#๐ฅ๏ฝanime
1 messages ยท Page 153 of 1
that's it's not ai generated XD
Ah lmao
Was just trying to find someone who was old looking but was actually a young kid or something. Eveline came to mind
Hopefully I didn't spoil the game or anything
You mean you can't figure out what we're talking about or don't know what to say about it lol
the second, that witch hunt, gives a bat feeling, but at the same time, that argument is like a punch in the belly, so, it's hard to not do anything about it
i feel like humanity is loosing something, it's hard to find the words
Which statements lolz and I agree that it is a very controversial topic
Honestly, was a bit reluctant in bringing it up
But I thought it was important to bring up anyway
yeah, i know, that was i was talking about
As I have mentioned, knowing the age of a character does hold validity when it comes to whether or not a character is underage. It's not the whole factor, but a part of it. In terms of AI Genenerated Art, most characters are completely made up and thus simply saying the age of a character doesn't make up for much
i had the same feeling reading lovecraft. like slowly falling is some sort of weird madness
it's hard to find the words
"The more you look into the abyss..."
but in apparence, everything is rational, normal
yeah
something is lurking under the abyss, something of destructive, but i can grasp it
just a feeling, a thought
yeah, i know, i read about it too
maybe that feeling is why i refuse realism and prefer stay on the cute side of art
but, oh well, I'm just getting lost in my thoughts, I digress
Well
I am the kind of person who embraces true creative freedom and I can take very progressive views and opinions, even though I can have unpopular opinions
It's probably discussion for DMs, but that's really up to you
yeah true
Know the difference.
uahahaah
With 2.5d models?
I was also like that at 17
:Clueless:
Hello playstation
Many days of not you in here
Pretty much. Also with mods for realism.
Elon is rolling out ai .. he's launching grok ai chatbot but only for paid users
I don't think I can spend $15 /mo for posting hobby pics
nad it's not funny if you can't develope your own models
id pay 15 but only if it has sexy mode
I think it's more for anime
mornin'
It's fine without weights but with weights it changes the facial tone too much
Morning
I tried with both anime and realistic portraits, it can make faces look a bit weathered
Jotaro aged backwards as time goes on though lol
Yeah lmao
Yeah sorry about that 
Well, I was active daily, then suddenly dropped off the face of planet earth for a little while
And now I'm back
It's been an eventful couple of weeks
My 0.02, I don't think the intention was to have a sexy young girl/kid or sexualized positions, but yes, it's kinda weird, and edge case
bodies that young are not like the first pic, and it's borderline sexualizing her
but I see no ill intent
Also, please call me Play. There's also an actual meaning behind my username too
maybe just warn @untold glacier that self reporting and self censoring is also valid
I don't really think it's entirely their fault. Some details can't be picked up by some individuals, whereas certain details do get picked up by other people.
Basically what I'm saying is that when things might seem obvious or apparent, it might not always be the case and might not obvious to everyone
"classical anime manga style, a chaotic 1980s manga magazine cover, over the top action"
so warm >.<
warm ๐
ai generated O.O ?
holyy moly, that's cool
all with SDXL (realisticfreedom)
pure blast, i love it
adding "gundam" dominates all, as usual
i don't like gundam XD
@steady grail
but you like neko, don't you? everyone does! look! even with cherry blossoms ๐
dropping the gundam from the prompt and here we go - a bit tight, that dress... if any mod thinks it's not SFW enough i'll remove it
we got even more sexualized images with marvel
some xman covers were so ๐คฏ
maybe the belly button
it's a little too much
but for the rest, is very very cool
no way this is generated
it's not, just a marvel cover to prove a point lol
ah
"no such thing as too many pokemon"? i disagree ๐
"a marvel comic book cover to prove a point "
ai generatewd? O.O
yep, you notice it in the merged hands and limbs - but at a first glance, it's the real deal, right?
