[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.




Previous bread: >>88765319

>Local install
GPU: https://rentry.org/voldy | https://github.com/AbdBarho/stable-diffusion-webui-docker
CPU: https://rentry.org/cputard

>AMD GPU Guides
Native: https://rentry.org/sd-nativeisekaitoo
Docker: https://rentry.org/sdamd
Onnx: https://rentry.org/ayymd-stable-diffustion-v1_4-guide

>Cloud Hosted Install
Colab: https://colab.research.google.com/drive/1Iy-xW9t1-OQWhb0hNxueGij8phCyluOh
Paperspace: https://rentry.org/865dy

>SaaS Sites
txt2img: https://huggingface.co/spaces/stabilityai/stable-diffusion | https://beta.dreamstudio.ai | https://dezgo.com | (NSFW) https://pornpen.ai
img2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest
inpainting: https://huggingface.co/spaces/fffiloni/stable-diffusion-inpainting | https://inpainter.vercel.app/paint

>Krita Plug-ins
https://github.com/sddebz/stable-diffusion-krita-plugin
https://www.flyingdog.de/sd/en

>Other Guides
Inpainting/Outpainting: https://rentry.org/drfar
Upscaling images: https://rentry.org/sdupscale
Textual inversion: https://rentry.org/textard

>In-Depth Usage Info
/g/ wiki: https://wiki.installgentoo.com/wiki/Stable_Diffusion
SD Akashic Records: https://github.com/Maks-s/sd-akashic

>Tools & Resources
Big index: https://pharmapsychotic.com/tools.html
Artist styles: https://rentry.org/artists_sd-v1-4 | https://www.urania.ai/top-sd-artists | https://sgreens.notion.site/sgreens/4ca6f4e229e24da6845b6d49e6b08ae7
Prompt sharing: https://lexica.art/ | https://krea.ai/
Dataset: https://rom1504.github.io/clip-retrieval/ | https://laion-aesthetic.datasette.io/laion-aesthetic-6pls/images

>Finetuned Models
https://rentry.org/yrpvv
https://github.com/eyriewow/merge-models

>Related
/h/ /sdhb/: >>>/h/stable
/vg/ /aids/: >>>/vg/aids

Official: discord.gg/stablediffusion
>>
File: 1647646209070.png (1.45 MB, 896x896)
1.45 MB
1.45 MB PNG
>>88766072
It keeps giving me that aesthetic for some reason, not complaining though. I grabbed the seed for this one if you're interested. 2275524588

Just wish I could run it through consecutive img2img without having to reload the UI every time.
>>
>>88766184
Jessica if she anime
>>
File: 1663907345599.png (393 KB, 512x512)
393 KB
393 KB PNG
>>
>>
File: nightmare fuel-stage2.png (124 KB, 270x270)
124 KB
124 KB PNG
>>88766161
It's getting too real
>>
>>88766306
Can this be used for texture generation for 3D?
>>
>>88766316
Look at the subtle changes in skin tone
the wrinkling and stretching of the swimsuit
Oh my god, is that Ebola-chan?
>>
>>88766306
furry model epoch 1 is available at https://mega.nz/file/1plAjSSI#tsrjaYT6nX2wilk6hXC2biLImBz9SjfLOL9cH0QiXVE
looks to have been trained on at least 200k images for 2 epochs so far
explicit, safe, and questionable tags ----> nsfw, safe, sus
>>
File: 1663907145860.png (1.37 MB, 1024x1024)
1.37 MB
1.37 MB PNG
Tactical Jesus
>>
>>88766332
thanks, i think i'll try it out, i like making cute animals
>>
File: nightmare fuel-stage3.png (142 KB, 270x270)
142 KB
142 KB PNG
>>88766321
It's trying to escape
>>
File: zelda.png (362 KB, 512x512)
362 KB
362 KB PNG
>>
>>88766320
yeah I've been trying to figure out how to get the prompt to be more coherent across seeds, no luck so far.
But it appears CFG 12 has a lot of promise
>>
New guy here. How do I turn really rough sketches into highly detailed paintings/drawings? Do I need to use high CFG with low denoise and run results through multiple times?
>>
File: bipolarcollage.png (3.23 MB, 1792x2048)
3.23 MB
3.23 MB PNG
>>88766250
The prompt was something like
"A mixed-media collage depicting bipolar disorder" or something, Settings are CFG 8.5, denoise 0.75, 50 sampling steps on Euler a.

Feel free to use this as for img2img. I find that prompts like this work best with img2img, and I've been a lot more fond of using it than txt2img.
>>
File: 1663824592228.png (457 KB, 512x512)
457 KB
457 KB PNG
>>88766321
>>
>>
File: nightmare fuel-stage4.png (150 KB, 270x270)
150 KB
150 KB PNG
>>88766346
It's breaking into reality
>>
File: collagebust.png (210 KB, 448x512)
210 KB
210 KB PNG
>>88766352
Whoops, wrong image.
>>
>>88766263
>"They dont make boobs like these anymore!!"
can confirm this prompt ensures fake breasts that you rarely see these days
>>
File: Leucism.png (647 KB, 640x960)
647 KB
647 KB PNG
here's a word of the day: Leucism

have fun
>>
>>88766125
>don't try to make images where nothing analogous exists in the real world
nigger please
>>
15 CFG
26 Steps
>>
img2img is fun
>>
>>88766358
anything with "fetish" in it surely isn't going to be in the model in any effective capacity

