Not sure if this belongs in offensive AI or here, but I've been trying to get the science man here
to plant some mushrooms on the head of the young lady over there (long story), but for some reason it keeps giving me some, ahem, God's chosen results. I used the exact same description for the guy in both settings, I think the machine is desperately trying to say something.
As the tilapia farms of rural China fail, a Confucian scholar-bureaucrat is dispatched to see what is poisoning the waters. He seeks an audience with the wise old catfish spirit of the lake.
The Vermont maple syrup farmers go to tap the trees, but they take more than is their right. In agony the elder forest spirit awakens, terrifying the men.
This one is far deficient of my vision. What I want is a very dark style, mostly black white white lines outlining objects. The tree is supposed to be a victim, gnarled up in an expression of agony and fear, while the men are also supposed to looking on in both terror and misunderstood hate. The only real color is cascading red-amber syrup from the tree's wounds as it attempts to wrench itself from the ground, the twisting of its roots cracking the topsoil like an earthquake. It is deep forest in a deep winter night, thick snow up to the knees.
An ecstatic frontier tent revival, a choir of crawdads and a catfish preacher, at the bottom of a Mississippi swamp.
Also falls short. Have to say "Lobster" to get it to do a crawdad.
This doesn't degrade art, I think, it makes me want to learn to paint/draw again (I get frustrated drawing anything longer than sketches and seem to have no talent at all for painting) because I can see my imagination with my real eyes and not just my mind's eye.
So quick question, I mainly use the art for a folder of wallpapers to shuffle through as my background therefore I need the art to be in a widescreen aspect ratio.
My proccess has been to take the image and resize it to 480 x 270 (1920 x 1080 / 4)
Then to take that into a ai upscaler (Upscayl) and 4x the resolution in order to get a widescreen aspect ratio widescreen. Then taking that into PS and exporting it as a JPEG there so it isn't 5mbs for every image.
While this does create pretty good results, the image becomes very smooth and loses alot of detail in this process.
Is there any better way to generate wallpapers in widescreen? I've been using Dalle-2 / 3 as the results there are usually the best. I tried another (Artbot or something like that) that let me define the aspect ratio for the generation but the art was way worse, would that be the only way to retain the resolution though?
I don't think I'm going to be able to get it better than this, at least not without letting Bing cool down.
Enchantment and terror in the City of Love. She ran away from home in the Summer of 69. Now its several years on and it's not fun anymore, but it's taken hold. Sinking into her skin. Working into her being. It's hard to smile but not impossible to dance.
From my understanding, yes, that's more or less the gist of it. Things like Stable Diffusion and DALL-E are both diffusion models, with the primary differences being the number of parameters (which also affects the amount of time/hardware necessary to train, SD still took $600k, despite being "small") and the availability of their model weights, which SD provides while OpenAI doesn't, hence the possibility of running it locally (alongside the smaller size allowing it to fit on consumer hardware.)
I haven't been following Google's development as closely, but I'm assuming it also works on a diffusion/transformer architecture.
She's not that pretty, actually, nor is it a very good picture, but the point is I can make ethnic women wear period clothing that it would be impossible to get them to do IRL even if I had one as a gf.
She's not that pretty, actually, nor is it a very good picture, but the point is I can make ethnic women wear period clothing that it would be impossible to get them to do IRL even if I had one as a gf.
The traditional suit jacket has never been period correct for women. A well fitted suit is designed to enhance masculine characteristics by broadening the chest and shoulders in a realistic ratio to the wearers body.
A well fit vest is more suited for women, as the role of a vest is to accentuate the body's natural shape, and draw attention to the extremities it isn't covering. Being skinny is usually easier for women in contrast to men who have to be big to really pull off a vest well.
The key words I use to get around the anti-psychedelic filter are "bizarre" and "abstract." The exact prompt is washed out of my recent history, but I think it was short and simple like "bizarre and abstract Simpsons portrait, 1993."
"Surreal" is another useful word I found, but didn't use this time.
She's not that pretty, actually, nor is it a very good picture, but the point is I can make ethnic women wear period clothing that it would be impossible to get them to do IRL even if I had one as a gf.
If I just say "Indian" it is likely to give me an American Indian.
I came to realize I had a preference for Indian women because they're the only race that has consistently Aryan facial features with brown skin and black hair, unlike Latinas (due to being part Amerindian) having Asiatic faces. The downside in real life is that they're usually bony, frail creatures but none of these AI babes really have representative body shapes.
If I specify dark it thinks I mean Black, which I know because it changes up the eyes and lips (though keeping straight hair like an Aryan Indian woman, or a Black woman that straightens her hair artificially).
BTW, not that I would have tried this or anything, but it's basically impossible to make it show a White man in a white suit and Black tradwaifu frolicking in a haystack behind a large mansion, it ALWAYS makes the man Black, like Bing can tell what you're thinking and isn't going to indulge it. I mean, that's what my friend told me.
My stock description for clothing is "pioneer American dress" or "pioneer American clothing." I almost always use "Norman Rockwell style." These are really not much like Norman Rockwell at all, they're almost more... digital? Too perfect and doll-like like very high quality video game graphics, not really like a true photograph. It's necessary, because the AI image quality just isn't good enough to avoid the uncanny valley. Making it more like a hyperrealistic painting kind of hides that under a layer of gloss.
Those you replied to were "Victorian three-piece suit" and "toga."
Bing doesn't seem to understand the difference between "petite" as in "little girls" and "petite" as in "very short adult woman," so it's not very useful as a qualifier. Best off saying "beautiful," but it makes beautiful women by default anyways.
And yes, I AM just as hunky and handsome and farmcore IRL, 100% accuracy.
The key words I use to get around the anti-psychedelic filter are "bizarre" and "abstract." The exact prompt is washed out of my recent history, but I think it was short and simple like "bizarre and abstract Simpsons portrait, 1993."
"Surreal" is another useful word I found, but didn't use this time.
I can't conceive of why psychedelic would be banned and Bing allows it, but to get better results I have to throw in stuff like "acid trip" and especially "colorful hazy sky."
Edit: God do I hate being confined to this plane of existence. I never got into waifuism because I hate 90% of the anime art style, they don't even have noses most of the time. It's cheap soulless garbage. Miyazaki is better, a lot of the 1980s stuff had charm to its cheapness, but nothing to idolize. I don't like schoolgirls, I like peasants. I like Western art styles. Now I see it. Now I can sit before my magic mirror and see any world I want and it tastes bitter. FML I'm going to bed.
Young Jewish man with beanie, horns made out of hair, golden teeth and golden star of david chain inside an oven that is on fire
Steps: 20, Sampler: Heun, CFG scale: 7, Seed: 2812222778, Size: 512x512, Model hash: 67d531eeb4, Model: artUniverse_v80, Version: v1.6.0
For this one the prompt was "Hodag running a hot wings and onion rings food stall in a small mountain town", didn't turn out like I was hoping, but it's not bad.