so damn cool O.O
for a fan of the genre, the characters were just weird
but really, it's hard to distinguish, for a noob
have one more then, even with a very american background ๐
eheheh
they are very tastefull
and spidey with iron man mask
cap mixed with thor? Xd
the thing under a diet? lol
but honestly, i love this model
the style is simply extraordinary
What the heck is going on?
Wyz, I don't know you very well so you're either trying to keep the peace or maintain order, but there's no borderline here. There's no sexualization of any kind in the image I posted. No gratuitous posing, no exposed skin, no abnormal focus on children, nothing in the image or prompt.
I'm willing to let many things go, and I love this server and the time and interactions that I've spent here. But being accused of sexualizing an image of someone WAY under the age of majority by a mod isn't something I can stand by and ignore. It's disturbing.
So please educate yourself or educate me on the difference between the server/Discord rules and what the issue is in my previous post, or call in the cavalry and let's have a judgement if you please.
yume, don't let it rub in the wrong way. it's the context that's is a little ambiguous
What image?
i come back form almost a year out, to this again
things don't change around here xD
i can see where the ambiguity could happen, but i assume your intentions was wholesome.
there's no need for confrontation, the first thing I stated is that I saw no ill intention or wanting to explicitly sexualize a little girl. that said, her body is more "mature", so it looks kinda weird, that's it. We didn't censor, we didn't timeout, we didn't even erase the pic, just made a comment about it, part of the discussion
you can just do maybe the "spoiler" blur and it's ok, but it's ok to question it, even if it is non intentional, can be mis-interpreted by someone
i think that the context is a little unsettling, being the bathroom a place considered private
Mature? We talking about the same pic? #๐ฅ๏ฝanime message
i got less offending material removed here in the past, don't sweat it, @untold glacier 
artsy tifa in adult lingerie, tasteful art, too.
just be mindful of cultural differences and how things can get interpreted
let's maintain thing charming and form an acceptable ethic for every one
let's not take things in the wrong direction
i have a funny story in that regard
yeah mate, she's very young but has curves and legs, maybe she's a hardcore bicycle rider, looks off for her overall age, but it's fine, I just made a comment that probably a spoiler thingy could be better in that case
the second pic is OK
in the past, 30 or 40 years ago, people got censored on tv, because they represented the act of going to the bathroom, or shown a wc, acts considered vulgar lol
mean while on tiktok: TWERK
not too long ago we had a case, someone very offended from the anime channel, pinging us and all staff and it got bad, insults and stuff, they were calling this a channel of pervs and stuff
some people are more sensitive
or have a little daughter or X
Oh, thought you were talking about the second one. In that case, I still don't see it with the first one but 
Thanks for trying to keep things sane and in-line with the rules.
mind the fact that this pic subjects and the camera angle gives a wrong impression, as portrayed in traditional medias. hiding one's face usually implies/connotes negativity
I still don't get this though, lol. Is the bathroom a super private place where you are?
that too
today not any more, but in the past, things were very private
You old? Used to the way things used to be? 
today is a place of the quotidian that usually others don't see
not so old, but i stats to feel the generational difference, things have changed very much in the past 50 years
wait till he/she sees civitai for the first time ๐
in the bath we act very differently, compared when we are in front of others
where are my OGs... @magic wave @cyan burrow @wispy canopy @umbral depot @queen void @cold talon @gloomy scroll
you cuties still hang around ?
I'm over 30 myself, I can see this pov.
so, that, it's considered an intimate place
i'm still here at times
hard at work making a merge ๐
Ayyyyh floydian

