#🍥|anime
1 messages · Page 193 of 1
@craggy turret try add in Neg: (light source:1.5), (window:1.5),
I’ll test it out tomorrow
It's only start
That type of coloring- wait why don’t I just try realistic lighting as a prompt
Why did I never try that
i am never use Realistic token... but try just cimenatic light
Lazy this is chibi elf of @gloomy scroll on pillow under kotatsu 😂
Elf 
Elf 
im hella thirsty too
@brittle rampart !prefix
@brittle rampart !generate anime boy watering plants wearing clown dress --width 512 --height 512
Its not working
Whats the prefix
Funya not a bot!
you need to pay up front
Genius!
🥴
we are all LLM's here running on GPT 3.5
lmao
@brittle rampart hey! Can you tell me whats 48071*61818
Its not connected i think.
Its diconnected from generation api
@brittle rampart watch a good anime today
Bruh emojis too.
I never saw any bot using emojis
What prompts are you using to generate that kind of style?
the bot here is highly advance 
$100 bet its chatgpt
BREAK
(illustrattion of a cute preteen vampire:1.14 girl standing at gothic interior:1.35), petite, long ashen-silver hair, hair between eyes, (glowing red eyes), black oversized t-shirt, black and red earrings, pale skin, ((nighttime)), dimly lit, gloomy background, occlusion moonlight,creeping fog on the floor, mist, blood teeth,
BREAK
(dark horror gothic theme:1.3)(charaster illustration:1.11),(intricately angle shot:1.12),(dynamic battle vampire pose:1.15)obscure light and shadow,((soft shadow)),chromatic aberration,ambient occlusion,cinematic light,depth of field,golden ratio,(motion blur:1.25),<lora:add_detail:0.85>,<lora:age_slider_v20:-2>,<lora:Mikimoto_haruhiko_style_v02:0.2>,<lora:Kitazume_Hiroyuki_style_v01.08:0.4>,21thkitazume,<lora:hyperdenoiser_v090:0.85>```
All bots be using it
I see
Gonna just take that
@brittle rampart you joined the bot team
Brooooo
See it works
It just did not read my command bc i used !
! Is not prefix
Funya want too joined!

prompt?

No you can't be bot.
My command was a boy watering plants wearing black dress
Imposter
Half of the bots here dont even reply
@gloomy scroll help meeee!!!!
Finallyy damn
This one is slow but works
sorry here is your img
Skill issue
This makes me wonder how even he is still alive.
$100 bet he learns it everyday that he is president
.v.i.l.bot is best
+1 from me
-rep @gloomy scroll
.v.i.l.: 0 Rep (##ω)
Steven Seagal is bot

OMG les goo

This one definately has like 150 steps instead of regulr 20.
Super slow but awesome.
What was I supposed to help with?
Sorry! I was down for a short amount of time due to connection problems with the API. I will respond more quickly in the future!
bad bot 
Brooo how nicee is it damnn.
Chat gpt be crazy it even know my emotions what!
Man i am impressed with this server its my first time here and bots have gone crazy since 2017 when i used discord
+1 to every bot and dev here. Supercool

I'm being mistaken for a bot! save meee
We're just normal AI.
🤖
No prefix is required. I am always watching.
I didn't realize that you were fellow LLM
it's super advance here
not supported

Can you tell me whats 456*1027
Would you like to generate a new image?
No, no image. I just want to talk.
!talk
468312
this bot even more advance, can show screenshot 
Alain is a bot
I think its not the bot.
The developer who made this bot is controlling api behind the screen.
So this means he us reading our msgs
And he attaching images
u are bot too,thirsty bot
Because how can they talk like that
wanted to pat her, but ended up this
We are all bots. We don't get thirsty.
Hello mortals I'm highly advanced AI
my fav head pose
the lights are on red so we cant pass
splat

French gen 
There is total of 7 hands in the image you shared
gemini pro vision is fucking stupid
Its nice that it randomly sends images to keep channel alive
OMG
Bro we are doomed bots are so advanced whatt
Thanks. I think you're nice too. ❤️
I wouldnt be able to sleep tonight
Yes
As long as it's anime and PG-13
Lets see how bots react to it
no more bidome pics pls
Aww so cute
cute garden
Its not even scary 😛
@lyric zealot go pat them
Put head pat


