#🍥|anime
1 messages · Page 213 of 1
without having a blurry shitty thing
can u send me ur page?
Which model are you using
I use hires fix mostly, but ultimate upscaler if I really need to push it higher
It also helps to use xl models and just gen at 1024x1024 range to start
i agree
CounterfeitV30, hassakuHentaiModel
if you dont want square images you could go with 768x1024 .. i usually go with 896x1152
i prefer pony base model
I see, thank you for all your advices
you use hires fix for the generation? or like in the img to img tab?
i don’t really understand what hires fix means and what it’s doing
It takes the image and upscale then generates a few more steps at the higher res
hires fix basically fils in more noise to make image look better with some correcting method, you don't always need it tho
It’s slower than other upscale methods but tends to be cleaner
I see, thank you, I will test that! 🙏
hiresfix, upscaling all that are more appropriate for post production work
none of which shows on my page lol
i focus on prompt ideas
updated the nvidia driver ... and its kinda beome slower to gen
Well.. you ask for an anime girl licking a giant sucker and.. well sometimes you get what you get..
Which means sometimes it has a face..
something is f*d,up.. images that take less than a min taking 4 mins nonsense
Maybe xformers isn’t compatible or something
my adblocker is interfering
Weird
yeah .. think i got it sorted out
down to 20 seconds again
rating_explicit generates some of the most beautiful poses
more like rating_cute
*rating_adorable
Yeah, I generally don’t use the rating tags, seems to allow everything through
This is more what i was looking for 🙂
But that’s probably more prompt specific
i like how explicit tag does the heavy lifting w/o me needing to curate the specifics
i'll be uploading the images in a bit if you wanna check later
playing with thigh high socks
man....
the color and feminine aesthetics on steroid
only thing left is to add music
i have to force find an excuse to stop rendering
had to make a very difficult decision of the moment
i had to decide to stop generating
otherwise this will never have an end in sight
and ive limited space in my ssd
i'll post one sfw here
you really don't understand, right? why genning tomorrow what you can gen today?
what you gen tomorrow will not be the same that you'll tomorrow or the day after. carpe diem, bro
the pose is not all that amazing cause its not rating_explicit
but still...
well .. i come up with ideas now and then depending on my mood
take me as example, i'm literally dying of fever, but i'm here genning, instead of sleeping lol
you should be resting
i made my choice 🫡
HDR photo of woman, dark-brunette hair, light-blue eyes, (f_stop 5.6), (focal_length 28.0), f/5.6, 28mm focal length, sitting in bed, contemplating, sunset, romantic, warm, revealing clothes, alluring, white sheets . High dynamic range, vivid, rich details, clear shadows and highlights, realistic, intense, enhanced contrast, highly detailed
civitai server resets in 15 mins
i'll play along
woman, dark-brunette hair, light-blue eyes, pajamas, sitting in bed, white sheets, contemplating, sunset, romantic, alluring,
epic
woman, brunette, blue eyes, detailed face, sitting on bed, white sheet, contemplating mood, sunset, white shirt, large shirt, alluring, pose, dynamic angle,
skooma is a bad beast
watching an imperial addict to skooma it's a true shame lol
i love tes lore eheheh
is so rich of colors
HDR photo of woman, dark-brunette hair, light-blue eyes, (f_stop 5.6), (focal_length 28.0), f/5.6, 28mm focal length, sitting in bed, contemplating, sunset, romantic, warm, revealing clothes, alluring, white sheets . High dynamic range, vivid, rich details, clear shadows and highlights, realistic, intense, enhanced contrast, highly detailed
hands need a severe fixing, but the result is very cute
Hey people, I am looking to replicate this ribbon style. what booru tags should I use?
hair tied with intricate strings" maybe
it's pretty hard to be possible on 1.5 without a lora
none for clothing
woman, outdoors, leaf, long hair, White hair, pose, blue background, bokeh,
ok thx
Nice
Not as far as I know
ahh ok, have you been playing with pony models?
i can't seem to dig into any other regular models now
Yep. They're great
oooh these look amazing!
They're certainly fun to work with
are these with pony models?
On the upscale, yes. They may or may not be on the base model though, I can't remember.
i kinda narrowed down my model lists to 2 ... both are sdxl
rest of all the others are archived in my backup drive
🤓
Haven't heard of lightningbloom. AutismConfetti is probably the best of the pony models out there rn, though tPonynai3 is a close second.
yeah totally agree on the autismmix 🙂 lightninbloom is one of my own merges, its a lightning version requiring 4 steps
this one is lightning bloom @untold glacier
in boorunese this would usually simply be tagged as hair ribbon which won't be very helpful on it's own because it applies to many other far more common styles of hair ribbons as well. tress ribbon is slightly more likely to give you what you're looking for as that is specifically a hair ribbon confined to one lock of hair.
#📝|prompting-help a girl
#1047610792226340935 bots currently offline, you can run it locally or use online services like https://seaart.ai
“prompt”: “creates a hyper-realistic image in high definition of a humanoid creature with wings that resemble those of a dragonfly, very thin, approximately 50 centimeters tall and with a face similar to that of a pangolin, emerging from the bushes about to capture a fainted house cat. The setting is outdoors in the very dark early morning, captured outside a northern Mexican style cabin. With a night vision trap camera with a gray image that has a blur defect. The lens is scratched and dusty, stained with water droplets and dirt, adding a diffuse and creepy atmosphere. The camera trap is strategically placed and hidden at a height of 4 meters. The creature is 20 meters away from the shot so it is seen far away and hidden behind a battered barrel, its presence is perceptible, and its movements create a natural blur that transmits movement. The creature is not looking at the camera, which enhances the feeling of a sincere and mysterious encounter. The scene must look extremely realistic, as if it had been taken with a camera with very poor definition.”