been soooo busy this year, but winter is starting to i intend to put back work around here
the thing here is that discord specifically prohibits public servers from doing any type of nsfw since minors are allowed to join, so we must enforce this, it got tighter from Discord's side
new models needs made, new recipes needs invented, new waifus needs to be cooked โค๏ธ
anime is a tough one so only the more extreme cases get deleted or go to horny jail for a while
I posted yesterday 

Okay it wasn't good but...
but we as a community have to be alert, we want this channel and server
I was testing out a model
hey guys I want to create anime style background environment and keep the same environment on the different perspectives and be able to keep the same style on different generations as well. Any idea on how should I do that?
Meen
what happenned, you had great flow

i take offence to this, >:3
Travel to the future... Is the short answer. 
this new model i'm working on, might be very good tho ^^
it's not possible yet?
it is -> just sleep
that's traveling to the future, technically
maybe with heavy controlnets you could do something like this but it's a tall order
you could try deforum, keeping same seed and settings, but move the camera around, it should translate well enough to make animations that you can use
Well, looking more closely at your request it seems doable. Just change keep the prompt the same aside from a stated camera angle (depending on the model you're using) and use different seeds.
It's never going to be exactly the same though.
I'd probably make the background in a 3D software and then apply a combination of controlnets
make a basic blocking and the shapes will be the same for every shot
What Wyz said is likely the best way to do it at the moment.
@terse jacinth pure txt2img
oh my, isn't that adorable 
you made a chibi-only model ?
I exist at times
do you have any tut suggestions on this? cuz I am so new to this
no... it's a generic model ๐
and you are very validated in existing, i still can't keep up with you ><
i mean, it also does this
That figure to the left tho 
so jelly of veggie-senpai
scawwy
smirk
prompted for a man but got woman instead
some things never change
horny model is horny
cowboybebop.... ๐
technically, you can make a room in the sims , screeshot it and use as a background of an image XD
This is a good idea. Using games for base i2i.
@wispy canopy @cold talon @magic wave @umbral depot we will need to get in Voice chats again sometimes and chat like the "good old days" ๐
i think i was in voice once
Voice sounds neat
especially considering that the sims today is very dope graphically
on the side i made some stuff with touch designer with plug-ins for stable diffusion
rough stuff, not too dissimilar to deforum, but way more stable to make consistent videos
I made an image browser for stable diffusion
I think I was never in voice
But I did play amogus with Trexdel and others once
this image is breaking me, i don't know why 
can always come and mute if you are not the talking type ๐
Mind void
TouchDesigner + Stable Diffusion

So pretty 
that's pretty sweet, does it go sub-folder too ? i need to find an image browser that will find image in many subfolders, i have too many
Yeee 
You're making me wish I had the ability to use SDXL on Auto1111 at decent speeds. 
@cold talon so close
it doesn't have a folder structure, it just fetches all images in the subfolders and sorts by date
subarashi~
you can filter by folder tho
do share if you ever put it public
could probably pull it off with qr-type controlnet
the only thing is it's standalone and not an extension
I embed the webui inside the image browser
oh, i see
it's like a reverse extension in a way
it's pretty neat regardless
D:
you should consider making this a standalone exe
i'll give a a few more goes, it should get the text right at some point
i got around publishing some outtakes we made here last year to my deviantart, i have not been keeping up since january
perhaps, using node to run it is pretty simple though
her legs are dissolving
also there's almost no one using it so I haven't bothered
They're just covered in soap
i think the hassle of installing node.js turns off a lot of people, like me.... ๐
fair
but i really like your approach
no image browser i know of will do sub-folder redundancy and present all the stuff in a single unified canvas
windows explorer was never designed with this intention either ><
yup
I didn't like the existing image browsers so I made it
also it's way more performant
finally a man of widom xD
We really need a new explorer. One that actually has more features than less of them. And stop getting rid of our scroll bar and mashing it into a tiny sliver of the screen or hiding it altogether until you mouse over where it's supposed to be.
"don't like something, make one yourself"
"but... taht's what i did !"
i used win11, so i have the "new" explorer, it's not half bad, but still archaic bones to it
Didn't they kill the context menu as we knew it? I set up my dad's PC and it was a nightmare to work with.
Yep. 
or. don't use win11 ๐
i honestly like it tho
except they hide legacy options i might sometimes need (like power plans or audio in/outs)
Sadly, when I upgrade I'll need win 11 for the CPU threading benefits.
nope, sirveggie is too much to ask
Sir eggie
that's me
S i I r e g g i 3
so. fucking, damn close
you've always been in my heart~
no, you're the paper itself
can also press shift+right click to open the other menu immediately
also movie is starting so bye bye
no fucking way....
what a chad, thanks
My pinkie finger after using win 11

you nevre know when you'll need the use of that pinky....
Eface.... Eface what model... Eface...
h o w c a n i g e t
doing some more tests right noww
i thought the previous iteration was good, but the upper layers were all borked

so. back to step one ๐ฆ
patience dear โค๏ธ
Wait, why haven't I followed you?
this model gets hands right so damn often โค๏ธ
got a onlyhands model ? ๐
๐ you wish
i'm more of a thighs person, myself....
to each their own ๐
Hmm, does it still work as well with SDE or 2M Karras instead of Restart?
didn't test, i usually run pretty high restart steps
Where can I get private lessons
you asking for a friend? ^^
teachers go on vacation too... damn, i think i'm love โค๏ธ
I didn't even prompt for it...
no need to prompt such things, they are simply there when we need to admire them :3
Lol
are my eyes deceiving me
floydian sound
omg
and not me
not me again
๐ญ
@untold glacier he tagged you first

i know this one...
this is a robbery, give me all your waifus!
literally me
you played in ff8? ๐
i actually wanted to tag you bhut couldn't remember you damn username ><
yeah weird username tbh
I mean my name on some things is already WeedEater
maybe if you tell me which one it is ๐
asking for a friend, of course
the upcoming update of https://civitai.com/models/148871
i found an error in the out block, so i had to do some mixing again today, but now it seems the highres upscale goes well too
so, with that -> it's even better image stability, with less errors, more creativity, better prompt following
how bout you
๐
busy with work, a lot.
did party a lot too.
back into the music and the arts soon enough. it's been too long
anything with nodes
this is full SD backend + bunch of audioreactive stuff to add noise into it
SALAD GUNDAM GO!
" @cold talon "
lol
actually
change of style and topic - SFW enough i guess
ok not what I tried to do but this image is so good
like this should be seen in a museum fr

