy. And there wasn't enough machinery! So I decided to push things a little further with a new prompt:
intricate details, inside bubblegum machinery, mucus, alien egg, (alien machinery, giant wheel), fluffy carvings, honey lightning, fantasy forest, purple sky, electric storm, blue fire fog, (metal veins by H.R. Giger)
A bit more sci-fi solid realism with Protogen x5.8 (scifi+anime)
This protogen model is designed to produce hyperrealistic sci-fi images, but if used properly it can do amazing stuff. I find it to be only second to CheeseDaddy's Landscape Mix (CDLM) in landscape quality, and that's only sometimes. In this case, I didn't start with it because I thought CDLM would do a better job. The following are 5122 CDLM images fed to 10242 Protogen (img2img). The results honestly astonished me, especially the first and 3rd one, due to their cinematic quality. It reminds me a bit of my first tests with Dreamlike Diffusion.
MeinaMix + Protogen crystal egg machinery
This one is done with MeinaMix, it's a 5122 miniature that contains a blurry person, not very useful, and then passed for a 10242 highres fix with Protogen
Conclusion & Inspiration
I had never tried anything like this. I have to give some super thanks to @julesquirin's beautiful art without which I would not have even thought of trying it. It had been bugging me for months that I chiefly make person designs with Stable Diffusion and I don't do enough landscapes and abstract art. Here's to changing that!
Thank you, Jules, for your art and inspiration today.
A Portal To A New World ( + settings used) by @julesquirin
Greetings, artists from all worlds and all realities!
Here are some more designs from the depths of my dreams!
