#๐ฅ๏ฝanime
1 messages ยท Page 119 of 1

gonna look for it now
there's lora for the previously pic tho, so here's the metadata
Unless your prompt is extremely big there is no need for 2
If you need to use something at 2 then it is probably not the right token/tokens to use
This very good. Needs upscaling tho
can you explain about the token use how that incorporates with output quality, for example im liking the texture when using the weight of 2
Try using something at weight 10 for example
It will just give you noise
Because it's too stromg
yeah 10 would be too strong but let me share the prompt that im using and i've tested those same parameters with 1.6 and 2 for low quality and worst quality....
I'm too lazy to upscale stuff
bad anatomy, bad proportions, (disfigured, deformed, mutilated, malformed, mangled:1.2), long neck, bad hands, unnatural hands, (fused fingers, interlocked fingers, poorly rendered fingers:1.2), muscular, asymmetric, conjoined, distorted, duplicate, cloned, out of focus, cropped, (low quality, worst quality:2), oversaturated, low contrast, (greyscale, monochrome:1.1), (underexposed, overexposed:1.1), (grainy, pixelated:1.1), jpeg artifacts, multiple views, morbid
When using strength of 2 it is bad because it takes too much attention from the other tokens and also it might make the image a bit more noisy because the model struggles with such a strong strenrght
I mean that's pretty long
But still I would rather not use 2
hmm ok so i tested 1.6 and jumped to 2, perhaps i could test 1.8
The longer the prompt the more it is viable to use higher strengths
Yay nubby is here
nowadays, most model don't need such a long negative prompt
Nubby pls give good explanation
you usually want to keep it short and simple, so that it doesnt affect the style of the model too much
Just like 2-5 embeds is enough
Someone send embeds
nubby 
yup, sometimes, you don't even wanna use embedding for negative prompt
but really depends on the model tho
yes about the embeddings i've had some weird outputs with anatomy then went back to use that prompt i carefully compiled mainly focusing on anatomy, and bit on the image in general
this is effectively the same as using:
(bad anatomy, bad proportions:0.5), (disfigured, deformed, mutilated, malformed, mangled:0.7), (long neck, bad hands, unnatural hands:0.5), (fused fingers, interlocked fingers, poorly rendered fingers:0.7), (muscular, asymmetric, conjoined, distorted, duplicate, cloned, out of focus, cropped:0.5), (low quality, worst quality:1.5), (oversaturated, low contrast:0.5), (greyscale, monochrome:0.6), (underexposed, overexposed:0.6), (grainy, pixelated:0.6), (jpeg artifacts, multiple views, morbid:0.5)
emphasis is normalized in WebUI
nubby how did you arritve at those values, im intrigued
by increasing the emphasis of low quality, worst quality, you're mainly lowering the emphasis of everything else.
the lowest emphasis value you used is 1, and the highest was 2. the midpoint is 1.5, or 0.5 lower than the max so you subtract 0.5 from everything. it's an estimation, but close enough.
hmm ok and the concept about applying weights that i had was the greater values gave more emphasis on a selected word and not affecting others
in ComfyUI, that is true because it doesn't normalize emphasis.
weight just gives priority, right?