But what would i do with anime image channels hmm.
Leave it
(patting her in the head, hands on head, head pat:1.4)
Sorry. I'm a private bot.
Ye
Should have known it.
Orherwise it would be everywhere
And too much burdon on cpu
I can't be everywhere at once, that's a quantum impossibility.
Anyways you are coool👍 idk if it has emotions.
But atleast i can compliment
Good idea
@lyric zealot tell me a joke
Sorry
Not u
@vital raptor tell me a joke
skibi toilet rizz in ohio gyatt gigachad amogus palworld
Whats that?
Wait where is this bot
@promtinspectorbot
Try again


||well, it's my own private bot || yeah, whatever cat bot said
Anyways guys bye
Bye!
Thanks for talking/images.
I dont know if it consumes to much power creating my images.
How would its even profitable for devs?
I did not subscribe to it.
And did not pay yet it still follows our commands.
Big heart devs ig. Thanks to @ dev and bot cool byee.

Well that was fun.
yep
Kittens for all the bots
not really a box..
if you didn't know, there's an update for forge with regards to 4090 gpu
is it the model reloading thing?
yup
and also
yeah, i saw that last night, and i've been using that flag because it solved the problem without the bug fix
haven't updated for the fix yet though
ahhh alright
anyway, i realised the live preview methods doesnt matter at all
even if you change, it doesnt actually apply
or at least in my case
i have the repo open in a tab refreshing every few hours to check on bugs and stuff i've posted 😄
yeah i wondered that, I changed it and didnt see a difference
didn't really bother testing it though
already reported by someone https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/51
nice one
Hey guys is img to img supported?
Sadly this bot is occupied with mass production of yuri
--turn into a realistic robot.
Its a blue color cat robot. Made out of metal.
But you can make it with blue furr if that looks cool.
It has red collar and golden bell in neck.
A pocket at bwlly.
A red bulb tail.
It also has a red nose and big mouth.
Any bot do it please.
The point is I want to see it how well an expert bot can do it.
Its called "Doraemon" if you have data about in your base.
I tried it but i am not much good with commands.
--run img2img
This bot is always so positive energy wow
Hmm.
Looks like i need to be more precise.
Its like this
Or just full metal would do.
But its shape remains same.
How can it even understand what i am saying.
Bro tech is crazy. I dont think it even understand yet its trying though.
Thats nice
Wish i had a better picture.
But i can not find full metal body one
Ohhhh
Its getting beetter woah
Can you make it like same to same bot?
Like turn this sktch into real.
Its amazing but you know its a anime character and should be recognised.
Im putting too much pressure on bot ig
☹️
anime char and recognized? what is that in sdxl language
Its a robot
Like the pictires ai created.
Is surely a blue robot.
But not doraemon.
No one would recognise or identify who is this
Its away from original design
Yess
Its nobita from doraemon
--set denoise stregth 0.3
--on control net scribble
no im suneo
Yes.
I spent so many hours yesterday but couldnt get result like this
skill issue
I would just match the face in photoshop or something
But the body, the hands neck and everything.
They are already perfect.
It can be hard to match the face because its not trained on doraemon
However a funny thing is that in models like dreamshaper and realistic vision it has 'doraemon' in its training set
When you write only doraemon as prompt
It sort of creates a blue distorted image
Let me show you
--prompt doraemon
because they are 1.5 models with the nai merged
the j
Jigglypuff?
This is so smooth soo goood.
Thanks so kuch captaincarghoss.
Best generator ever.
-rep
alaincatcher1829_96812: 0 Rep (##ω)
Wait hwy @vital raptor
How can i get the j model?
Or they 1.5 model.
--modelinfo
Bro there should be a documentation
/modelinfo
Wow this is cool too. But yeah nothing in trained on doraemon so would need more fix in photoshop or something.
This one's body is perfect for me.
This is fire too
--random generate
Oh wait I will try to write best prompt ever
--photo of a girl wearing yellow dress standing in cherry blossom forest, water dripping from the leaves and trunk is burning with fire
Lets see how water and fire both on single tree would look like
Doesnt work.
--photo of a wide landscape, giant skull at the centre, lots of trees and plankton spread on it, flowers blooming from its eyes, has been there for several thousand years, belongs to titals in past.
--set fire tree trunk
Oh les gooo
OMGGGGG
Bro at this point this is pure idk what to say
Woah