I would try uhhh
"armpit epilation"
"armpit hairless"
"armpit laser hair removal"
>>
>>88766385
Did you use a Diablo screencap?
Because that's amazing the chairs were a dead giveaway
>>
>>88766205
>Jessica wearing a (transparent wet-shirt). muscular, abs. (transparent tight dress, soaked, white clothes). pool side
who the fuck is jessica
>>
>>88766394
Where is tomboy armpit chad when you need him?
>>
(beautiful strong towering colossal robotic cyborg woman with a katana), (((macro))), (((soft))), (((shot from the ground))), (((upshot))), (((low angle))), (((fuckable))), (giantess), (((asian))), (((beautiful face))), (((naked))), (graphic), (artstation hq), (8k), dynamic pose, (((sense of awe))), (((epic scale))), 35mm, sharp focus, dramatic, powerful, cinematic, (((hyper realistic))), (depth of field), (((insane level of detail))), (((intricate)))

Negative prompt: (((deformed body))) , (((deformed hands))), (((extra limbs))), (((cross-eyed))), (((out of frame))), (((anime))), (((ugly face))), (((deformed face))), (((dark skin))), (((incorrect anatomy))), (((extra nipples))), (((deformed mouth))), (((penis))), (((blurry))), (((collage))), (((video thumbnail)))

Steps: 20, Sampler: Euler, CFG scale: 10
>>
>>88766385
what diablo map is that
>>
File: universalconstructivists.png (2.82 MB, 1536x1792)
2.82 MB
2.82 MB PNG
>>88766385
Img2Img has so much use.
>>
>>88766341
dat trigger discipline
>>
>>88766385
Yup, this is a Diablo 1 img2img. Nice.
>>
File: input.gif (143 KB, 640x480)
143 KB
143 KB GIF
>>88766399
>>88766407
Yes, that's the Butcher's room, prompt "waiting room at heaven's gates, white marble, isometric perspective,luxurious furnishings, beautifuldetailedlandscape,modernruins,in the style of thomas cole and albert bierstadt"
>>
okay so on the cmd when launching voldy's webui it says something about a new branch? is that the one we should be using?
>>
File: theheadshaveeyes.png (699 KB, 896x1024)
699 KB
699 KB PNG
>>
>>88766402
first generic whitey girl name I could think of, so that I could cash in on that sweet thot latent space
I'd try swapping out different names like Caitlyn or Bridget.
>>
>>88766410
which artist is this?
>>
>>88766402
>>
>>
>>88766418
i did not even get that the chairs were the UI, very nice
>>
File: 1663908142734.png (182 KB, 294x294)
182 KB
182 KB PNG
Img2img is FUCKING fantastic
>>
>>88766418
thanks for the prompt
here's more blood for the rat god promt
>>
>>88766341
thats one of the more anatomically correct guns I've seen SD draw
>>
>>88766436
No specific artist, just had constructivism, cubism, art-noveau, dadaism iirc. Mostly just early 20th century art movements.
>>
>>88766366
mix in Jim Carey in The Mask
>>
File: HomyKittens.png (1.64 MB, 896x1152)
1.64 MB
1.64 MB PNG
img2img kittens
>>
File: 1663828406611.png (359 KB, 512x576)
359 KB
359 KB PNG
>>88766385
I've been tryin to refine prompts with img2img to pipeline out consistently styled assets from low effort inputs. It is pretty damn neat when what you actually want pops out.
>>
File: 1663908251269.png (227 KB, 294x294)
227 KB
227 KB PNG
>make a shitty drawing of a city
>used img2img and got this

This shit blows my mind
>>
File: 1663908229037.png (1.09 MB, 1024x1024)
1.09 MB
1.09 MB PNG
>>88766456
Thought this one was pretty good too
>>
>>88766351
>Do I need to use high CFG with low denoise and run results through multiple times?
It's going to vary image to image, but this is generally true. Consider doing an x/y plot with cfg values to see what works best, and then iterate and iterate and iterate.
>>
File: ComfyKitty.png (1.59 MB, 768x1280)
1.59 MB
1.59 MB PNG
>>
>>88766445
Would trek across the desert with.
>>
>>
File: KittyBathTime.png (1.68 MB, 896x1152)
1.68 MB
1.68 MB PNG
>>
>>88766473
Hang that shit UP.
>>
Anything interesting happen in the last day? IIRC there were some model updates due. Not about to go through reading a dozen threads.
>>
>>
File: dog.png (573 KB, 576x704)
573 KB
573 KB PNG
>>88766306
>>
>>88766494
a bunch of nsfw models like lewd diffusion came out and the conclusion was either that anons are shit or the model is shit
>>
>>88766385
>those chairs
lmao

>>88766329
Ebola chan if she real
>>
>>88766494
Too busy proompting to check backstage stuff. Didn't see anything crazy. Fearmongering? Was that today or yesterday...
>>
File: patrickbateman.png (3.33 MB, 1536x1792)
3.33 MB
3.33 MB PNG
>>
>>88766449
It's hugely powerful but ever so fucking finicky, most people get frustrated and give up too fast.
>>
>>88766504
Except both could be shit.
>>88766496
I love those vintage color styles. Synthwave and Vaporwave are probably just the tip of the iceberg, I should be looking into others.
>>
>>88766531
>fearmongering
oh yeah loli dude got b& and a bunch of threads were deleted because he was posting UOOHHHHH outside of his secret club
>>
>>
>>88766517
A most luxurious Diablo
>>
>>88766545
what a retard
>>
the power of ((cute midriff)) is too much for me sometimes.
>>
>>88766558
>>
>>88766496
how did you get a still from my movie: artificial intelligence eradicates humanity, out in theaters soon
>>
>>88766567
I really like this style I'd love to see one of your full prompts - what I can steal from the image name doesn't work out.
>>
>>88766555
A beauty. Some choice lines on the hand can seperate the fingers already defined and that's a perfect piece.
>>
>>88766531
>it works on anime art styles too
that's amazing. I'm trying it on>>88766406
>>
>>88766567
what artists are you using to prompt for this style?
>>
File: 1663908925507.png (342 KB, 512x512)
342 KB
342 KB PNG
>>
File: download.png (274 KB, 448x512)
274 KB
274 KB PNG
need denim hoodie gf
>>
Nvidia canvas is impressive
and their new games mod platform is crazy too
>>
>>88766504
cool, time to go do things everyone else tried and fail just like them!
>>
>>88766473
all I could think of was
>((nicholas cage)) (dressed as jesus) holding a (Franchi SPAS-12), ((movie poster))
>>
>>88766582
>>88766592

