#🍥|anime
1 messages · Page 68 of 1
MJ was nice till i learned about running it locally then it became hard justifying paying them 40$ for it a month
possessed!
it being SD
yes...MJ is crazy expensive
atfirst when i had no idea about being able to run this locally it blown my mind, thikning it was cheap
but now knowing..
aye aye
Yeee xD
SD is where its at now 
some experimentation
Im turning an old midjourney gen into an SD gen, trying to add some different flavor to it. I should turn up denoise
hai ho smol
here's a smile for you @lyric zealot
img1 img2img SD, img2 MJ from September 2022
trying to get it colored but retaining details
Bunni femboi
Yould use controlnet :o
I am
attempting another one
Why
you are just used to aurora
also yeah aurora is usually more coherent
and like its style is a tiny bit more diverse (or at least it feels like it) but hell5-dpep follows prompt better and doesn't force some stuff aurora forces or at least less
like lolicon and certain hair colors
Fair, I still like the versatility of Aurora's style more tho
yeah
I also would say it's what it's purpose really is, isn't it?
Yeah I think so
granted it got a bit hazy on some details after 4k upscale
might be the upscaler, or the original 1k
as the successor of kenshi
maybe some inpaint would do wonders, but now imma watch Scream VI
good night bois and gurls
good hands, without controlnet
The hands belong to a woman dressed in a hotdog costume.
She's outside the new local hotdog shop, trying to sell hotdogs.
Bonus points if she is spinning a promo sign.
ok, trying now
The problem I have is if my model even knows what a hotdog costume looks like
@keen hollow am I allowed to use photoshop as an inbetween medium before img2img, or strictly SD without control net?
I can get everything down except the damn costume
like don't you just want to visit WeeSlap and have a Wiener slapping your face? She's in a mascot costume but...not a hot dog one
maybe neg prompt is the culprit
this got creepy, who the F is that girl behind? @keen hollow - this is as close as I could get 😛
weeslap
yes
Buget dineer.
They need to get a new sign painter.
Yeah they do
This is a merge between Meina's Pastel V3, Fikosofter's Lyriel V1.3 and Pastel-Mix by Andite. V2 adds majicMIX fantasy by 4044 in there too with a ...
Aurora
love it!
Hey guys, u can try this :3 https://civitai.com/models/73523/pikas-guilty-gear-style-lora
Easy to use LORA. Captures the style of the Guilty Gear game. Recommendations: Prompt: lora:GuiltyS:0.8, guiltys, by Akihiko Yoshida Enjoy!...
Looks sick, I'll try it :o
I'm pretty happy to see you tried. I love seeing people do different stuff! And I would definitely order the Siapip at the WeeSlap Sweet Diner! 🙂
is there any extension to automatically get the trigger words for loras? every time i want to apply them i need to open civitai and go to the lora to see the trigger words
That saves trigger words
oooh that sounds so good
gonna try that, thank you
Yeee, its pretty nice, and you welcome 
Also, LoRA is nice :O
cow level
boob level up, mooooooooooooooooooo!
Fair XD
Ooo das really cool :O
You are running xformers 0.0.16rc425. The program is tested to work with xformers 0.0.17.
Do I do it?
yeah you can upgrade
pip install clean-fid triton torch==2.0.0+cu118 torchvision torchaudio xformers --force-reinstall --extra-index-url https://download.pytorch.org/whl/cu118
im gettin back into model mixin
Fair fair 
husbando lora + list of feminine things in the negative
Abby the last of us
ty :]
hello I have a problem with creating images in stable diffusiuon
whats ur problem
Gm chat

anyone else planning on buying a 4060 ti 16gb when it comes out?
its all on price XD, but your better off saving and getting like 3090/4090 with 24 gigs if your serious about AI
Me, having nothing but a laptop: 🥹
its gonna be sold out so fast
How an image is generated with the IA?
@violet saffron
Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware - check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝|tech-support for any problems you encounter while installing it locally!
this is rlly cool, do you remember the prompt for this?
The style I achieved with this LoRA
Ooh I see
very nice