You need to stop making such fire images. I can't star each and every one of them.
thx^^ just browsing through the old images, making preview images for the model release
i think this model is going to be the 1.2 update
๐ค
So exciting. Such a great model
and whilst it's not all anime, i still like these ^^
sure, occasionally. I'm still alive anyway
You are telling me she isn't real?
No way
I have information to reveal and it saddens me to inform you all that in my recent investigation, one that's taken me from one corner of the world to the next. Into the very underbelly of Akihabara and Shibuya and even Kabukichล for an unreasonable amount of extended time.
2d girls are not IRL

anyone here watch the amazing digital circus?
The Amazing Digital Circus is a psychological dark comedy about cute cartoon characters who hate their lives and want to leave ๐ช๐
HELP US GREENLIGHT A FULL SEASON! LIMITED EDITION MERCH: http://digitalcircus.store/
Created by @Gooseworx and the GLITCH team
SUBSCRIBE, it really helps๐ https://www.youtube.com/glitchprod?sub_confirmation=1
Twitte...
i trained a 1.5 jax lora but rn sdxl training
got a bunch of loras and lost my cheat sheet of trigger words
ugh
that's not good
need a better way to manage these
use model previews extension
ill check it out
now i can finally check info about my 2100 loras
i felt the same way when i was watchin spongebob and saw a scene of squidward in the bathroom i started shakin and crying i ended up calling nickelodeon demanding them to take down the episode or i will sue them,they didnt listen to me so i went to their building the other day dressed as a giant toilet making fart noises to show them my maturity but since i got arrested for damages to public property i became a twitter/reddit activist now i can finally entertain ppl with my schizoid ramblings
you should study some more
legs are showing,ban imo 
Unless the trainer is a jerk that removes their metadata, just go to the loras tab and click this red button button. The trigger words will be any tags that appear frequently, usually just the top one or two in the list. For style loras, they often wont use any trigger word.
whys the spong melty?
interesting tip
evenin'
that's beautiful. the colors wow
what do ya'll use as a base model to slap loras onto?
meina,camelia,anylora or when all else fails my own mix
hmm I have meina and camelia .. ill try those
trying to see what i can get out of the sdxl huggingface page...
we can't see our true selves, (insert cringe dab here)
I can't find any info on ComfyUI's token padding for when things get too large for 75 tokens. It doesn't seem to recognize BREAK like Auto1111's does. Anyone know something I don't?
yup. it doesn't

you can't break early, you'd have to make a seperate clip conditioning, and average those
and goodmorning ^^
I can't even find any information on how to do that. Is there really so little info out on something that supposedly tons of people say "is better than Auto's"??!?
Top of the 'mornin to ya.
"you'll just have to figure it out"
so basically, what you want is two seperate clip conditionings
and then use the combine clip node
for all intents and purposes, that was one of the two dealbreakers for me with comfy
Oh there's a combine clip node? I tried putting something similar together but I guess that's what I was missing.
Yeah. It also doesn't have composable diffusion, one of my go-tos when breaking models. 
isn't composable diffusion the precessor to regional prompting?
but that's been kinda possible in comfy for a long time (never used it tho)
@rare cradle this is the character im trying to mold into something different sorry wrong person to ping
Nah, composable diffusion uses the AND keyword. It essentially takes two or more different prompts and mashes them together into a single gen.
They have that for Comfy? I was just about to check, as I have that extension with Auto--it's even better, as I love AND_PERP and AND_SALT
nope, i don't think there is
Then why did you say this 
and yeah, perp, salt and topk are aweseome!
Such a node for clip conditioning combining doesn't exist. Confound it all.
damn Dream has changed a lot since his youtuber days
Yeah lmao
it exists. i'll see if i can find which custome node it is
Ah, so yeah it must be a custom node.
Looks like I'll actually want WAS's conditioning concatenate node.
It's different than the above, and according to ComfyUI's creator, I believe, it's better?
But today I learned the word "concatenate."
shortest interesting prompt .... two phrases dynamic pose, nsfw
crap typos
on SD 1.5 its' really nice to play around these aspect ratios 512, 640, 768
then hires fix the image by 2x
cuts down render time significantly
Difference between Hires.Fix (30) and upscaler (80 steps)
send me the results in DM
I'll do a full analysis of them
@pure iris remember when you asked me for a fairy on a battlefield a year ago? models have improved quite a bit since then
remains quite an interesting prompt
Hello how is it going on here?