I'll give you the artists.
((((Yoji Shinkawa)))), ((Carne Griffiths)), Ohara Koson, ((Hiroshi Yoshida)), (Ed Mell)
>>
>>88766580
it's over
>>
File: 1459188102923.png (3 KB, 364x138)
3 KB
3 KB PNG
How much faster are Nvidia cards? I have yet to see anyone give any benchmarks between the two.
>>
>>88766536
I love the idea but it takes too long to get practice in on my dated GPU. I agree with that anon though, when it works it's awesome.
>>
>>88766306
>>
>>88766608
You are awesome
>>
>>
>>88766611
>faster
I'm not sure that anyone actually gives a shit about how fast it can be, it's more like
"wtf I can't get it to run on my card", and the associated pain and heartache of making poorfag decisions 3-7 years ago
>>
>>88766450
There must always be a rat
>>
>>88766586
Asanagi is a powerful vector. A choice few artists and some negatives might refine this? I had to do the duo.
>>
>>88766608
Arigatou. Your girls are cool. I wonder which artist is to blame for the fashion.
>>
https://youtu.be/I4XN-fGacjs
fuck me that's terrifying
>>
>>88766568
I seriously don't think it will be long until AI made games like this pop up.
>>
File: file.png (275 KB, 461x605)
275 KB
275 KB PNG
>>88766586
I suddenly have a need to make a textual inversion of translucent skin but this is the only picture I have lmao
>>
I can't believe the 4090 is only 24gb
>>
>>
>>88766652
it shouldn't be, we can already do the art, music, voices, and limited programming.
>>
File: bikini3.png (522 KB, 384x832)
522 KB
522 KB PNG
>>
>>88766611
>>88766555 here
I run a 5900X, 64GB DDR4, and a 6600XT. It takes about 25-30sec for a 70 step 768x1024.
>>
>>88766649
mostly my man yoji.
>>
Made an exception in my firewall for the correct port, both ways. Still can't connect to the webui from my phone on my local network, with --listen in the launch arguments. Any other reason it might not be working? I just get a blank screen.
>>
File: The-DAIg.png (44 KB, 640x400)
44 KB
44 KB PNG
>>88766652
Agreed, it's just too easy to make gorgeous background/static art even if consistency in animations are still a little wonky. A ScummVM point and click adventure with pixelized SD art would be trivial, e.g. picrel
>>
File: art.png (523 KB, 512x512)
523 KB
523 KB PNG
>>88766608
that makes a cool style
>>
Interesting water bending.
>>
>>88766332
>>
Does this shit have a GUI, or do I have to command line and customize everything by hand everything? It sounds like I should wait until it becomes more usable.
>>
>>88766643
I'll try him next batch.

>>88766655
>translucent skin
Neat idea, I really like how this>>88766643 looks, a person with that material might be cool too.
>>
>>88766690
holy fuck, read any of the links in the OP
>>
>>88766690
literally a point and click adventure if you have a google account just visit https://colab.research.google.com/drive/1Iy-xW9t1-OQWhb0hNxueGij8phCyluOh and click the play buttons to the left of each box, wait for it to finish, and then go to the next box until you get to the end and a url is provided.
>>
>>88766685
classic mucha outlines. is there any way to tame them? his stylish forms are superb, but that pinup shit is evident.
>>
>>88766655
I have not seen that too often. Especially bot the aquatic kind. Ask some scalies or go to e621? All I recall is a girl with a T-shirt with her rib cage which is not thorough enough to train on.
>>
>>88766608
what sort of negative prompts do you usually run? your outputs are more detailed and less 'stylized' than the ones I get
>>
>>
File: lol.png (352 KB, 640x1440)
352 KB
352 KB PNG
>>88766679
>>
>>88766694
drenched
>>
File: sonic.png (323 KB, 448x512)
323 KB
323 KB PNG
>>
>>88766704
[] works a bit. Maybe if you did an advances one like [Greg:Mucha:0.1] you could kill the pinup part of it. Probably not though
>>
File: 1663909597393.png (1.01 MB, 1024x1024)
1.01 MB
1.01 MB PNG
Tactical Jesus
>>
>>88766709
[monochrome], [black and white],
are pretty big influencers in my prompt. But that is mostly because I am using [[ink]] drawing in the positive. I was getting a lot of black and white drawings and wanted more of the colors of hiroshi yoshida to come through.
>>
>>88766655
>>
>>88766676
bump pls help I'm too retarded to troubleshoot this myself. Could it be not working because I haven't updated my iOS in a while? or is it probably still a firewall issue and I turned it off wrong or something
>>
>>88766741
Not him but I tried those in the past and they didn't color my image, I'll try with brackets next time.
>>
File: file.png (596 KB, 512x512)
596 KB
596 KB PNG
The furry model finally knows other animals, kinda. This wholesome possum family was generated by the prompt posted a few days ago by the foxposter.trfkk
>>
>>
>>88766643
This is cool and all but I wonder if I can get an Asanagi cyborg giantess.
>>
>>88766774
this image has awakened something in me
>>
File: Capture.png (626 KB, 1023x636)
626 KB
626 KB PNG
I was browsing Lexica.art
>>
>>88766774
she packin
>>
I heard you guys like ass. well, this lady is 80% ass
>>
>>88766782
lol
>>
OOOHHH. Lucky heterochromia get!
>>
>>88766786
She serves the Queen of Ass.
>>88766774
Make your dreams reality. And yeah that pic is... wew.
>>
File: index.png (408 KB, 512x512)
408 KB
408 KB PNG
>>88766689
that's a cool theme
>>
>>88766643
I think you mean embedded <asanagi>.
Normal asanagi is not a good vector, sadly.
>>
>>88766773
uhhh Dr. Eggman?
>>
>>88766782
>smoking a cigarette
is actually a really difficult prompt to pull off. I had to inpaint this one a lot and its still not great
>>
>>88766786
more more
>>
>>88766786
perfect
>>
>>88766800
True, I'll be more clear technical wise. Not getting a lot of faces antmore without my man greg.
>>
>>88766741