Follow the instructions it gives you and you'll fix it. I've had that happen before
I will see
@slate night soo still
Thanks glad you like it. 🙂
why black image?
That I'm not sure 🤔 maybe double check to make sure it was typed correctly?
maybe its your model?
have you tried another model
just to see if thats the problem
im not that serious but its only $500
i do a lot more than just ai
hmm im not sure then, maybe try restarting the web ui close the browser and the cmd window
also did you try changing the floatbit32 setting is said
bc you disabnled nans check so it skips it and shows you what happened anyways
soo what to do ?
yes
hmm im not sure then sorry, maybe someone else knows
okay thanks
Wide
yeup
It says 404 doesn’t exist, do u have the file by chance
Well...I guess she could pull off a nice OG Dragon Ball Bulma cosplay if I ControlNet this a bit.
Nun mommi supremacy
hey guess which anime this locon was trained on
Bocchi? xD
yep
Makes sense 
zombie eh
Just pale :O
And hot
Extremely hot
Mindbogglingly erotic
Lifechangingly luscious
XD
supercalifragilisticexpialidocious

@stable nimbus8
we all know whats under that dress... its a trap...
If there exists a god out there where what he wills is what is right/moral. Is it moral because he wills it or does he will it because it's moral? What do you think?
i use ghostmix
prompt: ((masterpiece, best quality)), the girl with red hair and eyepatch, standing at the edge of the ocean dock, looking over the horizon, dusk, warm colour scheme, makato shinkai art, anime screenshot, highly detailed, vivid, beautiful lightings

gm
Ah I didn’t notice that I thought she was an owl butterfly or fairy lol
my fault for not putting much emphasis on moth features 
I mean it's just the prompt bleed putting wings on everything made me think fairy haha
current state of ai images:
Keqing?
Noelle Silva
the hair is a lot closer to keqing
face is noelle tho
neither the face or the hair is the classic but it does feel more remniscient
long nose does seem more noelle tho
ghostmix, the rooftops of a cyberpunk city, bladerunner 2049, complex detailed style manga volume cover by sui ishida, sui ishida complex tokyo ghoul manga cover art, horror, flat shading, complex colourful pallete, complex shading, colourful
Damn that is so cool, lemme try that :o
Time to say goodbye to the nun mommy posting 
For now 
First output, Imma try using my Asteroid Ill LoRA tho :o

That one feels like a anime moment of like the MC and his best friend like "Are you sure you want to do this?" or something like that xD
Spooki place
cool prompt 🙂
Omg moth girl
my dm's are open 😄

