Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump

[EatYourVeggies]

Dumping cool generated images:

"ghost, woman"



"man forced to eat vegetables"



"dark souls"



"ruins of ancient civilization"



"dragon attacking city"



"mythical evil duck"



"slayer of demons"



"virus infecting cells"

 

[Catgirl Tyranid]

"slayer of demons"

This might be my favorite image that's come out of this thread. A fire breathing angel of vengeance holding the ripped off arm of one of its foes by its sinews. I don't even care that it's missing an arm since I can chalk that up to battle damage. And is it just me or does it have dinosaur skulls for wings?

 

[CwMkCW]

I have done a bad thing

Spoiler: autism

 

[️ronic]

Prompt:
"supreme gentleman"


Results:

---

Lovecraftian Nightmare, by Norman Rockwell:

 

[Puff]

prompts space, Adrian Smith, space battle, space ships, one planet
sampling steps 30
CFG scale 12
seed -1,
View attachment 3838794View attachment 3838797View attachment 3838800View attachment 3838803View attachment 3838812

FYI, "-1" just generates a random seed. After the first one they're sequential though.

 

[CansOfCant]

I got a very interesting result using the "Interrogate CLIP" feature.

 

[WelfareNiggerQueen]

Prompt:
"supreme gentleman"
View attachment 3841497

Results:
View attachment 3841479View attachment 3841485View attachment 3841491

---

Lovecraftian Nightmare, by Norman Rockwell:
View attachment 3842448

Here are some Sir Elliot's I just made
Denoising 1

Denoising (. I was having memory issues with this one and the next one and restarted my computer to fix it hence the black squares (wasted electricity my bills hnnnnng)

Denoising (.75)

Denoising (.5) I can't get this picture to upload. It keeps breaking.
View attachment 3843282

 

[Baraadmirer]

Thought I'd see what Nathan Fillion looks like in NovelAI:


And Marcia Gay Harden:

 

[AmpleApricots]

I wonder what the implications are of these pictures flooding the net in regards of future AIs being trained on piles of scraped images from the net. I'm no statistician but I feel there's some poisioning-the-well/entropy kind of problem in there somewhere.

Then again, I guess the pictures are pre-filtered already to begin with. There's probably a big number of pictures people decide to not show the world for the ones they actually upload e.g. here.

 

[Abeezo]

I am absolutely addicted to this shit now boyos, all I do in my free-time now is generate images

 

[Baraadmirer]

Behold, a young Gilbert Gottfried (rest in peace):

 

[Involuntary Celebrity]

I wonder what the implications are of these pictures flooding the net in regards of future AIs being trained on piles of scraped images from the net. I'm no statistician but I feel there's some poisioning-the-well/entropy kind of problem in there somewhere.

Then again, I guess the pictures are pre-filtered already to begin with. There's probably a big number of pictures people decide to not show the world for the ones they actually upload e.g. here.

Probably not that major. You don't scrape indiscriminately and there's a pre-classification step too (methods for which will also continue to improve) so if they suck they're basically in the same metaphorical bucket as whatever low quality amateur art makes it in there that nobody wants to gen anyway, and if they infiltrate sources you aren't expecting to contain non-human art then they're probably close to being indistinguishable anyway which is part of that self-selection you mention, yeah. In terms of feedback effect, I'd guess the bulk of art is from generic artists nobody could name who are imitating one another anyway, so it's probably either neutral or possibly even a positive effect insofar as it drifts away from that (especially given the benefit of users selecting "creative" flukes that stand out).

Validation schemes are a whole separate thing, there's a bunch of them and they do stuff like try to index human aesthetic preference, and those'll keep developing in their own direction too. Ultimately a self-training process (like how chess bots are trained by playing against themselves) might even take over entirely? People can only make so much art just like there can only be so many real chess matches in the book; imagine simulating a thousand years of speculative artistic development according to some art↔preference model in a day.

 

[Starved Artist]

Spoiler: NSFW

My wife has been running prompts and has been using it more than I have and essentially, stable diffusion is her realm now. I only exist to update, merge and troubleshoot.
I'll spare you guys the story of her 40K+ renders of everything that pops into her head.

Has anyone mastered clothing/dressing two separate characters without in-painting corrupting the fuck out of it?



The objective is to create a female knight holding a princess, and gently kissing each other. This she has managed to do.