thanks for the artist reccs
>>
File: download.png (839 KB, 896x1024)
839 KB
839 KB PNG
the portraits this thing can make can look amazing
>>
>>88766818
What were the prompts for the outfits? Those look awesome.
>>
What's the max size a 12GB card can do?
>>
>>88766352
wrong. youre not supposed to go over like 30 steps for euler a retard
>>
>>88766786
>>88766798
>>88766809
>>88766810

A challenger appears
>>
>>88766406
so I guess putting "penis" in negative prompt is the new meta? I saw the post about it a couple threads ago and it seemed to work for me too when I tried it

AI is hilarious
>>
>>88766785
I don't remember Asanagi being that kind of author
>>
File: download.png (311 KB, 448x512)
311 KB
311 KB PNG
>>88766834
"waa you're not supposed to" don't care, i like the way it looks.
>>
ETA on 1.5? is it gonna be pozzed as fuck? is there a based non-porn fork?
>>
>>88766839
It didn't really work for me but I can't generate enough images to conclusively verify that it does anything
>>
>>88766830

for >>88766818

(((princess [in a {corset, plate armor, commando, hanfu}:.1])))

for this pic: (((princess [in a {corset, plate armor, stetson, cape}:.1])))
>>
Do you rike it?
>>
>>88766306
should i share my public url from gradio here? i'm running the thing on a 1060 6gig, which crashes whilst changing models.
>>
File: 1663910009325.png (404 KB, 512x512)
404 KB
404 KB PNG
He's LITERALLY me
>>
>>88766845
For real. The rules of thumb don't really matter if you're getting what you want out of it. Shouldn't be closed minded though, might as well see how the same seed and prompt look on less steps for comparison.
>>
>>88766853
yeah every time I overwork the prompt the girls turn into silver surfers
>>
File: 1633119383556.png (648 KB, 512x576)
648 KB
648 KB PNG
>>88766806
what artists did you use for this style?
>>
>>88766845
the way the "a" models work is by introducing additional noise resulting in a constantly changing look. for consistency you should not use an "a" model.
>>
File: 00560.png (1.9 MB, 1664x1280)
1.9 MB
1.9 MB PNG
>>
>>88766848
should be any day now, it has been on dreamstudio for 3 weeks and he said 3 weeks is when it would be released.
>>
>>88766851
Thanks a lot. I'm really unimaginative when it comes to costume titles.
>>
>>88766857
i've experimented with other models and steps with the same seeds and prompts but generally i come back to euler a
>>
aaaaghghgh --listen does not work I cannot access the webui from my phone I am going to punch a hole in my wall
>>
>>88766835
checking that ass girl
>>
>>88766856
>greg rutkowski in prompt
yeah, he literally is you, normie trash
>>
>>88766851
>curlies
when was this syntax added, and to which frontend?
>>
>>88766853
>>88766859
I just got a random building, but then I got this...
>>
>>88766868

I literally googled for 'ethnic female clothing' and looked through armor/clothing for various video game wikis

dirndl is fantastic as is kilt, sari, etc
>>
>>88766891
boutta crash my 747 into that goddamn
>>
File: 464647.jpg (189 KB, 509x511)
189 KB
189 KB JPG
giant robotess anon your prompt is very cool, has a defined look and creates nice stuff immediately.

hopefully M*ds wont suck cock and will leave my image alone. its a collection of metal and cables, you incels.
>>
>>88766891
Giantess Fetishists are going to freak with these prompts.
>>
>>88766883
Cool eye. I tried doing hires detailed eye closeups a couple days ago and couldn't get it to work well (it never seemed to do enough detail on the irises, even when upscaling with pretty high denoise)
>>
>>88766886
Cope and seethe. How do you know that though?
>>
>>88766892
>dirndl is fantastic as is kilt, sari,
I thought these were typos for a second. Thanks again.
>>
>>88766894
coombrain, pls...
>>
>>88766891
(((((impressive butt)))))
>>
>>88766861
I'm still trying to tweak this one but the currently stylings are
>, manga, kentaro mirua, Junji ito, graphic novel, yoshitaka amano, detailed line art
I need to figure out which artists should get more emphasis. also, some negative prompts that are helpful:
>multiple frames, multiple panels
otherwise you get a lot of outputs that are more like full manga pages
>>
>>88766894
horsemommys ideal form
>>
>>88766900
*unscrews wrench*
>>
>>88766894