Not sure what happened to Miku #3 ๐
@wispy canopy you like these random things
as a matter of fact, i do!
explain
is that a sdxl thing or 1.5
For what info exists, you can find it on the github page that the tweet links to: https://github.com/openai/consistencydecoder
1.5
pipe = StableDiffusionPipeline.from_pretrained(
"runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16, device="cuda:0"
)

I have about zero clue how that thing works
it will me made into comfy components if it havent already most likely
maybe auto1111 components
so no worries
now open her mouth
Anime: Deus ex machena caffeina apocolypse
I think wombo does spell checking
starts up a1111 to see what he can do there
haz idea's now
are you sure? and not going to be offended? cause they are somewhat nsfw
am offended
btw how do you define nsfw properly? it something you want to avoid when you want to focus on work and can you also define work?
depends where u work,if u work in Iran for example only girls with burqa
so that involves social conditioning?
im curious specially cause when its necessary to understand what is intrinsic and what is bs
and how much constructs we are building around those two
yes but dont worry most ppl in twitter/reddit who use that word dont have a job so they mostly use it when they dont wanna see something they consider lewd
but what is lewd
terming one term with anohter i see but please define those
btw steven im not interrogating you ๐
WWE/TNA/AEW divas wrestling on national tv is not lewd because is rated pg-13 so you can see where their logic comes from ๐
so it all comes back to what we can call aesthetic ana acceptable?
mostly that are intrinsic to life styles?
i sure hope we dont train Ai based on the same human narrative that have led to all kinds of crap
torq
i can assure you they are not obscene but i think they are nsfw
send

yes also travis scott has a skin for fortnite a rated pg13 game,and we all know how safe and family friendly travis scott lyrics are
yeah thats what im talking about lol .. we have to focus on intrinsic values to human nature before we can develop a decent ai that does more than just carrying out tasks
the global system we have now are a disaster that have led to only ongoing conflicts... which includes basically three systems, economy, politics, and religion all of which have become unsustainable
and if we are going to keep building social structure around those we won't be here long to see a bright future with Ai
Yes, I'm sure if you get rid of those three, the world will be solved of all its problems. And its people. 
Further discussion in that vein should likely be relegated to the #1008652627204132904 channel, as I'd love to see more waifus and less inane takes on how to solve the world's problems here.
u better listen to the mod
it would only seem inane when arrogance is dominant, but yeah I'll acknowledge your recommendation about the channel
admin*
I too, hate it when there's an itch on my back that I just can't scratch.
owner**
sd 1.5 stock model gigachads
cackles
1.4 omega chad users
aimin
haha x2
annne