now take a look at this chungus negative with 27 embeddings (low quality, worst quality, lowres,third-party watermark:1.5),polymorphic, washed-out low-contrast (deep fried) watermark, cropped, out-of-frame, (mutated hands and fingers:1.4),blurry,Auroranegative,KFHB,bad_prompt_version2, ng_deepnegative_v1_75t,verybadimagenegative_v1.3,EasyNegative,bad-artist,bad-hands-5,bad-artist-anime,Vile_prompt3,Bad_quality,Bad-image-v2-39000, NEGS Bad Hands,negativeembed,NEGS Bad Image v4,NEGS Bad Prompt v2,badv5,deformityv6,wdbadprompt,badquality-test13-graupel-last ,badquality, colorfixerv1,re-badprompt,realisticvision-negative-embedding,an6,bhands-neg,negative_hand-neg,
uh oh
we need to go deeper with at least 50 neg embeddings
with so many negative, you don't even need positive prompts
Indeed
Beautiful one.
but honestly i wasn't getting my intended quality with embeddings but the prompts im using seem to be working farily well, however im still confused about not wanting to use a weight of 2 when the image seems to look what im aiming for, im somewhat puzzled
My knowledge is a little foggy on this, but as I understand it priority isn't really the right word. The prompt is tokenized, and those tokens are associated with different "weights" in the unet, which are values that ultimately are translated into images. Adding or removing emphasis acts as a multiplier applied to the weights associated with those tokens.
TY, I'm working on a new merge. I like this one so far, but I'm not sure it really fits what I was going for. It lost some of the "cute factor" that my previous model had.
nubby can you share one of your preferred render with metadata, i want to test it out on my end
i suppose it's more right
Nubby send him the cabbage one
I always share metadata unless the image is edited (in which case there simply isn't metadata to share). You can just search "from: bwm_nubby has:image" in here and grab one that you like.
how does that search work? not sure i understood that keyword
type this from: bwm_nubby has:image here?
ahhhh nice
discord search bar has commands. start typing from:nubby and it will give you the option to click my username, then start typing has:im and it'll give you an option to pick "image" which will show only my messages that include an image.
yes i see it now, didnt know about it
thanks, learned something new
I don't usually save art of characters I don't know but this was good
So you're like a landscape guy the mountains look nice
quick question, the default weight of a prompt w/o specifying any weight, is 1 or 0.1 ?
1.0 is default
ok. thanks
so if i used something like 0.5 that could actually negate the weight of a normal prompt
or how that works, cause i see A1111 usually starts with 1.1 when hitting cntrl+uparrow/downarrow
Ye
Or you can use () to increase by 1.1 or [] to decrease by 1.1/ increase by 0.9.
(()) = 1.1 x 1.1 = increases token weight by 1.21
Same goes for []
yeah, below 1.0 will reduce emphasis, and above will increases it.
yes, but later i started using values for more precise control
Ye
so im not getting what im wanting the image to look with weight other than 2 ....
the lowest it goes is 0 which is basically like not having the tag in the prompt at all because math.
yeah and as for my example i say from actual look of the outputs
feels like I lost too much of the toonyou-jp aesthetic in this new model 
i see a distinct difference between that image and this
smol sus' look more crisp
you see that too right?
yes, mine are intentionally a bit softer with flat colors and shading.
ahh that's intended then and perfectly good