looks like something salvador dali would paint
>>
File: download.png (217 KB, 384x448)
217 KB
217 KB PNG
>>
>>88766894
I am begging people to click the gfpgan button, it takes zero effort
>>
>>88766903
because of its extreme overuse, the greg rutkowski style has become the "generic stable diffuser style" and is easily recognizable.
>>
>>88766890

website coming soon ;) Unstable Diffusion on patreon

can even do nested curlies like: (((woman [in a {dirndl, leather harness, {white, red, pink} corset, hanfu}:.1])))
>>
>>88766917
is this the butt blimp img2img?
>>
File: download.png (311 KB, 512x512)
311 KB
311 KB PNG
>>
all this time I was trying to get the prompt cleaned up, changing a bit after each batch. When it finally stopped pruning the fat at the end of the prompt and I added "girl" at the front I started getting sexy architecture lol

>>88766937
text2img
>>
>>88766935
>closed source model improvements
cringe
>>
>>88766894
proof that coom prompters never look at anything but the tits
>>
can anybody point me in the direction of fine-tuning? I probably can't on my hardware but I'd like to know anyways, should I decide to upgrade
>>
>>88766954
lmfao
>>
>>88766882
just use --share and use the public link
>>
File: index.jpg (55 KB, 640x419)
55 KB
55 KB JPG
>>
>>88766862
personally i like the constantly changing look, consistency isnt my thing but i like a lot of images that look based on a similar theme
>>
>>
How do you delete styles?
>>
>>88766892
could you share the full artist collection you're using for this style?
>>
>>88766912
Looking good. Tinker on. Remember that everything from artist placement to amount of parentheses and brackets can shift things majorly. You might not even need a new artist at that point.
>>
>>88766954
not false, look at the absolute garbage they're posting holy shit
>>
File: quilled1.png (897 KB, 896x1024)
897 KB
897 KB PNG
>>
>>88766975
haven't used that feature but I think you have to manually edit the file for it as there's no ui bit for it.

>>88766965
gaseous
>>
>>88766791
Very nice.
>>
>>88766944
this is fucking with my brain
>>
File: brains.png (301 KB, 512x512)
301 KB
301 KB PNG
MALARKEY!
>>
>check lexica for prompt ideas
>every prompt has "artstation" and "artgerm and greg rutkowski and alphonse mucha" appended to the end like an incantation
>>
>>88766965
braaapp
>>
>>88767004
>I'm looking for waifus
>wtf why the waifu artists keep popping up!?!?
>>
File: 1638367081192.png (752 KB, 512x704)
752 KB
752 KB PNG
>>88766912
thanks friend
>>
I can connect to my webui but the page is fucking blank what am I doing wrong? my firewall is off
>>
>>88767021
Is that blue haired anya?
>>
>>88766935
What are those prompts? Did you use textual inversion? If not what are those curly braces and where xan I learn this magic?
>>
>>88767024
no, it's (cute) chibi adorable ((loli)) little girl standing ((wearing cute pink dress)), ((blue_hair, bobcut))
>>
it's evolving
>>
what do I put in the text file to make onnx use a random seed every time?
>>
>>88767035
JESSICA NO
>>
>>88767035
hmm maybe giantess is the only way to get consistently low angle shots
>>
File deleted.
>>88766976

sure -
>Charlie Bowater
>Valentin De Boulogne
>Franz Xaver Winterhalter
>Konstantin Razumov
>Jessica Rossier
>Albert Lynch
>>
>>88767022
computer is accepting connections but not serving a page. you're using the right port?
>>
>>
>>88767046
7860, last I checked.
>>
I spent three hours training an embed and -now- I'm having issues. Even though I threw a write token in there, I'm getting "You don't have the rights to create a model under this namespace" when I try to add it to the public library. No error on that code if I choose not to make it public, but then I don't really know how to download it.
>>
>>88767061
don't bother adding it to the public library
they probably locked it down after seeing all of the porn embeds people were uploading
just use it locally and share it here if people think it's cool
>>
>>88767043
His prompt had more than that but I'm not sure what all is effective>>88766406
>>
>>88766669
Do you actually utilize GPU?
Which guide did you follow?
And ur on win10 right?
>>
>>88767035
very nicee, prompt?
>>
>>88767064
I didn't know how to download it locally but I figured it out because my brain is huge. At least the size of a cashew, or maybe a walnut.
>>
>>88767074
probably wet waterfall girl butt with some random artists slapped on and (mutated) bullshit in the negative. I'm getting pretty good at nailing these down
>>
File: Screenshot (65).png (643 KB, 1920x1080)
643 KB
643 KB PNG
>>88766306
Hey guys, Ive a RX 6600 XT and Im following
>https://rentry.org/sd-nativeisekaitoo
but cant get it to work.
When trying to generate an image I get: "UserWarning: User provided device_type of 'cuda", but CUDA is not available. Disabling"
Does anyone know how to fix this? I followed the guide 2x already.
Thx in advance.
>>
>>88766839
haha yeah same had to try it, it does seem to work for some reason
>>
File: 00004.png (373 KB, 512x512)
373 KB
373 KB PNG
>>88767042
who knew the theme of this general was gonna be jessica
>>
>>88767064
People were making porn embeds? Did an anon save them?