'-'
just a normal day at the dentist ๐ฆท
:p
im downloading a mod that seemed to have merged several other mods on realism
but the mod focuses primarily on portrait
popuplar mods merged together Realistic Vision
epiCRealism
epiCPhotoGasm
QGO
Analog Madness
majicMIX realistic
CyberRealistic
dw i wont post realistic photos here lol
analog madness that one is good 
yeah it is
also some realism models can handle anime too if u prompt it 
oh yeah, i've tried that with dream shaper, i will check this one too
Hey Lewds. 
I'm not bad myself. Was trying to figure out the whole ComfyUI SDXL stuff. It's maddening at times. 
I just want to be able to say BREAK and then it'll take each part of my prompt equally... but it seems any small action or slight feature costs an hour's worth of work.
lol
i gave up on comfyui.. it was too much of hassle to manage workflow intead of working on prompts
Yeah I can see why. But it's the only access I have to SDXL at the moment. I wouldn't be using it for that unless I had no other recourse.
I LOVE Auto's build. I love all the features it has and the wealth of extensions. I have so many customizations and features crammed into it as a toolset that I can make nearly anything I think of as long as I'm creative enough.
It's been my go-to for forever. I just wish that it didn't cost 6+ minutes for a 2x upscale of 1024x with SDXL.
when sdxl came out i was very eager to run it, i had my new gpu ready, but a1111 didn't have support for it then, and i was almost at the verge of giving up on SD lol.. then a1111 caught up to the speed but after trying out sdxl for a while i kinda decided for hobby and fun id rather stick to sd 1.5
A small part of me hopes that Auto's build will have a optimization run or feature adjustment so that SDXL will run as 1.x does... but I know that that's a distant dream.
i think we have to wait for the next version for some of the optimiztion that a lot of users are hoping for
tbh im not gonna play with sdxl any longer, too long to render and the quality compared to trained sd 1.5 doens't appeal to me
Lemme ask though, in the last 3 major versions of the build, have any of them had optimizations? I can't think of any. Maby an xformers version bump, but aside from that...
Yeah, I can see that. But with Eface's model and others moving forward, the quality is making great jumps forward.
well i would say 1.5 has had decent success
We're already close to the current level of 1.x with trained/mixed models of SDXL.
2.0 and 2.1 werent that much of a hit
You betcha. It's insane what people have been able to do with it.
im still amazed by the quality of trained model on sd 1.5... altho rooms for improvements is reaching its limit
And now, with FreeU and this newfangled consistency decoder VAE, it'll bring even more life to it.
is comfy a lot faster for this?
ooh that sounds promising
Soooo much faster. So much.
i'll have to keep a look out for those updates
It's amazing. The big problem of terrible faces, distant objects, text weirdness a thing of the past.
At least, a big jump from where it currently is.
so far for faces i think adetailer works amazingly good
also adetailer is regularly updated
Imagine not having to use it in many cases.
a lot of the time i felt discourged to fire up the SD just cause of default weird glitches with most models and adetailer completely changed that
I wish it worked on hands
Yeah, you'd need the person model within Adetailer.
that link you just shared, is that under development, like a new version of vae coming out soon?
That has a much higher chance of working. But...
It's out, but there's still work that needs to be done to have it integrated into Automatic1111/ComfyUI
ahh sweet worth the wait then
Well that was quick. I may need to check an experimental branch of Auto's build if they already have it working: https://github.com/AUTOMATIC1111/stable-diffusion-webui/commits/consistency-dec
I can see why the speed though. If it just makes everything better.
Seeing these numbers brings me back to the early days of SD's release.
I also wonder if there's a way to change the context of Adetailer and what it's looking at. I know with face detailer in ComfyUI, that's a setting, but I have yet to experiment with such things using Adetailer.
I figure it's just as likely to fuck up the 2nd time as it did the first time
Context would mean that it would help the inpainting to know what direction the hand is supposed to be more than just inpainting directly in the spot.
if only something could do like, the 3d depth hands and place them in the same configuration in the same spot
Hmm. Adetailer does have controlnet possibilities. But we're talking more of an automatic solution. You'd have to do it... by h a n d . 
haha
good read here for not so technical readers https://www.reddit.com/r/StableDiffusion/comments/17pagxe/openai_open_sources_consistency_decoder_vae_can/
i think im gonna wait it out a bit longer ... not too tech saavy to test run it
i said the same thing when waiting for HAT and DAT upscalers,its been months and they havent added it yet ๐ค
same with hat upscalers they out perform ersgan but looks like they havent figured out how to do it
comfy does have them tho
I'll wait a bit more for it to officially hit master
Tummy
yeah comfyui has it