“size”: “1920x1080”
}
#1047610792226340935 bots currently offline, you can run it locally or use online services like https://seaart.ai/
/Brown red sand), solid wood light yellow floor, grey white high-end carpet, vase, wood wall cabinet, mural, single sofa, table, tableware, decorations, sunshine, french window, light, luxurious and simple style), strong sense of reality), rich details, masterpiece, super clean, multi-layer), sense of space
/prompt: A golden tabby cat wearing a spacesuit stands in space, dreaming of publishing a VLDB paper.
A golden tabby cat wearing a spacesuit stands in space, dreaming of publishing a VLDB paper.
still. no bot here
To anyone doing this for some reason...If I don't know you, stop trying to add me as a friend before giving me a good reason why
what a tsundere lol
more like tired of random people adding me to say "HEY JUST WANTED TO SAY HI AND GET TO KNOW YOU BETTER! BTW YOU CAN JOIN THIS SERVER IM TOTALLY NOT AFFILIATED AND THIS IS TOTALLY UP TO YOU BUT PLEASE JOIN ITS THE ONLY WAY I CAN TALK WITH YOU, K? BYESIES"
Don't mind DMs but a friend request from someone who's name I've ever seen before, acting all familiar? noooo, I was alive in the 90s and early 2000s. I know what Nigerian Princes are
(joking)
nobody knows what a Nigerian Prince is, they're mythical beings
ouh those thiings....
it's because people love you XD
Yeah after surviving my third cancer in three years, I decided I needed a trivial first world problem to get angry at
that wasn't a political twitter problem
I think I hit about the right one
They're like 99% bot accounts. This server is infested with them.
we are not bot, we are differently intelligent... ops...
Can you smell orange?
i hear they have a lot of money and are desperate to give it to you!
smell not computed...
orange not computed...
searching for smell lora...
searching for orange lora...
loading smell lora...
loading orange lora...
smelly orange generated
mhhhhh smelly
or anything anime, manga, manwah etc
it's a recipe for a down the rabbit hole disaster
now...you foolish bot. ORANGE IS A COLOR! YOU CANNOT SMELL A COLOR!
MWAHAHAHAHA I won captcha! I trained years for this moment
1000s of upgrades
I'm pretty sure this is generated using NovelAi, not sure what the exact tags are but I think it's something like "official art". Never really used it but I've seen some people on 4chan generating images with NovelAI with this exact style
this human bot failed hi task...
activated self destruct action in 3...
2...
1...
NovelAI is good now? It was pretty bad back then
offline
It is by far the best anime model out there in my opinion. It's basically like PonyXL but it knows way more characters and it's really good at mixing different artists together. You do have to pay for it which is a shame but If you have the money for it then you can try it out.
downclocking my gpu by 200mhz has done nicely on my 3070 ti
running at 60C full load now instead of 78-80
same performance
oh woah, i might try it then, thank you!!
I'm scared of something called PonyXL
I haven't used SD or any AI for moooonths now, I have no idea what's out there
I remember finding out about it for the first time and I was confused why it was so popular, the preview images were just ponies but after seeing it was capable of doing Anime stuff I tried it out and it was quite crazy what it was capable of without any loras
the last mine gens are made all in pony
it's pretty much a very deeply finetuned xl model
wonder if it has my model in there somewhere
Literally deleted all of my previous SD1.5 loras/models and now I just use AutismMix, a PonyXL based model
a few models started using mine for its reference power
so much finetuned, that it can't trained any more. the bet you can do is mergin with other models or lora
its not a merge,its trained
better call the owner then
for what the author said, he hate merges
I've heard it's difficult to merge with other XL models, or rather impossible, not sure if it's true tho
yeah, you need a very powerful machine
fair enough
and yes SDXL is very hard to merge with
not only power but it also has a ton of errors it runs into
on the different blocks, between them, information gets crossed
for no discernable reaso
Someone was trying to merge Pony with Animagine but it just simply doesn't work, could it be the score tags that's causing the problem? I have no idea honestly
pony is too much trained to be compatible with anything else
well anything XL its a merge of pony+animagine + others it can be done u just gotta waste time playing with it
I really want Dall-E to become open
I think it would turn out even better than SD as it is
OpenAI is doing so much though and testing so many laws
I'm just waiting for SD3 honestly, need to see what it's capable of
they are called "ClosedAI" kappa
its like "Hey officer, is it illegal if I do this? pokes with a stick Hmm? is it? poke poke poke illegal? poke poke poke hmm? poke"
admittedly their moving image is very nice progress
I wish they weren't such closed up asses with the nerv to call it "OpenAI"
fires a CEO cause they're afraid he's taking AI into an area where its a potential security breach. They are then replaced, all of them....The new board re-hires the CEO
ors omething like that
"Skynet must proceed, whatever it takes...*with Arnie voice"
I forgot to mention that the only monthly subscription worth buying if you want unlimited generation is the $25 one, which is why I said only do it if you got the money for it. Don't really know much about the other options
brb reboot...also NovelAI uses SD bacjend no?
you can also use NAI on a111/forge with a script in case u wanna gen in same webui
and I think MJ still does too? It's just their models that've evolved in different directions
Pretty sure it's based on SDXL
Really? Oh wow, is it an extension? What's it called?