this was the starting point for Kakigori https://civitai.com/models/92834?modelVersionId=98960
i love that kakigori mod btw, very merry feel to it
Btw how do you like choose which artstyle to apply and how detailed you want it to be because I see similar styles but with a more detailed skin shades and highlights
ToonYou-JP had major problems with anatomy and especially hands, so I decided to try merging it to keep as much of the style as possible while improving the problems that it had.
Classic prompt
I used block weight merging in several passes as well as merging in a few LoRAs to get to the point that my model is currently at. Basically just trial and error to figure out which parts of the model do the things that I wanted, then mixing and matching parts of other models that could fix the problems.
Don't feel bad, there's lots to know. A year ago I was just typing random words into a text box (yeah we only had one of them back then) with no idea what I was doing.
@lyric zealot damn, this one could have been cute if her shoes didn't come out weird.
i would just farm for the golden seed 
No one will notice
I guess the bad feet are why she fell down the stairs 
Does this look good it's form pixiv
It's not bad. The eyes, hand, and background in particular could use some work.
I never noticed these but from a normal person's pov I can tell you we barely notice details and even if there were slight mistakes it wouldn't matter
pretty good hand, too bad it grew out of her thigh
yeah, to be fair, I'm even more critical of details when I'm merging models like I am right now so they probably wouldn't have stood out to me as much normally.
that is scary to think about AI guided surgery
and i saw a video last night about high precision Ai robotics doing a surgical procedure on an egg
Her eyes are too far apart I think
Because it usually should be one eye between the two eyes
no
Why doesn't kether talk are they shy
They an OG but yeah they don't talk much idk why
blud wants a private conversation ๐
holly.......
No
im keeping the weight for low quality to 1.8
Are these OCS
OCS?
original character?
Original characters
original creation?
orange county
oh, no they are not, im just generating them with random ideas for prompt
Oh because the pink haired girl looked similar to some anime one
all anime more or less look the same ๐
Kather likes Hatsune Miku ๐ฑ๐ฑ๐ฑ
not the one from satoshi urushihara ๐
there is sweetness written all over that look of it
well, if you actually can see the now very hidden miku-mash, oh god
I like how the purple color mixes well with the green of the dress
wildcards created her and I liked it so I kept her prompt
Miku mash?
first they took our miku role, then they took our miku-mash, what will they take next? our miku stickers and emotes?!
yup, it's a very secret place in this discord, but it's currently down right now
good imo we need more steven seagal anime stickers
Who
what we really need are CoffeeGirl stickers
Did you just made this wow that was so fasttt
1200 x 480 on a 1.5 checkpoint ๐
i'm just sitting here making images. the new merge won't test itself sadly.
hmm yes, also quite stretched
2048x512 on sd1.5
1280 x 640
think i will try a few images myself too
She looks like she just fought in a war and still looking gorgeous
Uhh I wanna name all these characters ๐ญ๐ญ
touhou life is a tough life
havent thought of trying out huge dimensions but it seems fun
ultrawide ๐
the thing is i can't get the image to work on 1girl lol
heres another 1200 x 480 ๐ฅธ
this looks cool
how is your rendering not filling in with 2 characters for the wide space? weights on 1girl?
depends on the model,some models get deformed or output a black border
ahh ok, im trying weights to see if i can get 1girl to work on 2048 canvas
it does but with something weird going on
try with (full body:1.4)
ok lemme try
It looks like smolsus character
Wait
yea took the prompt
(full body, 1girl, solo:1.5), pink hair, high ponytail, yellow eyes, sci-fi, (detailed, aesthetic:1.1), cinematic, hyper realistic, dynamic angle
this hand looks bad
This literally looks like this mobuseka antagonist
well that does look quite generic
ikr
Thoughts on their art?
one of the arts out there
now thats wide ๐คฃ
last one from yae but still 1200 x 480 ๐
really nice
You're annoying but at least your art is good
speaking of 7 girls
It's crazy how their face and clothes are very detailed considering they're kinda distant
This is a marathon ๐ญ
i think that shows 14 girls .. and some more on the far left way in the back lol
One of the best things about your style is the lighting
she's got no name, one of my OC, but don't really bother naming
Can I name her please
Which type of name do you want like from which culture
I'm gonna be honest, i have no idea
also part of the reason why i didn't wanna name her