Do it nooooow
Touhou discord says it's not worth it for anime (consistency decoder)
prob not then
I dunno about that. I've seen many a skyscraper in my anime gens that were terrible to look at. From the examples I saw, it might change that. 
seems to work pretty well for non-anime but not so well for anime.
openai needs more weebs employed
My hopes... dashed...
can confirm. dall-e 3 can't even make reimu.
Ack. Nubby how do I tell webui that I don't even really want codeformer...
u using the decoder for this image?
Nope. It's an older image. We just needed more waifus.
That one may have been made with the help of FreeU. Which may or may not be great.
I think you'd have to rip out all mentions of codeformer in .\modules\launch_utils.py (lines 322, 328, 383, 389, and 390) Not sure if that will cause any other problems somewhere else though.
Alright. Thanks
fwiw, the entire codeformer repo is 29.2MB so it probably isn't worth the hassle to try to get rid of it.
i wonder, that decoder seems to work with dalle3 from openai, couldn't improve anime gens too?
I seem to have screwed it up in other ways. I was trying to use this: https://github.com/AUTOMATIC1111/stable-diffusion-webui/tree/consistency-dec
But it seems like it has way too many issues currently. I can't even have it install properly.
oh, i guess the model is a little bigger so if space is super tight it could be worth it.
hmm i think i read tht wrong at openai page, they might have been mentioning dalle3 api as part of their new dev work
I'm not super tight on space, though I appreciate the suggestions. I was merely encountering an issue in the install process of the branch. I may have accidently replaced a file with an older version before starting the install process so that may be my fault.
ah crap, thats an old post
got too excited about it
i'd really hope something can be done about hands
she's got freckles on her tops too
notice i didn't indicate full body, half body or portrait, its all driven by canvas size and character attributes
cute young girl standing, freckles, yellow eyes, long bob hair, black hair, hair between eyes, jewelry, halterneck, miniskirt, bare legs, sneakers, dynamic pose.
another example of the same prompt
now do an ultrawide vertical pic
naaaaawww
i mean i can try but that won't come out how i'd expect
you know easy fix for ultra wide .... inpaint
some models work while others dont,u right with outpaint most of them work
yeah i've tried some mods with you before, some are trained for that, but not all
so many cute poses coming out i can't pull myself to click cancel generate forever
these anime and korean kpop girls are kinda similar
holy crap, didnt realize sharing images from my blog using tablet made the image resolution look so terrible
something im just proud of making today
ooooo
here's the prompt if u wanted to try and improve it
oh, okay :p
Heres another one i got
took out the midjrny lora- blue eyes, blonde hair, outdoors, best quality, high quality, masterpiece, lora:darjeeling2:0.6, darjeeling, twin braids, black tie, darjeeling (girls und panzer), cinematic, 1girl, short hair, fine lighting, bikini, model pose, medium breasts, beach, full body shot, fine art, soft colors, black outline style, elegant, very detailed Cel-Shaded With shaded detail's, soft shadows
changes it some
Dew it
green eyes, blonde hair, outdoors, coffee, twin braids, black tie, coffee(girls und panzer), cinematic, 1girl, short hair, bikini, model pose, small breasts, sandy beach, full body shot, fine art, soft colors, black outline style, elegant, very detailed Cel-Shaded With shaded detail's, soft shadows
ponders
grinz
not what i had in mind but ok
I swear a this point I can't get her. ok then
wtf
oh I see
cackles evilly
wait .. I hope thats a leg
ummm
deletes
there we go
hmm the face....
He looks like the guy from Final fantasy for some reason
switches model
adds the mdjourny4 lora back in. oh gawd.. ummm
my cup it dun broke
changes gears
Lmao
lewd miku
PLEASE , I NEED a HELP . Does Anyone knows some PROMPTS that I can to use correctly for to create some images of the female characters " WITH STRECHED ARMS TO THE FORWARD " more or less " as if wish to be embraced " on Stable Diffusion ?
What model?
Guys I have a question. Do you prefer listening to music while doing something or not?
Me and my friend argued about it
I think it is better to listen to music while not doing anything else
He said the opposite
I think it's better to listen to music all the time whether you're doing something or not.
music 24/7
reaching towards viewer would be the most likely tag to work, unfortunately that covers many other poses that involve at least one hand reaching towards the viewer from any angle. Your best option is Controlnet Openpose with an input image that already has the pose that you want. Maybe one of these I grabbed from danbooru will fit what you want.
there's more merging to be done! i will get the ultimate model!
true, but it is more better to listen to it without doing something else
I'm very rarely not multitasking 
people these days
there is no time to be bored anymore
self reflection and deep thoughts come wwhen yyou are bored
that's why philosohpyy is dying
that's why there is an increase in mental health problems
that's why meditation so good
this is me rn
it's just doing this. over. and over. and over. again.
think "oh this model might be good"
then be like "oh this idea works even better, lets iterate some more!"
Pretty please
(chibi:1.3), cute aesthetic illustration, stylized digital art, abstract and colorful paint style, happy
aight, imma do that after this run ๐