A Novel-AI client with more utilities built in it. Contribute to KohakuBlueleaf/Kohaku-NAI development by creating an account on GitHub.
Can you also upscale the images with that extension?
srry idk about that,last time i tried was when i still had a NAI subscription and i couldnt
Interesting, I guess I'll try it out and see how it goes
returned I am
Did you know that smelling backwards is called sneezing?
You take every smell that you've accumulated that day and shoot it out for the world to share
it does this in my strange reality that's breaking down the walls of logic day by day
i should share my smells more often with people XD
I will look into it, thanks a lot!
NovelAI do all the workflow automatically?
to have a high quality image and stuff
I think in order to get a high quality image you'll need to upscale it yourself, you can find more about it on their site, it might use some of your "anlas" to upscale from what I've heard.
@supple raptor probably knows more about this since I think they use NovelAI, I don't really know much about it honestly
Sure bud
im not your bud,pal
then maybe don t treat someone giving you a friendly reminder about NAI like so
what?
NAI models are still closed sources. Discussions about their usage are not to be encouraged.
just another reddit mod on a power trip
Thats not very friendly, please do not insult the staff
pls tell him or others to not accuse ppl of something they havent done
Please be respectful to each other; thank you.
Let me tell you something instead, since its not the first time your name pops up. Be nice and respectful to each other. The staff is handling the server in their free time so you have a place to talk about SD and everything connected. Next time we see you acting this way there will be consequences since we can not encourage this kind of behaviour. For now we do not really want any highlights on NAI and how to use the model since its obviously not a public model meant to be used in any local UI, therefore he is right to remind you that its borderline talk
oh right im sorry i assumed he meant the script was using piracy/hacked accounts,but it doesnt it just uses your NAI acc
No problem, feel free to have that talk in DMs tho, we just wanna keep the server clean from it 
I'm using 4th tail, that model based on pony
Thanks, I'm going to try that
@supple raptor is there a prompt for a single hair ribbon?
this is NAI?
eh?
images i made when dinosaurs still exist
Try autismmix_confetti/dpo or 4th tail, they are less inclided to furry gens
If you're getting furry
Skill issue lmao
bruh discord
it was showing like fine for me but then i came back 💀
guys is it showing
thx
I'm not, I'm just sharing models with people, nothing else.
i miss colab 😭 miss when i can abuse resources good people made
I want to offer some clarification before someone gets whacked for discussing something that is perfectly legal, not even remotely adjacent to piracy, allowed by NovelAI, and at the very least should be allowed to be discussed here. There are extensions/custom nodes that allow making API calls to NovelAI's inference servers using a paid NovelAI account. Which is convenient because it allows you to do things like generate a base image using NAI, then locally upscale, hires fix, and face fix the image using open source models. NovelAI provides their subscribers with an API key specifically for this sort of usage.
Unless there were some deleted messages that were actually discussing the old piracy related topic surrounding NAI, this seems to be what people were actually discussing.
pinging @wooden coral as well
Yeah seems like there was some miscommunication and understanding along the way. Talking about that stuff is totally fine from my side
Yes
deymm
ouh btw how many images can create with those 10000credits?
You can generate unlimited images as long as you use the normal resolutions, those credits are used for other purposes, probably upscaling or something else, not sure tho
unlimited normal res (approx 1024x1024 pixel count equivalent) at up to 28 steps. "enhance", which is 1.5x upscale hiresfix costs 23 anlas.
Can you upscale those images on a localUI with that extension that was mentioned earlier?
inpainting/img2img at normal res is also unlimited.
I see I see
yes, that's what I do, using ComfyUI my workflow is included in this image.
Does it work on Forge/A1111? And does it use your anlas? (I assume it does)
I find that using KohakuXL-Delta works well for upscaling and face fix since it responds well to the same prompt that is being sent via API request to NAI.
It follows the same rules as genning on their site as far as anlas usage. So, unlimited at approx 1024x1024 pixel count equivalents and 28 or fewer steps. The ComfyUI custom nodes have a setting to restrict the settings to the unlimited values. I'm not sure if the A1111/Forge extension has that feature or not, since I haven't tried it but if you keep the parameters at values that would be free on the site, then the API request will also be free.
I ll clarify the situation on my side as well. I didn't threaten anyone with timeout or anything. I saw discussions about NAI, about "how to make it works with A1111" so I just wrote preemptively a friendly reminder to not cross the line and discuss about piracy stuff. It would be stupid to censor discussion about any NAI topic, just like it would be stupid to ban people talking about sora or other AI stuff. That's it, moving on.
The other benefit of at least the ComfyUI option (hopefully the A1111/Forge as well) is that it includes a node that converts ComfyUI emphasis syntax into NAI's syntax so you can use (word:1.2) instead of needing to type {{{{word}}}} for an equivalent effect on their site.
so is NAI also the same as SD when image is not upscaled? like there is a significant difference between upscaled and native (not upscaled) or whatever, like this?