amazing texture
there is an enormous frog in my room
and looks terribly sweet
sadly, that wasn't the vibe i was going for
what were you trying for?
the sound of it jumping around woke me up
contrasting theme of a vibe and happy girl in a dystopian horror train station
put the frog in a glass jar
well im certain there is happy vibe in her looks for the audience
coudl be a new pet....
How about Mako ,Airi ,ume,momo ,Hina , sana, kasumi,sarasa?
her expression in this image looks even more sensuous
time to get weird result due to having too many lora (probably?)
not sure
not sure what i vibe with tbh
I can think of other ones if you want
i think i need to think of her personailty first before even thinking about her name

Oh yeah that makes sense
well, time to ping the only two mods online and see if it works
#๐ฅ๏ฝanime message @rose tapir @dull jackal

looks offline
9 girls with adetailer is a lagfest with all those faces ๐ต
oh yeah that would lag a bit
i had my adetailer turned on too but the resolution is so big it ddint' quite focus on every faces
yea think u have to lower detection so it detects them all
What did I come back to
Doom eternal
About what
Oh
How old are you?
"69"
Damn
I don't believe
For some weird reason
Nah man I like talking to trollers
Omg smolsus that's a smart move
I mean everyone is seeking attention at some level
definitely, like what im doing now

Perhaps he has a deficiency of it in certain places so he has to get it somewhere else, like in here
I mean I don't wanna judge him too hard
Also sorry for assuming gender
Maybe they got bullied
I wouldn't go that far
Too many girls
that model looks beautiful ๐
same vibe lmao
using darksushi
They look like doctors
Makes me anxious
lol
Just looking at the image
Do you do ai art too
Then why did you join this server specifically?
banned dw
Omg they got banned lol
i always lurk
W moderator
next time i should just ping fruit first
Ye fr

u kinda should
Fruit always be lurking

well, i always try to just look at who's online
Wow her profile is so mesmerizing
this is the person that took away the miku-mash tho

do I look like someone who will do you wrongly? 
then where's the miku-mash 
gonna try kakigori for the wide landscape
We need to do a pow of Miku and whoever gets first place gets the Miku role
I'm definitely not winning that 
i'm ashamed of the lack of miku stickers in #1152260531982434434
i expected better..

Yeah guys wtf
fake fans
I mean I can't rn cause I am on vacation in japan
dont worry ill be making a set of chibi steven seagal stickers that i will post soon (tm)
I have my excuses
chibi ftw
[redacted]
Eh?
that reminds me to upload mine, 
Nai model?
get OUT
pls smol, your chibis are supreme
Of Japan?
i wanna make physical stickers
of winning submissions
so make some super cute onessss
I think that's what nubby does
i do enjoy boiling my chibis 
Like for work
nubby pog
Or something
mmmmmmm
Man I like these images so much
Also the jar ones
And also the ones where they cabbage
๐ซ
the one on the left hahahaha
have an idea..... if i can get to fill that image with smurfs
reminds me of touhou!
well, i'm really bad at generating stuff like stickers/emotes, so yeah
lol
lies and deceit
All the best models are made by touhou players
Projectai members are so far ahead of us
would this be allowed?
the only two, that i might be submitting might be these two
me2
L-L-L-LEWD 
omg it is like smol shuten from fate
hi nubby btw
Ch-ch-ch-chotto matte!
i see the similarities
Tis one is best
Hi fruit
i read it in her extra autotuned vibrato voice tysm
i got a good one, coming up
Composition is like duplicated. Model doesn't know how to handle that wide of an aspect ratio
it does know thats why it was able to make the other ones i posted
darksushi knows how to handle wide pics too ๐
didnt follow the prompt quite right tho lol
dreamshaper just lined up all the characters from left to right the whole width of the canvas
28 chars, imagine that with adetailer 28 faces it would prob take me like 30mins ๐คฃ
yeah lucky me heh, cause adetailer didn't detect any of them
can i have the metadata for it pls?
here you go metadata
thanks
im trying to generate stickers for the event
interesting, you have a big chance to win
buddy, respectfully, you should go take a look at the entries
but still glad you said that 
okay im at least getting fruit's vote on this at least. (not actual pic cos i want to submit it)
well if you interested, here https://discord.com/channels/1002292111942635562/1152260531982434434
why thanks ๐ i will take a look into the detail
angy
PLEAAAAASDE SUBMIT THIS
got to remove the background and stuff first 