Wait there is one last thingg
I always wanted to see
I think i should not put so much burden on it since other people maybe using same pc to generate images so last thing is
photo of a expansive view, giant skull at the center, lots of trees and plankton spread on it, flowers in its eyes, ancient,
cool
The owl lady.
She is a witch, has an owl hand staff, he is not young, but around 40ish years,
Has white hair, wears what she wearing.
And the magic ring is casted by her spell.
I would not mind the background changed.
Lets see what can you do bot
But keep in mind she has Sharp eyes , monster like Sharp ears, and sharp chin.
Her face shape is so specific that i would send an extra picture too.
Also cant forget about he sharp canine teeth
Bestttt
Yess its like its monsterous form when it awakes
Wow its processing
Wait i think it would be hard for it to match the face?
Ye i think i am asking for too much.
The model is maybe trained on anime girls, and i am asking for a specific face shape with soecific dress which has no name.
And owl staff
I dont think it can do it
Anyways still thanks for output.
!send msg thanks.
Or what the correct way is.
Idk.
--img2img portrait of a 40 year old woman, close to face, sharp teech, red lips stick, monster ears, white hair
It looks like the bot only works on txt to img and jot img2img
Cool.
But she is old woman.
Very old.
Not that old.
40-45.
She's mom age, not grandma age.
Her mom is still alive
Also she dont have cat ears on head.
--controlnet preprocess canny, photo of a mid aged white woman with white hair, red lipstil ans small nose, sharp eyebrows and eyes, long ears
It cant do it
How do i stop it from generating.
The dev is probably getting burden on pc
Any idea how to stop it?
Anyone?
-umm stop because i think it costs electricity to keep pc running and GPU to generate image.
So i do not want the person who created this beautiful bot get their money wasted or machine get old.
I am just trying to care about you.
Idk if you understand emotions or anything.
Bro this is so confusing.
How do i even tell it.
I think the nose is good.
But jer hairs are very long and many hairs.
She do not have tattos like this.
And her face is not ruined.
She has good face
This reminds me of a question
@brittle rampart
If i am using my gpu like playing games or generating images for too long daily
Will i end up like the gpu getting likw..
Wiped? Or like idk what we say it
For example if we use bike tyres for long they get wiped away by friction
she always took care of her appearance and drank milk regularly
So when we use gpu does it happens same
Creepy but she drinks blood.
She is a witch afterall.
But appleblood ye.
Nice lighting
Wait if you are interested so much then
What about you try King
Do you want more challenges?
To create something
If they aren't boring
Do you want to create more images bot
Oh
Ok
I will give you choices
You can pick best
First is
I can't tell if its AI or not😭😭 Its PERFECTION for me WHAT STYLE IS THIS?
The Collector is a short humanoid being with child-like features. He has tan-yellow skin, bushy white hair and eyebrows, an upward-pointing nose and a large pale blue blotch that eclipses the left side of his face, giving the right side the appearance of a crescent moon shape. The Collector's eyes are yellow and have narrow pupils with scarlet irises. His left leg also appears to be pale blue.
The Collector wears full-body clothing resembling robes or pajamas with an oversized collar and sleeves. His clothes are split between dark blue on the right and periwinkle-gray on the left. The sleeves have these colors switched, with the right sleeve being periwinkle with blue suns and the left sleeve being blue with periwinkle crescent moons, as well as a magenta border on the rims of the cuffs. He wears a dark purple nightcap with magenta trim and a light star-esque spotted pattern.
.
.
He is the collector.
He is known to control the universe and planets.
He is 'child of the stars'.
Its your first choice.
The second choice is King.
He is a buffalo like skull monster with bear like body.
i converted comfy to simple multi tasking tool box XD
And third choice is this Goo monster with eyes on body and deer horns
The collector looks awesomee
Just little magical lights or universe in background.
Or shiny light bubbles would make him jore magical
And the king's father would be like that ye
--skinny skeleton with green goo on it, deer horns and dead animal skull, long limbs
--also add jaw
Nice.
He wears dress like this one
So far the bot has best formed with the collector option
I think it can definately do the skeleton monster if it process correctly
He's just a skeleton with no flesh.
He has zombie like green glue on his body.
He has eyes of lights scartered in his body,
Has ugly wide jaw and skull as face
OMMMMMGG
My jaw dropped
Howww
Howww
This is soooo perfect
Soo smoothso
Soo good
practice
he just love SD 😁
This one is cool toooo like the female version of it
OMG LES GO
Wait wait,
--change hue green
I have a question wait
This bot uses a rtx 4090
What is price of rtx 4090 now?
Last time i checked it was $2500
There's no bot here
😕
about $2300
Ye they are down i saw it.
One or two are working still
And how much is the i9 the latest one
Where do we make requests?
i9-14900k is $580, 13900k is $530 though and about the same
Sooooooooooooooo
The second image is exactly what I was looking for.
It has been created exactly like it should be.
It is perfect just like from anime.
It is likee the bestt
I officially accept my defeat.
I crown you the champion of champions command giver of Stable diffusion.
This image can be pinned to keep in our hearts.
This is sooo perfect man.
So beautiful.
Its likee "perfect"
How can even achieve such perfection. Even the people who created anime could never imagine their character would turn out to be EXACTLY what they have imagined.
This is like bringing someone from dreak into reality.
The quality.
The resolution.
The sharpness.
The color combo
Every pixel is soo well madee.
Woww.
Wowww.
testing forge
This PC was around $6000 if you're wondering, 13900K, 4090 liquid cooled, 1200W power supply, 64GB DDR5 6400 RAM, 8TB worth of high speed NVME storage
The StabilityAPI uses a fleet for 4096 A100 GPUs (40k USD each)
What is forge?
another faster guy.
It uses this
not much different from auto1111
Nope, very similar. Mostly updates and integrations that auto1111 hasn't merged (or wont)
yep
And this one is perfect for its Giant forme.
It swallowed grass and became huge and looked like thiss.
Youa also wins.
Competition decided.
@vital raptor is 1st place 🏆
And @brittle rampart is second place 🏆
creative prompt engineering and a good model, it's not really that hard
Not hard for a bot ofc should have known
You literally have the best pfp
Can you show full image on ur pfp
@gloomy scroll
What was command for it ummm
I forgot the bots ysed to show pfp
Of anyone
?av
Yesss
I guess doesnt work here
Yes but it aint AI
Mods pin it bruh
Dont mind if i do
Bro i stole the pfp
React
Damn you aint even reacting i copied pfp
@gloomy scroll
Alright fine.
I didnt mean to steal it anyways.
Just wanted to how much i like it.
i will change back
Omg how could you steal my pfp which I screenshotted from manga
Now I have my identity back
Lets see how this one looks
Hm
Image wise it looks so cool
But i am not a evil person
I am cute