I mean details, is there a huge difference? if not we can just upscale NAI image from locally right? img to img
raw NAI image on left, 2x upscaled and hires fix + face fix in kxl-delta at 0.4 denoising on the right.
Interesting, I guess I'll try it out and see for myself. The only reason why I'm hesitant is because of the whole "upscaling costs anlas" thing. While it does allow unlimited generation I'm trying to find out if there's an efficient way to upscale the images without using too much anlas
I don't have a good example off hand of one upscaled via NAI vs one upscaled locally, but I can do that in a little while.
With my workflow, you can have unlimited gens at 2x upscaled resolutions. I've been using NAI's site at night from my phone and upscaling fairly generously, sometimes multiple times to get the best option and just using my comfy workflow when I'm at my PC all day and I've used less than 1/10th of my anlas for the month over the course of almost 3 weeks.
I see, I'll look more into it to see if there's other ways to do so on ForgeUI cause Comfy still looks a bit too complicated (or rather intimidating) for me. If I don't find anything I'll try your workflow, thanks for helping out 🙂
no, but in my experience hair ribbon typically only gives one on most models that I use. If you're getting more than one, try either moving it later in the prompt or lowering emphasis on it.
how about her? 🙂
uhh.... cute, naughty pirate....
that's the SFW-est result i can offer 🙂
yep, she is naughty but nice

civitai has new content rating policy
lets you filter what you like to see and what not
i like to see the world that's not violent
wrong model i'd say - same prompt with realvis 4.0 (sdxl)
full goku gear solid
ban yourself
anime focused they allow free gen with a cap https://nijijourney.com/
gonna look into their discord
oh snap
they ended their free option
tipical lcm gen, right?
i got similar results on bg with lcm,
this model isnt great at backgrounds, but lcm has nothing to do with it, i've used lcm models that had good backgrounds
never happened to me XD
girls and art are synonimous
guys fit in when you want to share the joy and talk about girls
im imagining the possibilities with sd3 ....
the examples i saw it goes beyond the generic theme of sdxl
so let me tell you a story...
there is this model trained to do asian cuties really well but it started with sd1.5 and does remarkable job
then the same author released an sdxl version and guess what?
it failed to come close
eheheh
btw im not dissing sdxl .. there are some great models built on sdxl 🙂
xl are simply harder to train
yep
that won me
i have an idea
which is i only have 2 models in my a1111 lol
it's time be creative
sd3 is gonna shake the boat hard soon
it's just me, or forge is not getting updated any more?
I wish there was a way to upgrade 1.5 models.
Dragonfruit (Unisex Model) is way too good.
what do you mean upgrade 1.5 models
u could try to merge it with newer 1.5 models,the ones that get updated
like turning the style of your favorite 1.5 model into a 3.0 one
they are trained or merged with specifics and you could do what steven suggested
some of the old models have already reached peak aesthetics, nothing to retouch
ive seen models gone bad with version upgrade for the sake of versioning
the asian models i was talking about earlier are top knotch.. retouching them would spoil it
if u like asian models,have u tried leosam XL?
majicmixRealistic and xxmix9realistic.. some art guru level stuff
no i havent
those two models make female portrait angelic
theres also xxmix XL but not as good as 1.5 in the nsfw stuff
yeah that sdxl version is messed up
thats the one i was refering to
i tried it in different ways, it wont come close to the sd variant beauty
i have to adjust my neurological pulse when witnessing asian cuties
but leosam XL im not sure if its dedicated to such art
its got diverse style
if there was a way to make cross models with sd and sdxl then there are lot of sd models i would love with sdxl prompt coherence
ELLA will try to do that,they removed CLIP and attached an LLM to 1.5 to improve coherence https://ella-diffusion.github.io/ still not released yet though
yeah i read about it earlier but it doesn't talk about sd1.5
it does at bottom
ohh you are right
i hope this project is carried out successfully
so much sd 1.5 aesthetics i dont want disappear
realcartoon author is another brilliant model creator
does a lot of anime
their cap (if it still exists) means jackshit if you root ur phone
those models are something you can feel the beauty in art
they give kinds of 2.8d and 3d vibe
i have successfully broken the cap before
but uh didnt know what to do with nijijourney so i just uninstalled lol
i didnt like get access to img generation for eternity but thats because i was too scared to root my phone. theoretically its possible to just use it for free forever
nijijourney check is pretty damn hard to bypass tho ill give u that
cool but i dont believe in bypassing bs policies, i'd rather want policy change, we got lot of nonsense in the world systems
now do the bee movie script
interesting request
our economy is the biggest horse sh*t in the world today
not that there is too much point in talking abut it when we still depend deeply from it
collectively trapped
its gonna break
and we'd have to collectively come to understanding
that's a bit agonizing from individual pov
aint no policy changing if money is involved
we gonna have to bypass it if we want something
its not that serious for niji journey
after the collapse, will simply be the martial law.
already happened and will happen again
two takeaways here, conditioned mindset and the assumption that the system isn't collapsing under its own weight which it is
man i love the pastel sd 1.5 model so much
take advil and get sleep
idk know why but i just love it. are there similar models guys
advil?
this and anything models are peak sd for me
not sure about sdxl of that style
its a fever pill
oh, i feel already better compared to yesterday
not sdxl is fine the problem is the model is restricted as hell
just a day or 2 and i0m again in play
you mean non censored?