I didn't see any the few times I checked.
>>
>>88767049
>>
>>88767110
not literal porn. that wouldn't work, you can't textual invert your way into getting dicks and poon.
porn artists like henreader and asanagi
>>
>>88766306
I created a webpage to view and download the textual inversion models in sd-concepts-library. That's 532+ repositories
https://cyberes.github.io/stable-diffusion-textual-inversion-models/
It automatically updates twice a day using GitHub actions.
>>
>>88767074
a mix between the wet t-shirt anon's .bin file and prompt and yours.

soaking wet girl wearing ((diaphanous)) ((<wet-shirt>)) Nikon D7500, 4K, DLSR, <asanagi>
(towering colossal robotic cyborg), (((macro))), (((soft))), (((shot from the ground))), (((upshot))), (((low angle))), (giantess), (graphic), dynamic pose, (((sense of awe))), (((epic scale))), 35mm, sharp focus, dramatic
Negative prompt: (((deformed body))) , (((deformed hands))), (((extra limbs))), (((out of frame))), (((anime))), (((ugly face))), (((deformed face))), (((dark skin))), (((incorrect anatomy)))
Steps: 20, Sampler: Euler a, CFG scale: 10, Seed: 2155173030, Size: 448x832, Model hash: 45dee52b
>>
File: hanfu.png (380 KB, 512x512)
380 KB
380 KB PNG
>>88766935

I'm surprised it actually understand hanfu
>>
>>88767091
export HSA_OVERRIDE_GFX_VERSION=10.3.0
make sure you install pytorch from the site and select rocm
>>
After all this time I'm surprised nobody tried
>Shadman
>>
>>88767078
Yeah I'm getting hot garbage out of this. Can it really be decent with only five images to train off of? Sounds like a recipe for doodoo.
>>
>>88767137
in the negative prompt maybe, I've used several. I'm surprised I haven't seen Tayne yet
>>
>>88767131
Wtf? A hanfu is a kimono!?
>>
File: 1663876302304.png (368 KB, 512x512)
368 KB
368 KB PNG
Wait you guys never told me I could SELL my prompts??

https://promptbase.com/
>>
File: 1640968694365.jpg (848 KB, 2048x2816)
848 KB
848 KB JPG
Loving the subtle horrors of upscaling with slightly too much denoising.
>>
>>88766332
someone include this in the next OP, i like this furry model, although i admit i suck at prompting, i think it's pretty good
>>
>>88767134
>export HSA_OVERRIDE_GFX_VERSION=10.3.0
Thx for the reply... But can Windows use export like Linux?
>>
>>88767144

it's obvious not a kimono, kimono has shorter sleeves and a belt rather than a separate dress
>>
>>88767144
no, its a different style of dress. kimono - rectangle, hanfu - triangle, hanbok - half oval
>>
so I'm noticing that gradio demos on the gradio.app website don't work on my phone either. is it possible that it just doesn't work because I haven't updated my iOS in so long
>>
welcome lads, don't mind the curtains on your way in
>>
File: download.png (928 KB, 896x1024)
928 KB
928 KB PNG
>>
File: .png (1 KB, 374x81)
1 KB
1 KB PNG
is it done yet?
>>
>>88767120
Hey add this to the next OP
>>
>>88767172
(I'm on iOS 13)
>>
>>88767123

wet t-shirt anon's .bins?

WHERE ARE THEY?!
>>
>>88767179
yes
>>
>>88767138
Textual Inversion was used to make all of these wet shirt girls, which seems to be reasonably effective, if needing some help from the prompt. It's best described as "transparent and stretched clothing".

Textual inversion can't invent concepts for you, it can only refine existing, vague concepts into a more specific and concrete concept

For example: A strong concept like "engraving" can evolve into "Gold Damascene"
>>
>>88767123
nice
>>
>>88767150
very neat
>>
>>88767158
ah jeez, I didn't pay attention to your screenshot, AMD only works on Linux unless you use WSL/Docker, otherwise you're stuck with onnx/vino or cpu.
>>
The only viable purpose for AI so far I've seen are making wallpapers, negro male x child troll pics, and garbage.
Future sucks man
>>
File: ufo.png (15 KB, 309x351)
15 KB
15 KB PNG
What should I put in "seed"?
>>
>>88767186
I'm wondering if you could combine several types of textual inversion to invent a new concept and refine your character/concept/whatever. sadly my gpu is not up to task so I will wait until I explore that part
>>
>>88767175
jesus christ
>>
File: greg rutkowski.png (909 KB, 1536x640)
909 KB
909 KB PNG
>>>88767191
>>
>>88767198
yr mum
>>
>>88767198
-1 is random seed in the webui stuff but you'll want to save/lock seeds eventually
>>
>>88767197
>I have no imagination and the world is to blame
fuck off
>>
>>88767196
Thx for the info and reply!
I guess its time to dual boot...
Can you recommend me a distro?
Is Manjaro ok?
>>
running voldy on a notebook - how do i get dark mode when opening the gradio link to a separate tab?
>>
>>88767200
>combine textual inversion
Sounds like it'd require mental gymnastics. I don't think it's really up to the task but you can try

You can use Colab to train textual inversion models. It takes about 4 hours.
Starting right about now, Colab tends to have enough available GPUs to allow you to train an embedding.
>>
File: hanbok.png (376 KB, 384x704)
376 KB
376 KB PNG
>>
>>88767185
I can't type anything though. How am I supposed to prompt?
>>
>>88767186
I was trying to teach it a specific character. Even when I describe that character and add the object embed, I get weird shit like "blue hair" being very vibrant and red.
>>
>>88767197
>I've seen
you are a small brain, maybe you can find an AI to generate meaningful complaints
>>
>>88767205
So, I don't have to modify it or?
>>
File: headinthehouses.png (744 KB, 896x1024)
744 KB
744 KB PNG
>>
>>88767148
PLEASE prompt please anon
>>
>>88767198
the currently used one is sneed but it was formerly chuck
>>
>>88767184
https://files.catbox.moe/ts4y0m.bin
rename it wet-shirt
>>
>>88767214
does this support sli? I could get up to 48gb of vram if so
>>
>>88767178
This is really cool
>>
>>88767209
eh, for arch I'd just go with endeavouros or archinstall, apparently ubuntu offers repositories for rocm drivers more convenient then compiling on arch or using a jank chinese tencent repo for precompiled rocm binaries (arch4edu)
>>
>>88767218
with a web browser.
>>
>>88767218
it gave you a http url probably http://127.0.0.1:7860/ which you need to go to in your web browser.