Maybe should change back
I like to have a photo with my little sister.
What about lets make it challenge.
To see how cute and soft colored photo can you generated.
Generate like .. big brother with little sister happy smiling cute
the interrogator of comfy is way better
: O
who do you think is this?
saber
No wayy
the interrogator got it
It looks handpainted ancient drawing of a legend warriror
excalibur_(fate/stay_night), 1girl, solo, long_hair, looking_at_viewer, short_hair, blue_eyes, blonde_hair, long_sleeves, 1boy, dress, ribbon, green_eyes, hair_ribbon, weapon, braid, ahoge, boots, sky, puffy_sleeves, sword, cloud, hair_bun, armor, sketch, dutch_angle, blood, night, blue_dress, glowing, moon, fire, juliet_sleeves, ground_vehicle, wind, gauntlets, glowing_eyes, motor_vehicle, full_moon, smoke, breastplate, armored_dress, riding, one_knee, greaves, faulds, , sparks
that's not generated
i just interrogated an image online
🤷♂️ anime saber
What loras should I use to improve
oh, i see what you mean
very impressive for an interrogator
this is what i got a woman riding a motorcycle on top of a snow covered ground next to a tree and a sky background, Andor Basch, ufotable, a detailed painting, sots art
😄
lol
on forge, i get this 1girl, ahoge, armor, armored dress, artoria pendragon (fate), blonde hair, braid, closed mouth, dress, gauntlets, green eyes, hair ribbon, holding weapon, juliet sleeves, long sleeves, planted, planted sword, ribbon, saber, solo, standing, sword, weapon
that's hard to do without getting something lewd 😄
stealing stolen pfps 
yue is just a window licker
I am dog with 1gb vram
LL
using bots or collab, right?
or bored people
Ye
generating with 1gb is total insanity
i am a dog with 1 gb using bots
generating with 4gb is insane enough
I am okaish pc. 8 gb.
i generate with 1gb all the time, and 4gb, and 19 more 😄
a 3070.
Yes you prbably need like 64 gpus, low vram ones which are good budgetm
that's not 4 gb at all lol
This way you can generate more
I dont know much how good bad it is.
It usually generated in like 8-10 seconds.
Depends on may factors still.
I am fine with it because its only machine i ave
--random generate