I'm conflicted, not sure to submit at expression or chibi
I mean like half the images you posted with that wide ar have weird/duplicated composition.
I am pretty sure the model is trained (/the models inside it) with bucketing with mostly images that aren't that wide ar-wise.
If you train a model with let's say only portraits at tall ars and only landscapes at wide ars then when you make a tall image it will struggle with landscapes and be better at faces and might actually force a face into it even if your prompt didn't include anything related to that. The same thing applies here. If a model is less trained in a certain ar then when you make images with that ar it will also be worse and will actually try to use what it learned from the other ars, which will result in duplicated/weird compositions.
@supple raptor I am pretty sure I am wrong about at least one thing here. Please correct if needed.
blud pulled out the wall of text just to say "I am pretty sure I am wrong" ๐
Bruh
16:9 ar... 1200x672 dimension...
btw is this rule to be strictly followed for actual image detail.... cause i do see large image dimensions tend to lose focus on faces
yeah, that's pretty much right.
apparently gotta keep it close to 1megapixel
within that range there's some leeway, but if you go much outside of it you'll likely see problems. imo XL is even worse at genning outside of it's trained resolutions than SD1.5 was.
i need to try out some of those sample res to see how good or bad images come out
better with or without text?
hmm ok, since you were trying out 2048x512 on SD 1.5 maybe i could apply that XL chart with sd 1.5?
I will never understand what bucketing does
This needs upscaling
Very good
that one has deformed hands so no upscaling ๐
not really, those ARs are based on buckets for 1024x1024 base training resolution. 1.5 was originally trained at 512x512 with no bucketing at all (meaning every image was cropped to exactly 512x512). recent 1.5 finetunes have mostly been done at around 768 with bucketing, though 512 with bucketing is also still a thing, there may be some that have used odd base resolutions. I remember hearing of at least one person who trained at 960 with bucketing on 1.5.
Just allows the latent blocks to be in different arrangements instead of needing to be arranged in a square.
What does that mean I need visual explanation or something
ok i think i get the idea of it, which is why it's obvious that extra large images lose focus on content, or what the model resolution is trained on .. but i dunno the term bucketing is..
it's a training technique that allows training on non-square images
ahh ok
so far ive seen if the expression of a character has to be the focus of an image .. all those large images aren't delivering it
this image here you can see the faces but with loss of detail
has to be a thing about aspect ratio
So latent is actually just 64x64 vector or something but bucketing allows for something like 48x64 vectors????
I am confusion
the background is pretty good but the subject / char is losing some detail
Most these images have 2 vanishing points
in fact those background are petty amazing
close. as one dimension gets smaller, the other will get larger so that it's still the same number of vectors, just arranged differently.
Or something like that I don't know the exact term
this one prob the best one of all
Wait fr
Now I just need to study UNETs for the third time so I will finally be able to understand that, because each time I tried to, I failed
i think in general sense we should be able to generate images of any size and dimensions freely, but we have to go along with technical limitations... for now...
What sampler are you using?
euler a
I wonder if someone can generate a Pokemon gijinka wallpaper for me
I could probably but I don't think that resolution will work well for 1.5
๐คฎ
yep, something like this
the AI is not trained enough to generate images on loose canvas
also what model are you using
Thanks for visual explanation now I understand it much better
Loose canvas?
any size
Oh
I mean
Technically
If you were to use a model that is trained on pixels
You could I think
But then you need a hella beefy gpu
cutemix
yea
u can,just play with prompts
im gonna play around with 1344x768 and 1536x640 to see the output quality i get
Just checking but by 1.5 you mean models based on 1.5 right?
ill try with 768 x 1024 cuz these ultrawide take too much time on a 1060
yes, not sdxl
which is what I thought just bc the aspect ratio and amount of detail
generate one or two at a time for testing purpose
Nah man we could do insane details and ars for a pretty long time now (in ai technology time)
That's pretty cool
do you need a special extension or
does it just work
maybe I'm just under a rock
We all been there
the ones i made had 4 loras thats prob why it had those details
ah
still
yeah most my generation is messing around or 512x768 with low expectations
u have to use high res fix ( latent) so it adds more detail , if u just generate it normally it wont have good details