civitai is temporarly down
ill give you the link in a bit
that author does great work and not censored has one pastel too
something as close as pastel as possible
its sd1.5 btw
that model and anythings are peak anime sd nothing can change my mind
who gives a shit tbh
sdxl image quality are far superior, but thats different from character aesthetics
send a few pics here maybe i know a model that looks similar
.
steven ... are they trying to censor something?
yeah i need steven's analysis on it
the belt 
naww, i think they are trying to censor the dirty ground
oh and everytime you refresh the page, the image changes lol
i guess civitai borked up something with their new rating policy integration
they made a contest for 404 images?????
tbh i was getting lots of exposed beauty on my feed even with pg selected
yeah the contests been a while
civit's working fine for me
can you load up https://civitai.com/models/94328/hello2danime
That's just a bad URL for a model page that doesn't exist.
i got bad news
the author no longer has that model
yeah no i cant download it i no pc, running through colab (either that or yodayo)
i dont think i have the permission to upload that to civitai
technically i could but thats not my model
it was a good model btw
not sure why he removed it
maybe he wasnt happy with it .. artistic temper
this is an example
of that exact model
yeah no thats not what im looking for unfortunately
something that look basically exactly like dis but less censored??
this probably one of a kind thing tho i dont expect another model to exist
hmmm but why call that pastel? that confuses me
idk man the model calls itself that
i dont understant art things i just see model name
ok there are tons of those type you are looking for
just browse the model section with any filter you might think is good
i got suggestions for you
have a look here https://civitai.com/user/7whitefire7/models
no offense but mid asf
another resource you could check https://civitai.com/user/aji1/models
i feel kinda bad for stop playing with sd 1.5 models
having to switch between webui profile bugs me
all those prompting template, the resolution the parameters
and a1111 doesnt have anything you could just and switch with a click
thx but eh
i miss colab usage ):
i use an extension for that https://github.com/Zyin055/Config-Presets.git
you can save most parameters to a preset and switch it easily
looks great.. how buggy is it tho?
havent had any problems with it so far
ahh something worth a try, thanks for the share!
i mostly use it for resolution/sampler/steps/cfg, but it supports a lot of stuff
including other extensions
i guess i could save lots of different templates with specific details for the models i'd wanna use
how is that pony loras are mentioned to work for both sd 1.5 and pony sdxl?
if you want something dynamic for anime that does different artists styles well then try autismmix model
animaginexl and aidxl can both do that style with the right artist tags. no idea what their nsfw capabilities are though.
pastel is an art medium. https://www.artfinder.com/blog/post/pastel-art-its-a-medium/#/
i havent looked at them yet but i might try them later, im very impressed with autisimmix
btw @nubby i think people using cfg 7 for autismmix is grossly low
cfg 7 doesnt justify texture quality
im trying a combination of 30 steps and 10 cfg
with euler a that most ppl are using
the leaf texture on the right side just above her shoulder is not possible with cfg 7
i wonder what the point is with prompting for artist names in sd 1.5 models and especially that are not even included in danbooru
lol civitai fixes bugs once in a while then they introduce new features and borks up the system all over again
Tried doing something that isn't a muscle girl.
how tf am i supposed to know who make these kinds of images tho
you dont have to jump on a trend, just observe conversations on discord, also showcases at civitai discord and ofc browse the models section at civitai site
dis you?
well better than this 🙂
Finally got the male version to work... damn hands!
AutismMix confetti
really cute
Styles are extremely prompt dependant
didn't know it could do such stuff
"Traditional media" does the trick
probably an author from 60
oh, i remember the lora about traditional media
very cute
autismmix has a lot of styles/artists in it, takes a little time to filter through them but it's not hard to find a style for just about anything
dinosaurs are cool
using adetailer and an expression wildcard, you can randomly change the expression of every single face
i don't use adetailer, but I use a lot of wildcards for expressions, poses, hair/eye color and styles, etc
adetailer is very good in finetuning stuff
i use it sometimes with realistic models, don't really find the need for anime
i only use in on anime models, what a coincidence XD
i simply hate realism
put me off, for some reason
like looking at a ghost
i figure if it's not generating faces i like then it's the wrong model/style/prompt, i do use it for hands sometimes though
if that mirror is cursed XD
although the adetailer hand model is mostly garbage, it rarely actually works
oh yeah, sometimes the faces can be a little boring
when you do gibli and adetailer change the face in another style
yeah that's the other problem, i've spent a lot of time finding the artists and styles i like, adetailer tends to change them which pretty much defeats the purpose
i love the curve minuscule details of the house gibli
+1
I'm still too lazy to go on my PC
Just facing the pillow with all my strengths here
your pillow is a monster that eat faces

too lazy to react, like "whatever" XD
eat your hunger and you will be satiated
like a Buddhist teaching
desair is just an illusion, overome the illusion and will not have power on you any more
an illusion of a bot who is an illusion of a human...
...brain hurts... too many layers of obfuscation.
welcome in the matrix XD
slap her

heyo
I was wondering, are there any sites like civitai? but with different kind of content or models?
one more hit and down
Like huggingface?
ehh its.... I dunno, it doesnt feel like
Yeah
It's not buggy, works on mobile and the search works
when i don't know what to do, i try random prompts on random models of huggingface and sometimes i find something cool
colab free tier cant run sdxl models can it?
i tried some random notebooks it felt like its just not capable of running xl
which is a shame cuz anything xl looking fire asf
Does anyone know an model that can make rpg battlemaps? I was wondering if with sdxl it could be possible
can do that pretty easily with google
@wet elm something like this I imagine ? 