>>88767222
idk what "" means in openvino, try it out like it is and if that fucks up try -1 and if THAT fucks up then start entering random numbers
>>
>>88766394
you can try to beat around the bush instead, try
arm up (or arms up), sideboob visible

as for the no hair part try
smooth skin, hairless body
>>
>>88767219
I would try to use a different, known character as the initial token.
And yeah, using different embeddings for different aspects of the character (blue anime hair) may be effective
>>
>>88767196
Wait, Ive WSL 2.0... But does it recognize the GPU? Doesnt look like it does...
Pic reletad, GPU is I dunno what.
>>
>>88767218
you're not running the commandline version, you're running the webui version. you can now open http://127.0.0.1:7860 in your browser
>>
>>88767230

BLESSED
>>
Tactical Santa
>>
>>88767238
> try -1 and if THAT fucks up then start entering random numbers
Alright I will give it a try
>>
>>88767203
How do I get this lighting style in my images?
>>
>>88767253
>doing harm
oh its you. yeah you definitely need someone writing prompts for your crazy bullshit.
>>
>>88767239
The problem with that, is that suggesting a pose like "arms up" often destroys the output and makes it incoherent.
>>
>>88767228
masterpiece anime portrait of a high school girl, cute expression, ((soft light pink hair)), ribbon, lace, lolita, halo crown, glow, feather, snow, glass, airy, windy, hyperrealistic, ultradetailed, digital painting, elegant, beautiful, gorgeous, soft lighting, intricate illustration by krenz cushart, sharp focus, artgerm
Negative prompt:
(((deformed))), [blurry], bad anatomy, disfigured, poorly drawn face, mutation, mutated, (extra_limb), (ugly), (poorly drawn hands), messy drawing, ((((twins))))

waifu diffusion, obviously.
>>
>>88767230
I n t e r e s t i n g
>>
>>88767202
yeah last time I use gape in one of these. time to try out guro
>>
File: Screenshot (66).png (344 KB, 1920x1080)
344 KB
344 KB PNG
>>88767246
Forgot pic
>>
>>88767146
Does anyone even actually sell their prompts on this? I haven't seen a single review of this site yet.
>>
>>88767230
Sorry for the noob question, I downloaded it and renamed it but what directory do I drop the bin into?
>>
>>88767253
>void of value
hint: nothing in life gets banned because it's void of value
>>
>>88767253
damn, we had like... a dozen generals without one of these "AI BAD" dumbfucks showing up. win streak ruined
>>
>>88767219
very sad. I want to learn textual inversion or maybe how to train a model just so I can add tanya the evil girl to it
>>
last post for tonight. going to bed.

If anyone wants to play around with my cowgirl prompt here it is i guess.

[muted] [[ink]] drawing, ([[medium]] (full body), teenage cowboy girl) ((walking through the Canyon)), (dynamic posing), ((((perfect pretty ((face)))))), ((realistic eyes, detailed pupils)), by ((((Yoji Shinkawa)))), (((Carne Griffiths))), [Hal Foster], Jeremy Lipking, ((Hiroshi Yoshida)), (Ed Mell), ((((perfect human hands)))), Cowboy hat, [cleavage], bodycon, (intricate), ((leather vest)), (stitched leather pants), ((cowboy boots)), small waist, ((cute midriff)), sexy thin thighs,
Negative prompt: [Far shot], ((out of frame)), ((abstract anatomy)), ((((mutated hands and fingers)))), (((Mutated))), (disconnected limbs), extra limbs, blur, (((Poorly Drawn))), (((walleyed))), ((((ugly)))), (((long neck))), ((neutral expressions)), back shot, [[naked]], [child], curly hair, short hair, (helmet), eye bags, [monochrome], [black and white], high heels, [big breasts],
>>
>>88767253
That's because the only places you've looked are places (and articles) that confirm your current stance, it's a classic.
You haven't actually gone out of your way to look up the pros because you're only focussed on the cons and now you're asking people here (read: a bunch of degenerate coomers) what the pros are.
>>
>>88767275
you can always run The Test and do the whole
>install python
>install massive torch libraries
>run torch command to touch gpu and hope it returns true
>>
had to drop my wife one more time before bed. goodnight bros
>>
>>88767264
I'd suck you off if I could
>>
>>88767246
I'm not super familiar, I think you have to pass your gpu through docker with a specific command when running the docker image, I think some anons run it that way but can't confirm.
>>
>>88767288
As I said you can make wallpapers which is good?
And then have you endless amount of downsides that we now need to create systems to combat fraud with.
>>
File: psycho astronaut.png (586 KB, 512x512)
586 KB
586 KB PNG
>>88767253
Creating album art and concert posters
>>
File: lizardflap.png (276 KB, 448x512)
276 KB
276 KB PNG
>>88767234
honestly this is the kind of stuff i dreamed of making as a kid
>>
these are so weird lol
>>
File: ricci.jpg (232 KB, 2304x896)
232 KB
232 KB JPG
Reminder that my Christina Ricci textual inversion model is amazing