Examples
Do png info on last one
heres before and after ( high res fix latent )
I like using upscalers instead of latent
Idk why
I just like it better
Tho yeah latent should probably be better
But takes more time
sometimes latent goes crazy and adds bunch of stuff u dont want so i guess it depends on the picture
I mean just change denoising strenght
i keep it simple
i did with some pics but if i pick a low denoising strenght it goes blocky and if its too high it gets deformed so prob some loras i use are incompatible with latent
I like higher denoising strngth
setting denoise too high can change images from the intended output i think
0.6 my fave
when latent looks good, it's the best option. but it's success rate is far lower than using a GAN upscaler with a bit lower denoising.
It makes them change more from the before the hires fix
*from before the hires fix?
Ye
ahh ok
Damn you were ready
i got that shit favorited XD
im intimidated by math
same i mostly copy pasted and changed numbers at random til it seemed to work
i used desmos ill know if it worked in a few minutes
Normal math intimidates. Linear algebra not.
i kinda lost touch with algebra from lack of practical use on daily life
this one prob the best one i got with latent at 4K
looks nice
Same. I was good at math in school, then I never used it IRL
old gen
oooh the texture came out so nice
another from the same time
lightning looks beautiful ๐
kinda looks you pasted anime character on top of a real garden photo
this one worked for a 2048 x 1536 gen
constant zoom might look better
most of your image backgrounds are really cool
i use 3 loras for backgrounds thats prob why
ahhhhh.. i dont use any LoRA at all
it gets deformed on latent most of the time but when it works it works
those must weigh in on the render time
yea it goes up to 15mins
wow
more oldies (2nd has some edits)
i kinda like it more to put a model do the entire work
this one has a few edits too
thats the terminator bubble
Bruh
Just merge the lora with the model theb
well yea, im ok with downloading some of the merged mods too
im actually waiitng for eface to finish merging and perfecting a XL mod
I mean there are already some very very good ones for sdxl
naa there isnt
this one has no loras and no neg embeddings ( prob why it got deformed )
Even for anime
almost all of them are falling short badly
i mean i get more kicks out of SD 1.5 than XL mods atm
FateZero: Fusing Attentions for Zero-shot Text-based Video Editing
Didn't try this one but seems good https://civitai.com/models/124189?modelVersionId=164729
ๆจกๅไป็ป๏ผChinese Part๏ผ I ๅผ่จ ๅจๆฌไป็ปไธญ๏ผๆจๅฐไบ่งฃ๏ผ ๆจกๅไป็ป๏ผ่ง II ้จๅ๏ผ๏ผ ไฝฟ็จๆๅ๏ผ่ง III ้จๅ๏ผ๏ผ ่ฎญ็ปๅๆฐ๏ผ่ง IV ้จๅ๏ผ๏ผ ่งฆๅ่ฏๅ่กจ๏ผ่ง้ๅฝ A ้จๅ๏ผ II ๆจกๅไป็ป ๅจๆผซๆ็ป่ฎพ่ฎกXL๏ผๆ็งฐ AIDXL ๆฏไธๆฌพไธ็จไบ็ๆไบๆฌกๅ ๆๅพ็ๆจกๅใๅฎๅ ็ฝฎไบ 2...
how are the gallery pics looking plastered?
i was using this for a while https://civitai.com/models/118729/reproduction-or-sdxl its ok but there are some lackings with the output you can get
Madlads
yea they all look like sd 2.1
but other than using Mod i have issues with resources and render time XL needs
Man all the recent research papers and projects have hella weird names
by the time XL finishes rendering 1 pic i already done like 10 pics at normal resolution on 1.5
Either anime/weeb references or very weird abbreviations like LoRA, LoCon, LoHa, and some other stuff I forgot
lycoris too ๐ค
Lycoris is anime reference
Not weird abbreviation
I mean technically both
But they all are
must be SDXL
might be, but there's nothing there that would surprise me from a 1.5 model.
thanks but no
spam enough bottons and you can make anything
Could easily by 1.5
really? the cushion behind her looks like its XL quality
its 1.5 im still waiting on models i want to be moved to xl
ahh interesting ..
text and arguably realism are the only things that XL is really noticably better at. It's significantly worse than 1.5 in a lot of ways for anime.
yeah im kinda good at this stuff when im not making random crap to keep my mind from making sad books
think so too, the base XL model images kinda looked too artificial compared to SD models, but few trained XL models kinda had very rich texture to it
might have to fix that hand tomorrow
found it like this one here people were asking if it was XL too a while back
She is still as fabulous as ever.
she's got the face to steal any heart
she certainly stole my heart 
were you using calicomix for these images?
I found a problem with SD
https://reddit.com/r/greentext/s/T93TY34oXt
ive hit the point where im not sure which setting im doing wrong https://i.imgur.com/o1HC0B1.mp4
its like not morphing at all
i think i got it but ive said that like 5 gens now
Trying to make a new alien OC
She looks like a ghost.
im glad to have camera distancing working out nicely
I'm feeling strong warrior girls today
Blue alien momm
little meow
@nova remnant love your work
Good morning everyone
Meowning
I got a cool image but I forgot the nsfw negative