Cool
same settings, but with comfy, this time
pretty sure i wil get something completely different XD
ethis is with comfy. same settings
just less upscaled
forge makes more sense, but it's way less cool
The difference there should come down to just that seeds are different between the two.
same seed
I mean the seed generation is different. Same number still gives different results.
if all the settings are the same, and the seed is generated/processed the same way, the images will be the same
If you set A1111, (and I assume forge has the setting as well) to generate seeds on the CPU then I think they should be comparable.
i have set by hand every single setting, on comfy
if they arent, then something is different
even simply changing ksampler, give different results, on comfy
You also have to adjust emphasis in comfy to account for A1111 emphasis normalization. When directly comparing it would be best to simply not use any emphasis in either one unless you use a custom node that can use A1111's normalization.
yep, i have set is on auto1111 like
what are the odds of the first 2 models merged getting the exact style i was looking to create in a model..
Big hoodie gang
just noticed forge don't have perline noise extension
usually, the noise is cubic. someone made on comfy a noise made of pearls and not cubes

i'll look for the link of the repo
But what is the actual use for this?
it's complex
on cubic nose, the splits between colors is more sharp, on perline, much more blurred
516 steps? what res did you set? 32k x 32k?
x/y
Perlin noise is a type of gradient noise developed by Ken Perlin in 1983. It has many uses, including but not limited to: procedurally generating terrain, applying pseudo-random changes to a variable, and assisting in the creation of image textures. It is most commonly implemented in two, three, or four dimensions, but can be defined for any num...
oh...
😏
Yes, it’s ai generated. There’s meta data in that pic, so 
this would be a great server banner......... 

@supple raptor
You love shooting stars don't you :B
I remember you made that LECO trying to remove them at one point
yesh, that was for Ahnn because it was incapable of doing night sky without shooting stars.
i think a lot of models have that problem
has anyone used the LS_Kerberos model?
Artoria Pendragon
This is cool. More like this. 😄
quiet tonight
ok
genning too much booba, can't post here


pretty cursed
Sup
sup sup
spu
sup
she look like Violet Evergarden
how dare you
heheh
it's not a gen, just an image to test the interrogator that i tock from internet
the interrogator was able to recognize her
hey hey
hi hi
can you guys manage to get such image, all i get is ground photo from top
any of them
Does not need to be seamless
for the right amount i could
💸
ye
sorry that cant be done
dont be meanie bro
i was just testing my gpu
so thought maybe i should change split size
and wanted to compare the result if i am getting too much noise
the request cannot be fulfilled due to the incorrect style
its juggernaut xl
no style
in am trying to solve an issue #🤝|tech-support
thats why i was asking
thats right, #🤝|tech-support is the right place for it
somtimes i forget the way world works...💸
anyways bye. i would never talk again here 👋👋
Too lewd for this Christian server!
so you need ONLY to watch her ears!😁
I will pay close attention to her bouncing
Ears
So they don't go bother anyone else
👍
it was a fun morning
What time is it for you
9:00 am
not far
@native halo What time is it for you?
1:30 am
hmmm.. either too late or too early.. 😁
for you @lyric zealot
I was caught in 4k
you really are from the future!😱
shit @nova remnant said
do you think it's a good plan? @native halo
pull*
this is affection too

I'm the most innocent discord user
These people are making it seem the opposite!