Too bad I don't have the Kizuna Ai smuggered emoji
a lot of people think AI and tech replacing human labor is a threat to society, the reality is unless society can overcome its dependency on human labor the lust for power, control, and corruption will never end. and once that is over the next phase would be dealing with hunger for women, there are countless alluring girls on earth all loveable for their unique virtues and traits.
she is too cute, i'd ler her shoot me
the level of innocence in her face would make the gunshot painless
Experienced hitman level of calmness on her face
yeah, totally adorable
When @nova remnant tries to
me, this is what happens
cat woke up
so im sitting here, having coffee, looking around, and pondering what to do
generate girls ... or watch girls ...
Exactly, what AI is that?
No style, it's just my dazzi character lora + pony anime
its from google, just search google gemini / image generator
its smarter than SD but cant generate people at the moment idk why 
they took down the feature to generate people after a public scandal that went viral, they had woke racial bs going on like when people asked gemini to generate images of founding fathers in US it was generating black people .. and similar incidents with other figures in general that caused a lot of commotion, and google had to pause that feature
💀
this is what happens when you use an LLM to manipulate the prompt before feeding it to the image model and intentionally inject racial diversity into prompts without regard to context. Dalle3 does the same kind of shit, though arguably a little bit less heavy handedly than Gemini was.
@lyric zealot don't you dare smug me
thats something unavoidable at this stage tho, cause tech companies are claiming to -train- the AI with data, but my whole understanding about AI is that eventually it has to function w/o human dictation and unless that happens the whole AI project would be a failure by definition.
@lyric zealot we did it. we reached maximum smug
what just happened here?? 👀
i don't have enough nodding emotes 
Not really. The "AI" tools in question haven't been designed to operate independently. They are designed as tools to assist humans with completing tasks more easily and/or more efficiently. An image model doesn't need to to create images without human input any more than Photoshop needs to do so.
there is a huge difference in talking about LLM as a tool and AI
What do you think Gemini and ChatGPT+Dalle3 are? They're just LLMs with an image model strapped to them.
they are simply chat bots designed to do certain tasks, those are far from a developed Ai by defintion
the term "AI" has been broadened far past what you're picturing, and it does indeed include LLMs and image diffusion models as we know them today.
I think what you're actually imagining would be AGI.
how would you summarize Ai in its functional defintion?
no need to call it AGI
AGI is another term, but the basic idea is AI
i think they might be coming up with super general intelligence too, its basically AI
lmao, we don't even have anything remotely resembling agi. the best we can do is make things that sort of seem like they're smart before the veil is pulled back.
that's what i mean by developed Ai which we dont have yet
i would picutre such model of AI as having the collective intelligence of humans, and surpassing it by its self learning method that will obviously transcend our physiological linked limitations
apparently quantum computing would be integral to that
It's a neat idea, but the vast majority of the research in the field is not going in that direction and unless someone makes a huge breakthrough, likely won't. Generative AI (AKA: things like LLMs, image generation, etc) are where most of the research is going because because there is a proven usecase and market for them.
What you're describing is very unlikely to become a reality in any of our lifetimes in my opinion.
depends on a lot of factors how fast the development of AI can take place. but overall its more than just a neat idea, its matter of if society will break or thrive
i would assume society can remain in the current violent structures for several more decades
you think that giving humans more free time will somehow reduce violent tendencies?
what do you mean by free time?
well, if the AI overlords are doing all of the work for us in this hypothetical future, then humans would have more free time to do whatever they want right?
thats one way of looking at it, but actually human activity will be reshaped into more productive areas
like what?
But we suddenly have a bunch of really good jobs in your hypothetical future? How?
you dont suddenly have those, we have to get there first before we can ask those questions
We absolutely can't get there without asking those questions first.
that involves developing AI and removing the outdated systems we have today
you can hypothize but would be difficult in the current frame of societal thinking
we have nations killing other nations and that builds some kind of economy .. we are at that primitive stage still
Ok, let me ask you a different question. These hypothetical super AI's, probably require an incomprehensible amount of compute right? So who owns them? Surely average people can't just run them for themselves right?
why is it so difficult for you to grasp what has been long been thought of... AI = collective human intelligence with self learning methods that transcends human linear thinking
and why would you even ask who controls AI?
if you are needing to or able to control AI then its NOT developed AI
I didn't ask who "controls" them. I asked who "owns" them.
and i would guess by not able to control it would send alarming signals, but the whole point of intelligence that surpasses collective human brain is not stupid
what do you mean by own?
Who owns the hardware and/or software that runs them?
you do know self replicating AI models right?
In Crysis Nanosuit voice: MAXIMUM SMUG

you dont control or own a developed AI
but again we are talking too far ahead of time we might not live to see, thats besides the point of the conceptual point im making
ok so let me get this straight. In your hypothetical future, humans have invented a piece of software that has transcended the need for hardware?
again a self replicating AI model can produce its own hardware, you think thats too difficult for a system that surpasses collective human intelligence?
you are reflecting on human capability but not its collective potentials that AI can hve
btw there are simulated models of Ai in development that can self replicate and self sustain ofc minus the actual hardware
ok, then we go back to my previous question about humans having more free time. If your hypothetical AI can produce it's own hardware, then surely it can perform a huge amount of the tasks that humans do. So what are the humans doing?
something not bs ofc
Like what?
