What is the best AI generated stuff you´ve seen RE Sailor Moon?

  • This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn more.

Rika-Chicchi

Staff member
Site Admin
May 7, 2009
47,230
10,323
1,665
#82
As an artist I’m not really a fan of AI art but I tried it out and I’m pretty impressed with what I could do once I got the hang of it,



The fingers are obviously still problematic, & I don't like those "ant waists," lol but yeah, overall they look impressive. :cool:
 
Likes: Jawshx

foenyanko

Solaris Luna
Sep 21, 2010
2,189
1,905
1,665
#83
Even though there's a lot of problems with it in terms of construction and anatomy, the moody and dream-like nature of the last image of Moon is really nice. I wonder whose art they ripped off sampled to achieve that look.
 
Likes: Jawshx

Jawshx

Aurorae Lunares
Dec 1, 2020
1,555
4,827
1,665
33
#84
Yeah some of the anatomy can be really off. And I’m really not sure about the ethics of it all. Plus, as someone who loves to draw and support other artists the whole situation sort of makes me feel a bit depressed but I think it could be a good tool to gain inspiration for my own art, perhaps.
 

Rika-Chicchi

Staff member
Site Admin
May 7, 2009
47,230
10,323
1,665
#85
Yeah some of the anatomy can be really off. And I’m really not sure about the ethics of it all. Plus, as someone who loves to draw and support other artists the whole situation sort of makes me feel a bit depressed but I think it could be a good tool to gain inspiration for my own art, perhaps.
Even a human artist has already seen many works by other artists before creating their own ones, which're likely influenced/inspired by various aspects in those works, whether consciously or subconsciously, & AI software are essentially doing the same thing - the difference is just that they probably have seen way more other works but still lack the knowledge in other fields, such as human life experiences, among others, to perfect their works. So the ethical issue are blurry here.
 
Likes: Jawshx
Sep 19, 2022
3,733
7,271
1,665
20
#86
Yeah some of the anatomy can be really off. And I’m really not sure about the ethics of it all. Plus, as someone who loves to draw and support other artists the whole situation sort of makes me feel a bit depressed but I think it could be a good tool to gain inspiration for my own art, perhaps.
I've actually been hearing many people supportive of artists say that AI should be used as a tool [instead of a replacement]. This included things like auto coloring /inbetweening in animation or gaining inspiration for an artist's own work. So, the way you're saying you're using it would slide for many.
 
Likes: Jawshx

Clow

Gurges Ater
Jul 29, 2012
9,607
8,030
1,665
#87
My favorite artists were born in the beginning of the twentieth century, so I naturally tend to gravitate towards their works…

That being said, I believe that artists born in this day and age will use AI as a tool … and the good ones will find ways to subvert it.
 
Likes: YuYu Yuichiro

Rika-Chicchi

Staff member
Site Admin
May 7, 2009
47,230
10,323
1,665
#88
I've actually been hearing many people supportive of artists say that AI should be used as a tool [instead of a replacement]. This included things like auto coloring /inbetweening in animation or gaining inspiration for an artist's own work. So, the way you're saying you're using it would slide for many.
Moreover, w/ good, creative prompts, the more advanced AI software can even enable ordinary people who lack the technical skills & professional training of artists to produce their own original works. :)
 
Sep 19, 2022
3,733
7,271
1,665
20
#89
So, asking out of curiosity since I don't really plan on using AI [ at least as we speak]. AI is basically a commission right? You give demands and someone [something] produces it for you (I'm assuming for free) Do you get to choose the exact final position or is it randomized. I'm assuming the AI doesn't make a file that breaks down the piece so what would happen if you loved the picture but wanted one single thing slightly altered. Is it like scanimation where you can't produce an exact result more than once.


"Love it. Absolutely flawless but can you move that one line slightly to the left pweez ^_^"
 

Rika-Chicchi

Staff member
Site Admin
May 7, 2009
47,230
10,323
1,665
#90
So, asking out of curiosity since I don't really plan on using AI [ at least as we speak]. AI is basically a commission right? You give demands and someone [something] produces it for you (I'm assuming for free) Do you get to choose the exact final position or is it randomized. I'm assuming the AI doesn't make a file that breaks down the piece so what would happen if you loved the picture but wanted one single thing slightly altered. Is it like scanimation where you can't produce an exact result more than once.


"Love it. Absolutely flawless but can you move that one line slightly to the left pweez ^_^"
I think someday one will be able to do that. As of now the result all depends on what you "feed" the AI software & how you train it - maybe someone fluent w/ the technicalities can give you a more specific, operational answer. :)
 

Maraviollantes

Sailor Moon fan #1
Staff member
Site Admin
Jan 3, 2006
10,419
1,879
1,665
Akihabara
shoujocity.com
#91
So, asking out of curiosity since I don't really plan on using AI [ at least as we speak]. AI is basically a commission right? You give demands and someone [something] produces it for you (I'm assuming for free) Do you get to choose the exact final position or is it randomized.
It's a program you install on your PC. Stable Diffusion is by far the most popular, and free: GitHub - AUTOMATIC1111/stable-diffusion-webui: Stable Diffusion web UI
Body poses, clothes, background, and everything else can be specified in prompts.