for some people, deep thought and meditation comes when they are doing something relaxing. for example, I might pick an image I want to fix and work on that for an hour if I want to meditate.
also "people these days" I'm pretty sure I'm like twice your age ๐
not sure which one is the best with this prompt xD
Time for me to decipher my old merge recipe and do it again so I can use one of my old prompts
i just browsed the "anime archive" for some prompt, threw it in
lets see what happens ^^
eh. boring anime
what was i expecting
boring. anime. as proof.
more than twice
lol
yeah true
but thinking thoughts in general is harder when listening to music
esspecially when listennig to music while doing somehting else
I mean, when the only thing I am doing is listennig to music I could at least have some deep thoughts
but if I am doing something else at the same time then no
doing highres tests, hoping the merges dont'fall apart
ok I have nitro again
"which model did it best" -> FUCK! they're both decent
why not mix them?
ty
th're both descendants
I guessed from their names
you can still mix them
try different ratios
was trying the upscale with those two, but they both seem very stable on upscale, which is a thing i wanted to have
i already know it fucks up at 1280x1280
it generally doesn't like going over 1024
unless it has a 'repeating pattern' it can as background material
well, as far as highres goes, seems pretty stable/usuable, now the only thing remaining is creativity/dumb model errors
lets go with @brittle jay's idea, and make a 3rd model, a 50/50 blend of them both
neg embeds vs no negative prompt
sd1.5 with no neg prompts?
sd1.5 model
obviously a 1.5 model type
i mean, i know 1.5 base wouldn't be able to pull such a megumin
sdxl doesn't need neg prompt rly
sdxl pretty damn fckn good
but still needs to neg prompt to nudge into right direction
right now evaluating my final candidates (yet again)
does sdxl work in a1111 yet?
it does
since when
which was released when
august or so?
it's worked in a1111 since like a week or two after release. 1.6 made it work better.
as long as i feel left 2 models beat the right 2 models, i should be good XD
come back to SD, the uhh... uhh... boobies are good!
i mean. dalle denies that
@brittle jay my model chibi's love you!
oh I never left SD ๐
I gotta stop though... I should be working through my steam game backlog
this crap is a neverending black hole ๐
shhh...
come back to the SD side, we have boobi... cookies
yeah
cookies
I only post the anime-style ones here.. more supergirls in #1071935886800982108 ๐
and batman and kratos, and of course Tifas
@keen hollow
oof
I want to know the other cards underneath in the deck ๐
omnia seems to be a hard word
๐ฆ
i think the mn is the hard part
aborted generation, want to get rid of the sad expression, but, omnia IS possible! @keen hollow
sdxl?
sdxl, my current WIP model merge
right on
that huggingface page is fun to use for sdxl since it's so fast
sdxl is a little slow on my machine because of the higher res and sample requirements
for vanilla sdxl, try https://poe.com/ you get unlimted gens. I use it for chat bots
ill scope it out
oof, phone number required. that's a no go
realistic miku... uhh... i'm not sure you want to post that one here
it might get flagged for all the right or wrong reasons
ugh
no, not dirty minds, just keeping in mind a similar policy that will get you a strike on civitai for similar reasons
@supple raptor
It's developed by Quora fwiw, fair enough if you don't want to provide phone number though.
ok.. noted
yeah, I'm weird about it ๐
hey, nubby is me 
you're having more success with that than i am with omnia
maybe if I add "greg rutkowski"...
LOL
@keen hollow they're all trying . can't complain about the image quality tho โค๏ธ
yeah "omnia" is tough
especially nr3 -> a proper sword!
omnia is srsly hard it seems
she's testing your vision
nope, can't... nope. can't read it.
perhaps that i've got so much more in my prompt too?
or is your prompt filled wwith stuff too?
dalle vs sdxl ๐
the one that looks like an eye test is dalle
this is sdxl...
3d animation pixar style chibi anime woman holding a sign that says "Omnia"
idk wtf dall-e was trying to do with her face
challenge mode
uh. i give up
put in sign "E-Face" not sure where it went or is.. shrugs
i HAD confidence in my model
have u tried with freeu if it can maybe fix wrong text?
i have it on by default
it's really the MNI part that it struggles with
dalle is the same sometime it struggles with some letters that are too similar to each other
yeah sdxl is TOUGH... just a lot of this..
redeeming factor: the hands on my model seem to be great ^^ (0 negatives used)
damn, too bad the signs are both cut off a bit
zezz
you are now simply "face"
what is "SDXL Distilled"?
its like a pruned version of sdxl
someone thought "what if we got rid of basically all of the technical benefits that SDXL has over SD1.5 to make it smaller and run faster and then they did just that" Basically it's SDXL with parameter counts reduced to make the model smaller and faster.
not thicc enough
we can't go much thiccer on this server, we're already at critical mass capn
we need a thicc one like her
good ol' crushinator
massive trunk