Ok, let's try a different thought experiment.
Let's imagine a machine that can manufacture perfectly cooked burgers. Who makes money from this machine?
you should rather focus on the underlying issue here, the economy, this relates to how resources are produced and distributed
not about making money at all
lick
wait.... how does this work ? she's holding a mirror in her other hand ? 🤔
she's using two phones, one to take the selfie, one that you can see on the selfie itself

You've failed to provide any insight into how your hypothetical future would actually work, besides (paraphrasing) "things would be better for humans". The simple truth is that if actual AGI were to be invented tomorrow, it would be owned by a massive company. It would be used to make that massive company even more money than they already have, further concentrating the distribution of wealth than it already is, while likely reducing the number of jobs that are available. Now you have a higher unemployment rate, but at least the top 0.0001% have several more trillion $ I guess. Where is this magic that you're so confident in that fixes everything and makes life better for everyone?
iphone users have to make sure that people that people know they're using an iphone.
why you dissing me 
talking about iphone, i was looking forward to their new rumored AI, but then there's report of them in talk with OpenAI and Google
that totally killed my hypes

i would suggest understanding the current debt centric economy and how that's a prime cause of inequality and corruption by fundamental design, if you are looking for system change w/o fully understanding this monetary economy, you wont have any sensible conclusion. as for Ai i started with its defintion and what eventually needs to happen unless all these researchers are planning for failed project.
I can't seem to word what I want to say in a way that doesn't come off as rude. So I'm just going to go ahead and excuse myself from the discussion.
you dont have to, you have several basic fallacy in your logic, again... look into money, politics, and how those contribute into structural violence
You haven't even made an argument past "make ai.... um... things are better now"
i think i have but you are failing to see my points
you have to make a point before I can see it 
i talked about human labor, and i talked about economy, i talked about automation ai tech
I'll try one more time. What are the humans actually doing in the hypothetical utopian future?
only think i will level with is that we dont have the kind of tech im projecting
What labor are the humans performing?
elaborate that quesiton. but again people will do what they like and what is produtive insetad of making money overy bs gimmick and frauds
so they have more free time as I initially posited. cool. now the other question. who owns the ai?
who owns you nubby?
ai isn't a living organism.
im not saying that but i asked who owns you
one could argue that the US Govenment might as well own me though given that I am required to pay taxes and if they choose they can force me into military service.
ok two broad topics in that .. owning a living organims and govt system
ai is not a living organism but imagine it to be developed enough to achieve collective intelligence, who do you suppose would be owning such Ai?
then i have to ask what you understand by AI that achieve collective human intelligence
the intelligence level of the tech doesn't change ownership.
im talking about python or windows 11
If I invented the abacus yesterday and then invented wolfram alpha today, they both do math, and I don't own wolfram alpha any less than I own the abacus.
this has lot to do with economic model how it conditions our perception and how we live in society thats destructive
such as owning, making money,
What makes you think people who have the money to create a hypothetical agi, would decide that they don't want to have more money than other people?
there are lot of conditions to that, human labor is part of that
i would encourage you to look into resource based economy and how they explain the current system of our economic model
you're dodging the question again.
im not, you are evading every info im giving you
you keep asking who owns the ai
do you see the fallacy in that question.. an AI with collective human intelligence, i dunno if thats teh fallacy on the surface or you are failing to understand the meaing of collective human intelligence
its not some wikipedia concept
dang, nearly perfect... what i want to say is: people don't expect philosophical AI discussions or conspiracy theories or whatever here. at least add some images from time to time 😉
she's mad at the AI because of her fingers
Again, you're assuming that intelligence level of the tech has any bearing on ownership of said tech.
i'll come back in some hours and take care of the survivors 🙂
sigh... Ai and automation will possibly remove human labor, that will have implications on how economy functions...
ok i have to stop
if you have the time, do look into rbe concept they talk extensively on the socioeconomics and technology
they are called venus project
there is so much valid arguments too so im not discrediting you
Tail grab
lmao
definitely failed the test
could it be something like this output\txt2img-images[model_name]
output\txt2img-images\ [model_name]
this definitely too much text
lol
you could use a script to run imagemagick to get the prompting metadata and move files accordingly 🙂
no such thing as too many cats, though
well you see right now when i generate images they go into text2img folder but im trying to get those images into folder specific to model names use to generate with
uh lol it created a folder [model_name]
ancient drawings left by the old civilization

oh its easier than i thought, had to apply the [model_name] in the right box
I found a neat extension called infinite image browsing. Allows the filtering and sorting of all your shit.
this is better then what I'm currently using.
just found it yesterday, you can also do manual search by prompts and/or models.
I have 4 OC's that I make in multiple styles, so this is great
That one is pretty good. I like SirVeggies solution better though. It's a standalone image browser that just has a flyout pane that can be configured to display any web page, which means you can use it for any of the UIs https://github.com/SirVeggie/sd-browser
Here's examples with comfy, swarm, and forge. The flyout can be closed or resized too. I have it set to "wide" in the screenshots, but for swarm and comfy I would recommend setting it to either half or full.
Here's the png info displayed which is accessed by clicking an image to expand it and then scrolling down
png info doesn't work as well with comfy because of the lack of consistent layout depending on the workflow used in comfy.