The issue is the skin color, clothing and height. Whenever one of those is and I quote "JUST FUCKING PERFECT WHY CAN'T IT DO THE REST RIGHT", the other is absolutely wrong. Even when negative prompting men, it randomly turns one of the women into a feminine dude which absolutely rustles her jimmies. Or gives them both dresses or it gives them both armor.
The Princess is supposed to be fair skinned/ tanned/ lighter than chocolate black, short with wavy hair.
The Knight is supposed to be Pale and tall, with red wavy hair and wearing knights armor.



In experimenting on how to do this, I've tried the novel Ai specification route and have gotten nowhere.

.

Anyone can figure out how to specify exactly what to change?

 

[Puff]

My wife has been running prompts and has been using it more than I have and essentially, stable diffusion is her realm now. I only exist to update, merge and troubleshoot.
I'll spare you guys the story of her 40K+ renders of everything that pops into her head.

Has anyone mastered clothing/dressing two separate characters without in-painting corrupting the fuck out of it?

View attachment 3846015

The objective is to create a female knight holding a princess, and gently kissing each other. This she has managed to do.

The issue is the skin color, clothing and height. Whenever one of those is and I quote "JUST FUCKING PERFECT WHY CAN'T IT DO THE REST RIGHT", the other is absolutely wrong. Even when negative prompting men, it randomly turns one of the women into a feminine dude which absolutely rustles her jimmies. Or gives them both dresses or it gives them both armor.
The Princess is supposed to be fair skinned/ tanned/ lighter than chocolate black, short with wavy hair.
The Knight is supposed to be Pale and tall, with red wavy hair and wearing knights armor.
View attachment 3845910View attachment 3845967View attachment 3845988


In experimenting on how to do this, I've tried the novel Ai specification route and have gotten nowhere.

View attachment 3846009.

Anyone can figure out how to specify exactly what to change?

I'm going to level with you. Your wife is going to be the expert at prompting over anyone here.
I guess my advice would be that over specifying breaks stuff that you don't notice. Start removing tags and emphasis while generating on the same seed and seeing what changes. Basically get rid of anything not significantly changing the image.
Further, if you're not already, use image to image to remove some complexity. You can almost eliminate posing prompts by starting with something with good posing and colors.

I am absolutely addicted to this shit now boyos, all I do in my free-time now is generate images

I wouldn't go that far, but it's become my "listening to a podcast" or "pretending to pay attention to a conference call" activity.

 

[Woe-B-Gon (TM)]

I am absolutely addicted to this shit now boyos, all I do in my free-time now is generate images

Ah, my favorite anime, LNKJU. The Etaoin Shrdlu arc was awesome, so much better than that fgsfds filler arc they stuck in so the manga could catch up.

(Yeah, lazy joke, AI can't letter, etc.) My actual question: Any idea what that image was trying to emulate with the oil spill in the corner? Some attempt at an issue number or title I'd get, but it looks almost like some kind of livestream chat sidebar gone wrong.

I wouldn't go that far, but it's become my "listening to a podcast" or "pretending to pay attention to a conference call" activity.

Now there's something I could never do, I mumble too much. I don't think I'd want to explain why midway through the head of HR complaining about something no one in the call has any control over I suddenly asked what I need to do to make women have less than 3 boobs.

 

[There Is Light At The End]

novelai

 

[️ronic]

Here are some Sir Elliot's I just made
Denoising 1

Denoising (. I was having memory issues with this one and the next one and restarted my computer to fix it hence the black squares (wasted electricity my bills hnnnnng)


Denoising (.75)


Denoising (.5) I can't get this picture to upload. It keeps breaking.

Vanilla Eliot:

Gentlemantron Supreme:

Chad Rogers:

Anime Chef Rogers:

Better Call Eliot:

Vault-Tec Employee 1NC311:

Norman Rockwell Universe Variant:

 

[Baraadmirer]

I once again tried to make a character that wasn't eating something with their hands and I succeeded... sorta. Used NovelAI.


Has anyone noticed when using NovelAI, the curves are less smoother than if you don't add anime to the prompt when using img2img?

 

[bifftango]

The somewhat-a-hell-ride continues...

Spoiler: Tacos!

"How many heifers are on the head of a needle, depth of field, crimp, cool"


"How many heifers are on the head of a pin, depth of field, vivid"


"The gloaming, vivid, inspired"


"Taco Tuesday ruined by acid rain on the surface of Venus, gruesome, vignette"


"Pinhead eating a Tootsie Pop, big tongue, morbid"

 

[GermanPotato]

I fucking love this thing. Still scary, but fun as well.