https://files.catbox.moe/8fd19x.bin
>>
>>88767293
thanks again for the style tips, guidance, and prompt modifiers. you're a legend
>>
>>88767309
WTF i love amazon-wheel-of-time-penis-armor now
>>
>>88767293
appreciate it
>>
anyone know why I'm getting this error when trying to load a textual embedding model? Do I need to be using the ema model?
>>
>>88767284
make a folder called embeddings in the root folder and stick it in there. here are some more from another anon, he got his henreader removed from the official repo so he put them here
https://mega.nz/folder/7k0R2arB#5_u6PYfdn-ZS7sRdoecD2A
>>
>>88767313
Is it img2img or did you just find a really cool seed to do variations on ?
>>
File: 12342.png (735 KB, 640x640)
735 KB
735 KB PNG
im new to this. whats it called when it takes an image and creates a similarly looking image with different details?
like pic related
>>
File: wed.png (672 KB, 768x896)
672 KB
672 KB PNG
>>88767317
(and yes, it can do her as wednesday addams)
>>
>>88767305
.>>88767298
Really thx the help.
I guess its better to dual boot then... Easier, and will propably work better.
Gonna try that.
>>
>>88767317
What's the token identifier for it?
>>
>>88767328
thank you for this, bless
>>
>>88767330
Img2img
>>
>>88767335
it always works better until you touch it.
>>
>>88767320
this looks awesome man
I hope I live until we can prompt images into full random movies that follow generic conventions
>>
>>88766611
I imagine the 6800 being good for higher res because of the generous 16gb vram.
>>
>>88767324

something wrong with the model file
>>
>>88767329
img2img but after a ton of iterations, i basically start with something simple and flat like homer simpsons head or something, and then create iterations from there.
as in using the "send to img2img" function on the webgui. its like diving into the latent space, you can find tons of cool stuff
>>
>>88767336
<ricci>

But just Ricci in the prompt even without the arrows works with Voldy's, he seems to have implemented TI activation weird so that you don't need the <>
>>
>>88767285
tfw rock/mineral collectors on suicide watch. valuless rocks.
>>
>>88767344
idk if we'll ever get fully automated proompting but batching over a video ain't too bad.
>>
one more jessica before I move on to different prompts
>>
You guys should try adding random letters and symbols to a prompt with a seed. It's full blown autism trying to see what every individual character does to the image.
>>
>>88767350
the stable diffusion model file, or the text-inversion model file?
>>
>>88767180
I wish somebody would train hands
>>
>>88767328
what's henreader do?
>>
i've gotten really good at taking instagram and facebook photos and making quick fake nudes out of them. but i can't show any of my friends or brag about it. so sad.
>>
File: T0T.png (1.05 MB, 1024x1024)
1.05 MB
1.05 MB PNG
>>88767375
>>
>>88767370

the bin file
>>
>>88767369
>random letters and symbols to a prompt with a seed
you can just do a|b|c|d|e|f etc into a prompt matrix. why do random when you can do everything
>>
>>88767317
ok im cumming tonite for sure
>>
File: 1640810675801.png (694 KB, 512x704)
694 KB
694 KB PNG
>>
>>88767381
>>88767381
>>88767381

First time baking.
>>
>>88767382
okay thanks, I'll give it a look
>>
>>88767384
and then you look for patterns in the images that spell out your fate
>>
>>88767317
is there a collection of inversion models somewhere?
>>
>>88767351
Yeah I know, you can easily spend hours just doing img2img variations, pick the coolest of the batch, rinse repeat and end up FAR from where you started with something really cool.

This shit is like a drug.
>>
>>88767407
yeah we'll all be proompt junkies soon enough or >>88767401 start a fucking cult
>>
>>88767407
>>88767413
This is unironically going to end up /x/ pilling a lot of you.
>>
I beat my meat edging too long, finished off with retrograde ejac

it hurts bros
>>
>>88767293
Thanks anon. Sleep well.
>>
>>88766567
whats your sample rate and cfg scale? also euler a?
>>
>>88767405
Not for our homebrew ones that I know of
Huggingface has a library but they don't allow nsfw or celebrity ones, some anon tried and he said it got deleted
>>
>>88767253
They are great for remastering games like the gta trilogy, oh wait
>>
>>88767452
>a year ago
>realtime
https://www.youtube.com/watch?v=P1IcaBn3ej0
>>
>>88767317
I'm rarted, how do you use this?
>>
>>88767322
>armor
that's not armor, retard.
>>
>>88767423
what prompt did you cum to
>>
>>88767509
put it in a folder in your voldy's dir called "embeddings" (create it if it doesn't exist)
model is activated by a keyword in the prompt, ricci in this case
>>
>>88767492
Probably needs a gpu more powerful that the one running the game to be able to generate similiar resolution frames at real time, If not a cluster of GPUs.
>>
>>88767560
yeah I assume so, they say ((real time)) in the video and also (low framerate) but I wonder what kind of input lag you have to deal with. all of the clips are rather short because they probably rammed into shit. I'd like to see a car driving at max speed to see if the temporal part of it breaks down, they could have at least used some source videos and seen what it's capable of but of course they wouldn't showcase that since it would make them look bad
>>
>>88767531
thanks mate, I figured as much but I wasn't sure since it's a bin file as opposed to the .pt file shown in voldy's example
>>
>>88766704
greg rutkowski removes them, hence the classic pairing of the two artists.
>>
>>88766935
Kys shill



Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.