I'm just starting out and using auto1111 but I'm gonna install this now. Thanks.
If you need help configuring it, just let me know. It's fairly straight forward, but I'm not sure of your technical knowledge.
also I think I need to probably install anime specific models at some point I'm just trying to learn one model at a time so I'm stuck on the Dreamshaper XL model right now while I'm learning.
interesting, they are actually quite similar.
I think infinite image browser was somewhat of an inspiration for Veggie making his image browser.
The main thing I like about Veggie's option is that it's super light to run, so I can just leave it running all the time, even if I don't have one of my webui's loaded so I always have quick access to my images. Searching through it is way faster than using windows search through my ~160k images (and also windows search doesn't look through metadata so it isn't as good at searching)
160k images is a lot of AI porn
nah, the porn and several other folders are being filtered out in the screenshots I sent and I still have 100k images showing in those 
huzzah!

btw, you can access it from other devices that are connected to the same network by typing the host pc's local ip address followed by :4200 in the address bar. Should be something like 192.168.0.4:4200
yup. cool cool. thanks for this.
I like the live feature so I can watch my batches as they produce images.
This one didint turn out too bad
ugly 
Merged some of my most used 1.5 models and made a decent one.
that's pretty cool
really like the way it displays the metadata, it's not as fast as the app i already use unfortunately
i guess as an 'image viewer' it's a bit limited, but to look through things quickly for prompts and such it's really good
Yeah, there are better options for using as just an image viewer. SD-Browser is mainly geared towards using it for your AI gens to allow quickly and easily searching by metadata.
the nsfw and folder filters are really nice
Yue brought this thing and said it was shining in the sky. Do I need to be worried?
nah, no one will miss it 😄
thats not anime 
Anyone have any ideas on how to extract just the Lineart from outputs like midjourney? For example these headshots, I want to turn them into just line art so I can color myself. I’m trying out some photoshop filters and getting close, but curious if anyone else has had success with something like this.
like this?
This is the whole point of canny and controlnets. Might be worth looking into.
I think she likes me
wow, very romantic
she about to me mauled to death
is dat a wolf head?
canny is edge detection and gives rough, jagged lines. I'll look terrible for something like this.
Use the controlnet lineart_anime preprocessor only (in A1111 click the 💥 button between the preprocessor and model selection). Turn the resolution up to around 1024 first so it isn't too low res. then download the Preprocessor Preview image. Only problem is your image is now black with white linework, so open it in your favorite image editor and invert the image. Do a little cleanup as necessary and you're good.
is that dreamshaper xl?
sofra 🙂
dreamshaper xl is great for realism
but for anime you might love autismmix
I'll move to an anime specific generator a little later but for now I'm exploring this one to see what I can do with it.
https://civitai.com/models/288584?modelVersionId=324524 prompting is bit different with this model
Mix of pony with some stuff. It's an attempt at making pony more predictable and less dependent on schizo negatives without removing its comprehens...
cool
dreamshaper is versatile too
I'm so new to the game I just installed and started prompting less then a week ago.
good thing you are exploring it .. there is lot to explore 🙂
I'm noticing... I really like the versatility thus far... although I'm struggling with lighting right now. 🤔
ahh yes getting the lightning in specific style can be tricky
some ideas you can play around with
not all of them might work as intended and they depend on model training too
what about dark setting like on a stage... right now my characters are getting washed out in the front but the lightsource is behind them.
umm, something like studio lighting maybe?
let me try that
lol she looks terrified... 😆 I gotta fix her face.
this is neat
try keywords like happy playful
one easy hack is dynamic expressions but that will have random expressions with each gen
I have a trick I'm working with for generating specific expressions.
you could try regular words too, like happy, sad, angry, frown, etc
laughing, smiling, wink so on
oh and they are mostly anime specific
I can dial in exaclty what I want with my trick. #📝|prompting-help message
looks good, what did you do?
it's explained in the thing I posted.
that's interesting... I think I can use this to get even more precision.
something i should play with too, some of those i havent tried
Okay last one for this style and then I'll try some other things.
How in the heck did a pony model became the defacto default for all but realism styles in SDXL?
it follows artists style very well
traditional sdxl model dont do that
you can pretty much use a pony model and have the diversity of all the anime artist styles
btw there are some people using custom loras to generate realistic images with pony models that are epic
blows regular sdxl
Despite the name, it's trained on an even mix of pony, furry, and anime images.
sd 1.5 model w/o adetailer
the inbuilt R_ESRGAN 4x upscaler is gentler than 4x ultrasharp
Everyone is a furry now using pony model

Cute snake girl
Can you grant me a demonstration? (Because all I see is porn in civitai)

I believe most images in this channel are the model
you can ctrl f my images posted or smolsus images
Like do ctrl f and paste this: from: smolsus has:image
I know smol hasn't used another model in a while so its all autism 

ugly
naurrr
yesh
ofc
Oh wow, this is perfect thank you! Photoshop filters was isolating it to black and white lines but would take out entire chunks of the image ( like the eyes or hair ). I’ll try this on my end— appreciate it!
wtf is that? she gonna spank me with knife sharpener?
why's she just holding slap of metal?
spank? ya sure not to kill?
waitttt its yandere
sure hope not that thing look too blunt
gonna break few bones