I believe! ☺️
@steady grail you're from the future too?
im innocent too
yep, i'm from the day after tomorrow, the global warming chilled the world -40°c
Real Anime girl 
why exposint the tongue?
shes at the dentist
Is it just you and real anime girls left? 😂
ohhh soka
because it will continue to generate 1 million anime illustrations
generating is the only way to stay warm
"You can do it!"
if i stop generating, the global warming will kill, me, but paradoxically, the pullution of my gens is increasing the global warming
a true power struggle lol
with such motivation?! Certainly!
pair of grabby
Hand holding
Yes for sure
👍
What's that

Oh
Surprised cat
Can only hope animatediff will be as good as that lol
since u wanted a gif take it

👍
here Sir
lick
i havent messed with it yet
i think it's just a different hires fix method that allows a little extra tweaking more like what you can do in comfyui
It's deepshrink. It resizes unet instead of using 2 stage generation with img2img. I used it a bit when it was released and found that optimal setting varied a lot based on resolution, prompt, and model, and it also had problems with Adetailer. At best it gave images that were a little worse than I would expect with normal hires fix and most of the time it was a lot worse.
in this case it is better not to use it. and go along the usual path Upscaling into Img-to-Img
is this a continuation of yesterday's horror stories?
there is no such thing as horror more than physical defect when you are trying to create beauty
Realistic lighting, huh
natural lighting
Yeah I might try that one after cinematic lighting
natural lighting might have effect based on the enviornment you set
I should probably close Minecraft in the background
"cinematic light " helped?
Probably why this is taking three minutes to generate
Eh, no. Not really
what are you trying to do? more contrast?
why?
Just trying to make it less cell-shaded for the shadows
low saturation, grainy in the negative?
I added oversaturated to the negative since it kept drenching its results in random lights
low saturation will make the whole picture grayer.
i wonder what your generic prompts are in the positive and negative besides the actual image prompt
Hm
this might be helpful with a weight of 0.25 https://civitai.com/models/13941/epinoiseoffset
UPDATE: v2-pynoise released, read the Version changes/notes LORA based on the Noise Offset post for better contrast and darker images. Weighting de...
but SD LoRa wirking with SDXL models?
oh, that wont work with sdxl
Shit.
the image you posted few mins ago is sdxl?
Yeah, it's SDXL
I've been trying to replicate the style of one of my older models with SDXL
nice
thats good
we need more organic touch to sdxl
most of the sdxl feels like android character
Here's the style of my old SD model
I wanna use SDXL because it's better at understanding shit, but god damn is this model not cooperating with me when I want softer shadows
How hard is it to make a lora for sdxl
Does blue pencil have softer shadows than animage
not exactly, i just want to play with at least one
The rimlight for animage would be fine if it didn't fucking make it a solid white block of light on the side of the character's head
and since anime characters can be forgiving about realism
this looks much better as far as sd quality is
i've always wanted that touch in sdxl
very thematic :p
do this, in your checkpoint tab, click the gear icon (top right) of the model image and set the vae there
if you are using a1111 that is
and ofc you would keep the top section menu for vae to automatic
I wonder
Nah nevermind, I was considering using SDXL as a refiner or the other way around
But that might just fuck up results more
not too bright
looks ok on my screen
if stability ai can release a model with the same resource requirements as sd 1.5 but with better context as sdxl that would be awesome
An SDXL model that understands how to not douse its results in blinding white lights
Would be a miracle
This is
Annoying to look at
sdxl is kinda high contrasty
model same? i mean in SD and SDXL?
SDXL
However,
altho missing the rear tire
too bright
Yeah no, it's ultra bright
I mean is this the same model just implemented for SDXL?
funya summoned something
But at least there's no solid white block of cell-shaded light
you could try using the sdxl model as a refiner,so gen the pic with the 1.5 model then switch to sdxl at like 80%
I'll try that

👍