@nova remnant no booli ):
can you give me the info of these 2 images ? 
just redo it with same seed
and how do I get the same seed
DM me that pic and I'll tell you what seed it was
damn this new model are surprisingly good at gore
too bad can't send them
Its there, why would I need to send them to you, its not my image even

cuz I'm curious what it was
๐คฃ
don't worry I don't judge
beside if it wasn't your there's a chance there won't be any info on it to begin with
and since you're making excuse after specifically say that you forget nsfw negative is even more sus
is it so bad that you need to shift blame onto someone else
๐
But the face expression changes her entire personality
Wait
I read that wrong
Lmao
You were curious about the picture oh
Like I thought you wanted me to send vcm07's picture to you
lmao
I got very confuzzled
meow
D:
funny those are what I made based on YOUR prompt
She ate too much chocolate and strawberries
diff neg tho
Yeah but I wanted even closer xd
she really loved those jam
Who says she isn't 
loves them so much forgot to wear clothes
want ghost? here
Needs some blood on the tee 
hello there
Very beautiful pic.
one side is an eraser
the left one certainly isn't eraser
@oak spoke are you using 1.5 or sdxl
sdxl
o:
I open the open the enfjogntyf
If anyone wanna buy I am selling this card for 200
200 of I am not sure what yet
Yes
people
This is epix
And cats
you want male cat wearing skirt?
Bro what
you asked for it
and boots
Nah man
yes you did
Boots and cats and boots and cats and boots and cats
Look at me, professional beatboxer
hello purple man
Hello cat with puny graphics card
i wish i could ifnd my battle cats account number
I thought cat said "people" as in "hello people" or like to take the attention of the people as in "people!". So I said "and cats" as to mean that the audience of his exclamation should include cats as he himself is a cat.
Assuming gender feels bad
But I will just assume I am correct
I fucking hat ethat wmoji

Why you bully them
No bully, name says 4gb
btw cat
You know what they say about big hands eh
yes I can reach lots of keys in the piano
They take longer to wash?
that they're short?
They need a bigger mouse?
I would put spoiler on this and say blood warning
What no
True
MX Vertical > others...no issues with wrists due to mousing in an unnatural position
that was just.....her makeup
Tell that in court
skill issue
Skill issue
get old, talk then
Might as well get an ergonomic keyboard haha
wait til you actually see my 'abused' prompts
Arisu
Ayo
I'm old already stop, want me to turn to dust ? 
I always had trouble making black eyes
Don't let your dreams be dreams.
Guys and girls and cats and boots and others
Listen
This channel
Is amazing
I watch this for a year already
better than 1.5 generally
I have like 400 pics of various characters holding guns, it's an obsession
bullets are sharp
Indeed. Having dreams is amount to suffering. And also as Schopenhauer said:
What disturbs and depresses young people is the hunt for happiness on the firm assumption that it must be met with in life. From this arises constantly deluded hope and so also dissatisfaction. Deceptive images of a vague happiness hover before us in our dreams, and we search in vain for their original. Much would have been gained if, through timely advice and instruction, young people could have had eradicated from their minds the erroneous notion that the world has a great deal to offer them.
huh didn't remember making this
Literally me when