...what would happen if you loved the picture but wanted one single thing slightly altered.
The feature called "inpainting" is used exactly for that. You paint a black mask over area you want to be altered (e.g., hand), and program will redraw it. Another approach is using "image to image". You upload your generated image, and then add in prompt details you want changed (e.g., "holding book", "short hair", "angry", etc.). Program will generate a new image closely resembling original one, but with those additional details added or changed. There are other, more advanced techniques for fine-tuning generated images.
 
Jul 5, 2009
12,883
5,535
1,665
Philippines
#92
It's a program you install on your PC. Stable Diffusion is by far the most popular, and free: GitHub - AUTOMATIC1111/stable-diffusion-webui: Stable Diffusion web UI
Body poses, clothes, background, and everything else can be specified in prompts.
By chance do you happen to know what app/software they use for scams like that Youtube commercial of Brad Garlinghouse telling you to double your XRP if you send him your XRP. It's more for making GIFs and memes. ^_^'

They are just so freakin' realistic. Just imagine the memes and videos I can make with it. I'd love to make more Joan Crawford, Dr. Frank N Furter and Baldur's Gate memes.


Admittedly this would have fooled be back in 2021 when I was so new to crypto. ^_^'
 
Last edited:
Likes: Starlight

Rika-Chicchi

Staff member
Site Admin
May 7, 2009
47,230
10,323
1,665
#93
BTW, a member of the idol group GNZ48 has just posted some private pics of herself & a groupmate w/ AI alterations in various styles on her blog: lol

微博

[EDIT]
^ So Yoshi may also use AI to do his pixel arts & conversions. lol
 
Jul 5, 2009
12,883
5,535
1,665
Philippines
#94
^ So Yoshi may also use AI to do his pixel arts & conversions. lol
From what I'm told by my animator friend pixel art AI's results lacks details. Kinda like the problem with human digits with these popular AI art apps.

Would be nice to have one perfected for pixel art. It's mentally, emotionally, neurologically and physically exhaustive. I have to both have a lot of short breaks and a long break after 2 or 3 months working on it.
 
Likes: Rika-Chicchi
Sep 19, 2022
3,733
7,271
1,665
20
#98
The feature called "inpainting" is used exactly for that. You paint a black mask over area you want to be altered (e.g., hand), and program will redraw it. Another approach is using "image to image". You upload your generated image, and then add in prompt details you want changed (e.g., "holding book", "short hair", "angry", etc.). Program will generate a new image closely resembling original one, but with those additional details added or changed. There are other, more advanced techniques for fine-tuning generated images.
Body poses, clothes, background, and everything else can be specified in prompts.
In that case why is anatomy such a standout problem people bring up with AI. You can't specify you want 5 fingers on each hand? Does nobody take the time to try and fix the hands? Since you're not doing it manually wouldn't it be easier to fix this stuff at a fraction of the speed?
 

Maraviollantes

Sailor Moon fan #1
Staff member
Site Admin
Jan 3, 2006
10,419
1,879
1,665
Akihabara
shoujocity.com
#99
You can't specify you want 5 fingers on each hand? Does nobody take the time to try and fix the hands?
Both. Algorithm is not precise, so "five fingers" will be always interpreted as "something similar to human fingers with amount close to number 5" (which still can be 3 or 6). If you are patient enough and keep sifting through generated samples, sooner or later you randomly will find one with decently shaped hands, and where all other details will be satisfactory. Depending on one's GPU, it can take an hour or more, and sometimes can be slower than manually fixing all problematic details in Photoshop, not faster. Many people just don't bother investing hours of their time into each image, so they share one of first outputs they generated (same as me just uploading a few pages of BSSM "manga" earlier in this thread, with clearly visible errors).
 
Likes: YuYu Yuichiro
Sep 19, 2022
3,733
7,271
1,665
20
Yeah, even the currently most advanced AI software are still relatively primitive & can still make many (embarrassing) errors, but after a certain number of generations of upgrading, I think they'll become indistinguishable from humans - perhaps only except that they're markedly smarter & better than humans in handling various tasks lol - & something qualitatively very different in nature could emerge in them, whether humans could tell that or not. And by then arguing whether they really have "souls," "consciousness," "ghosts," or w/e will become moot, since they'll be behaving exactly as if they have, whether they really have or not.
But if AI is better than humans at anything wouldn't that distinguish them immediately?

AI must just be the full sandwich! It will apparently be able to authenticate human interactions while possibly being able to answer "impossible" questions, release a deadly neurotoxin whenever it pleases
Next thing you know it will be able to reproduce on its own, photosynthesize, replicate the opinions of dead people, revive the dead rendering the last point mute, experience time travel, all sorts of illegal stuff I don't want to actually write out. And it will be able to simulate human errors making it indistinguishable somehow.

Forget being inspired by other human works! I can be inspired by AI disregarding the fact it can be better than any human hypothetically and I'll never be able to live up to something like that

Note* I'm just kidding around. I don't really know much about AI and don't currently plan on using it to produce art but will try not to have any strong opinions about it [for now at least]
 
Last edited: