🧵 Untitled Thread
Anonymous at Sun, 6 Nov 2022 19:53:03 UTC No. 722453
Draw something and i'll tell AI to redraw it
Oekaki Post (Time: 5m, Replay: View)
Anonymous at Sun, 6 Nov 2022 20:30:56 UTC No. 722465
Can you link the specific AI you're using for these? Looks goofy yet decent but there's so many AIs now that I can't keep up
Anonymous at Sun, 6 Nov 2022 23:38:49 UTC No. 722495
>>722464
Doing it
>>722465
locally running AUTOMATIC11 with NovelAI shenanigans. Go to the hentai board and search for diffusion in the search bar, lots of guides there.
Anonymous at Sun, 6 Nov 2022 23:52:44 UTC No. 722497
>>722464
This one generated lots of interesting results. Here's one of them.
By the way i'm trying to make it still it look like the drawing, so i'm not using too much denoising, or else it'll make a whole different image
Anonymous at Mon, 7 Nov 2022 07:51:47 UTC No. 722547
https://www.youtube.com/watch?v=tjS
Anonymous at Mon, 7 Nov 2022 09:37:05 UTC No. 722560
>>722500
"Determines how little respect the algo should have for image's content".
The reason why >>722498 is more realistic is because i'm using some default prompts that will help in making the image less fucked up. i.e. masterpiece, best quality. And some "negative prompts", which determine what the image should not have. "bad quality" etc.
Anonymous at Mon, 7 Nov 2022 09:53:27 UTC No. 722564
>>722500
I forgot, NovelAI is a model (?) that can be used in stable diffusion. It's pretty good for anime so ppl use it a lot.
here's a guide
https://rentry.org/nai-speedrun
here's another guide for everything automatic1111
https://rentry.org/voldy
>>722506
reminds me of spirit from dbd, cool. trying to get something very similar but here's a result so far
Anonymous at Mon, 7 Nov 2022 21:48:36 UTC No. 722656
Anonymous at Mon, 7 Nov 2022 22:19:56 UTC No. 722660
>>722659
Heyyyy, that one is very cool, but also >>722657
is.
>>722658
That's some bedtime fuel. Hardcore mode.
Anonymous at Mon, 7 Nov 2022 22:50:04 UTC No. 722671
>>722666
Thanks for doing it anyway, even if the results aren't ideal. What does the AI like to eat? Do they require a list of text prompts as well as a sketch? I'm guessing they work better with less visual information in the drawing? Booba combined with coomer tropes seems to be their forte.
I will try to assemble a drawing that combines more iconic elements next time. Should I include the list of prompt words that I was thinking of when I draw it? I'm not really interested in running these AI myself I prefer to let you curate the results. Thanks again OP!
Anonymous at Mon, 7 Nov 2022 22:57:02 UTC No. 722674
>>722635
Well, I guess those were it for the day.
>>722656
Glad you liked it. Second one was a surprise. Tho i'm only telling the AI what to do so, idk lol
>>722660
Pogg true.
>>722671
Pog. Well, most times I just kind of try to describe the picture to the AI, so it uses the picture uploaded and does its thing. Lots of experimentation after, and sometimes there's a good result. But yeah, with a lot of stuff going in the drawing makes it kind of hard for me to describe it to the AI. It can still work with Inpainting, which is pretty much doing the picture by parts, kind like >>722661 . I haven't tried Inpainting too much before so it looks kinda goofy. But yeah, glad you liked it. Don't worry too much about the difficulty of the prompts tho, experimenting is fun
Anonymous at Mon, 7 Nov 2022 22:58:45 UTC No. 722675
>>722671
If you hover over the filename you'll be able to see some of the prompts used. Some get cut by 4chan, probably because it's too long of a filename
Anonymous at Mon, 7 Nov 2022 23:00:51 UTC No. 722677
by the way, fuckers, for anyone interested,
https://rentry.org/voldy
here's the guide for everything AI picture proompting related.
Anonymous at Tue, 8 Nov 2022 00:12:44 UTC No. 722689
OP could you do : >>722688
wanna see what it churns out
Anonymous at Wed, 9 Nov 2022 14:38:02 UTC No. 722913
>>722910
>>722911
>>722912
hmmm not what i was expecting lol, silly how it interpreted the face mask as a dog head. guess i will have to figure it out on my own lol
Anonymous at Wed, 9 Nov 2022 14:50:51 UTC No. 722914
Anonymous at Wed, 9 Nov 2022 18:43:33 UTC No. 722932
>>722564
>>722677
Very informative, thanks.
3 questions, if you don't mind:
1- So WebUI is the real AI, Automatic1111/stable diffusion is a model and novel AI is a webpage default configurations and also a model, is that it?
2- Can we use many different models at the same time? For example, could we use a supposed digimon.ckpt and pokemon.ckpt to make more random monsters?
3- What is AI training? Is it making more models? Or is it really like making someones trained AI different from others AI plain installation?
This AI thing is amazing.
Oekaki Post (Time: 20m, Replay: View)
🗑️ Anonymous at Wed, 9 Nov 2022 20:38:43 UTC No. 722945
>>722932
Well, nv at the bottom of the webpage there are some links on the subject.
Anonymous at Thu, 10 Nov 2022 02:55:31 UTC No. 723004
>>722932
After some lurk i have figured out some things:
1- Stable Diffusion is the AI, webUI is the way they have to work with SD, thus its web interface;
2- Yes;
3- Someone can train his AI so he can associate new tabs to images then distributes it as hypernetwork/VAE. To make a model someone would need a beefy machine.
Is that right?
Oekaki Post (Time: 10m, Replay: View)
Anonymous at Thu, 10 Nov 2022 12:09:10 UTC No. 723039
>>722932
AFAIK:
1: WebUI stands for Web User Interface. Automatic1111 is the version for the UI. NovelAI is a very known model.
2: Not that I know. You can merge models to make a new one based on other 3, but you can keep merging and experiment. Berry Mix is a known merged model that's pretty good and used by a lot of folks
3: So, i'm not an AI expert, just a guy that decided to install the shit from Voldy. But, (afaik), AI training is making models, embeddings, hypernetworks that give differente results based on the dataset (the pictures you feed the AI). I never trained a model, because it requires a REALLY good setup. Plus it seems complicated.
IDK THOUGH
Anonymous at Thu, 10 Nov 2022 12:57:33 UTC No. 723052
>>723037
oh thanks!
thats uhh interesting...
Ciccio !8zrq655DwE at Thu, 10 Nov 2022 18:34:29 UTC No. 723083
>>723040
>yo, waddup ciccio
Yo anon friend, it's just a little bit okay but I try not to get too tired!
>excellent redraw lmao.
btw how are you buddy
Ciccio !8zrq655DwE at Thu, 10 Nov 2022 18:36:24 UTC No. 723084
>btw how are you buddy?
I hope everything is fine
Anonymous at Tue, 22 Nov 2022 04:03:38 UTC No. 724806
what is this specific version which takes shittily drawn input and polishes it to look good called?
too many different things out there am confused
Anonymous at Tue, 22 Nov 2022 18:39:50 UTC No. 724920
>>724911
Ty
Anonymous at Tue, 22 Nov 2022 23:27:30 UTC No. 724953
>>724909
LESGOOOO he looks so cute
Anonymous at Wed, 23 Nov 2022 03:35:32 UTC No. 724979
>>724904
Is that Ryan Gosling?
Anonymous at Fri, 25 Nov 2022 10:31:09 UTC No. 725312
>>724905
that's fucking hilarious
Anonymous at Mon, 19 Dec 2022 12:37:11 UTC No. 729109
You can run some of my drawings from other threads if you want. I'm too busy to draw new things for your thread at the moment. I don't mind if you use the word 'fixed'. The AI learned from much, MUCH more talented artist than me.
>>727079
>>726167
>>728720
>>725765
>>707959
>>707443
>>707288
>>707015
>>723914
>>723672
>>708825
>>701434
>>725364
>>724797
>>724527
>>724227
>>724204
Anonymous at Mon, 19 Dec 2022 20:00:52 UTC No. 729158
>>729127
Lol, knew it wouldn’t capture the 4 balls, but thanks for trying anon
Anonymous at Mon, 19 Dec 2022 20:39:23 UTC No. 729164
Anonymous at Tue, 20 Dec 2022 02:21:33 UTC No. 729209
>>729130
huh
still pretty neat regardless, danke anon
Anonymous at Wed, 28 Dec 2022 04:22:19 UTC No. 730287
>>729864
Ew twitter
Anonymous at Wed, 4 Jan 2023 05:12:39 UTC No. 731162
>>731056
sigh at my dick or the result lol
Anonymous at Wed, 4 Jan 2023 09:38:58 UTC No. 731172
>>722453
greetings, ai guy. while i thought i was satisfied, it turns out that satisfaction is on a meter that slowly goes down once it's full; eventually leaving the satisfied hungry for more. i can has more,
Oekaki Post (Time: 21m, Replay: View)
Anonymous at Thu, 19 Jan 2023 17:11:38 UTC No. 733549
>>731162
yes
Anonymous at Fri, 20 Jan 2023 00:20:01 UTC No. 733637
>>733546
damn did the ai unlearn caveman or something? can it do that? still saving it. what the hell did it turn the tusk into lol
Anonymous at Sat, 18 Feb 2023 21:54:31 UTC No. 737346
>>722662
Salvador DALL-E