god of butterfly have snuck into my images
It would be better if there were nothing. Since there is more pain than pleasure on earth, every satisfaction is only transitory, creating new desires and new distresses, and the agony of the devoured animal is always far greater than the pleasure of the devourer
Me and the boys at 3am be like: "Human life must be some kind of mistake. The truth of this will be sufficiently obvious if we only remember that man is a compound of needs and necessities hard to satisfy; and that even when they are satisfied, all he obtains is a state of painlessness, where nothing remains to him but abandonment to boredom. This is direct proof that existence has no
real value in itself; for what is boredom but the feeling of the emptiness of life? If lifeโthe craving for which is the very essence of our beingโwere possessed of any positive intrinsic value, there would be no such thing as boredom at all: mere existence would satisfy us in itself, and we should want for nothing."
Sorry for spamming pessimistic philosophy btw. But Schopenhauer is just such an edgy boy.
stop thinking jordan peterson is deep
nah thats fair
Diogenes = gigachad, sigma grindset
Socrates = probably slept with little boys
Nietzsche = literally mentally insane. Also horse moment funny.
Schopenhauer = edgy kid who thinks he hella smarter than everyone else but is actually just annoying. Tho has some good points.
Albert Camus = gigachad 2.0, actually an alpha male, absolute player. I have crush on him fr ong, and I'm not even gay.
reminded me of this
https://www.youtube.com/watch?v=7dGJi0Rt4xY
now do one letter for every color 
you can do it I'm about to eat

Omg eow
Insane
Oh damn already
Didn't even need to ping
โ ใญใใจใฎๆฎใใใซๆฅฝใใๆฐ็ฟๆ
ฃใใญใใใใฃใจใใจใใใไธๆ่ญฐใช้ๅ
ทใใญใใใใใใ
https://amzn.to/3HwSUNU
โ ้ฉใใปใฉใดใใฝใชๆใๆฏใๅใใใใฉใทใ
ใใกใผใใใผใฟใผ ๅฐๅ็ซ S ้ทๆฏ็จฎ็จ ใใใ ใใฉใท ๆใๆฏ ้คๆฏ ใฐใซใผใใณใฐ ใขใณใใผใณใผใ
https://amzn.to/3QOKllD
โๅฝใใฃใณใใซใงใใๅบใฆใใไธๆ่ญฐใช?ใใฎใใฉใท2็จฎ้กๅใๅใใใใใ้ ใใฎใง่ณผๅ ฅใตใคใใ่ฒผใฃใฆใใใพใใ๏ผใขใใพใณใขใฝใทใจใคใใชใณใฏใไฝฟ็จใใฆใใพใ๏ผ
ใๆๅ็ซๅ็ปใไปๆใฎไบบๆฐๅ็ปใในใ๏ผ
็ซๅณถใงๅบงใฃใใใใฃใจใใ้ใซ็ซใซๅ
ๅฒใใใ
https://youtu.be/M5PbLfVGOQs
ใชใใจใใฆใ่ใฎไธใซไนใใใ้่ฏ็ซ้ #shorts
https...
watching preying mantis eat is terrifying
Did you try using adetailer?
True and real
willingly clicking on shorts 
This image is so cool
I mean it's cats
What can I say
I see cat I click
Actually smart move
Blocking shorts
The missing G is me
this is awesome
trained on jwst images
i wonder if u could generate like, solar systems in the shape of boobs or something
blinks, the actor mark harmon of ncis just send me an invite to follow his page.. what?
ummm (smells) hmm fish.. fishy
meh his dogs the reason abby left
anyhow back to art
Can't tell if Mumford style knows detective pikachu, but it's certainly trying.
This Haru Okumura there?
no, Shirley Temple
used a lora this time
@supple raptor https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/issues/241
i submitted issue and i think it should be safe to update now
larger one for those th at want it :
๐ค she needs wings
not a single one of my opensea artworks sold
Just came to say goodnight