you using hyperbreast right?
Nope
https://civitai.com/models/73240 This LoRA
I love the style, but also I wonder what even is in the dataset that allows boobs to get that humongous
anyway do you want to try my model?
Also I'm pretty sure it was trained with offset noise
Maybe :o
If its too big I'll be lazy to download it
xD
/dream: Ford bronco, old, retro, custom, chrome wheels, red, white, nu-girls
wrong channel
@summer sun
Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware - check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝|tech-support for any problems you encounter while installing it locally!
Sorry, thnx
reminds me of my first sd images lol
Yeah, and if people adopt it, these will be kind of everyone's first images again, as this is SDXL. Back to the basics in a weird way. It does generate much superior anime stuff, though.
im happy i kept my old ones
this type of typical old SD eye style is still there in SDXL, but with much better quality now
prompt is "yes", animation of iirc 150 steps on euler a (this is from november 2nd)
Any tutorial to train a lora of a character?
was this trained on something in particular? All of these characters have massive head to face ratios
It's SDXL, the training is still going.
emphasis on XL har har har
https://media.discordapp.net/attachments/1108000871729614939/1110573032164048968/00042-4172079284.png?width=1159&height=657
it only went through 210 iterations of adaptive+36 hi rez, i tried difrent samplers but for some reason whenever i try to do somehting only adaptive looks good...
pretty! I like how the clouds and sky blend together
it looks like the sky interrupts the moon which is super cool
model?
Cetus-Mix v4 with some lora/locon
nice!! what prompts did you use? :)
But in exif wrote breakdomainanime model
breakdomainanime is awesome :>
im about to try it
I'm guessing that's the one you used just now
A professional dark anime upper-body artwork portrait wallpaper of 1 tomboy girl, (1 girl only:2, very handsome tomboy girl:2), (in a formal male imperial marshal military uniform:2, long military pants:2, wearing a military cap:1.8, boob bag:1.6, luxurious royal earring:1.8), ((big breasts)), (in a luxurious royal palace:2), masterpiece, (best quality:1.4), highly detailed, vivid colours, (black hair:1.8, very boyish short hairstyle:2), ((perfect anime face)), (very handsome girl face:2), ((perfect aqua anime eyes)), ((perfect anime mouth)), (closed mouth), [perfect teeth], pale skin texture, ((perfect mature female body)), (oiled shiny body:1.9), (looking at the camera, front facing the camera), ((low light)), (((midnight))), ((afternoon)), (excited expression:2, light handsome smile:1.6),
I believe it is this one
it makes awesome results like that
i love it
one of the top 3 of my models i use
Tart card
im having fun using pikas model but that looks great too
u should try blue pencil is my favorite so far!
thanks!! I’m gonna try this
i will try it for sure. in civitai?
man imagine this: 10 years into the future a nerve gear exists and LLM NPCs are good(not with that TTS shit, i mean where you can really interact like in a real conversation), couple that shit and boom! all my dreams come true.. when do ya guys think the tech gonna be usable(i just set 10 years for example, something realistic would maybe be 15-20 years or more(i think so at least))?
i am no neuroscientist so idk man
i mean real nerve gear full on for the public
no "only for the big guys" or some shit
Ermm maybe 10 then
Too expensive
thats what i mean
for the public, everyone could buy it
when do ya think
?
It just depends how long they’ll gatekeep it first
My guess is that anything resembling full dive VR is probably still a long way off from being in a state that enough people can afford it for anyone to put any real effort into developing things for it. As far as having LLM controlled NPCS, I think that will be easy once mid range GPUs get more VRAM.
They've already paused gpt stuff so
i think, for now you think you want it. But till this what you imagine now comes out, there will be something as interesting as that
which does not need alot of time
who
Elon
elon didnt pause gpt lol
Not just him yeah
ye this is another thing
Elon left OpenAI in 2018. He signed some petition thing requesting that AI research be paused, while simultaneously purchasing several thousand GPUs for his own AI startup.
bruh i heard about that
dumb as hell
Oh, it's smart. He's trying to get himself an opportunity to play catch up
not dumb he's who he is a silly person who want to control peoples mind
with his twitter he can
now:D
its still dumb
Businessman at heart
idk what to call it man, theres a german word for it
There's a german word for everything
selbstironie i think
scheiß Schnitzel
self-irony lol

alter kannste mal mein armes schnitzel in ruhe lassen?
die sind lecker
man alter
now i'm hungry
Dedication
how long did that take?
uh , around 5 hours but i was watching netflix for a bit of it xD
lotta impainting huh?
you gonna upload that model?
initial gen -> inpaint -> import into clipstudio , manual edits -> inpaint to smooth low effort edits -> repeat -> upscale -> repeat -> create more gens for the billboards & gun -> post processing editing and fine detailing (the UI, rain effect, text, wet detailing, eye reflections) -> done
I was gonna say nice gun gen 
I was gonna say, damn, thats extremely good for an initial gen.
me too xd
there it is, even before inpainting oof lol
ook
someone wanna have a dataset? idk what to do with it
sample image
these were NOT generated with SD
now i need to make a lucy
hey, can someone show me how tf you use impainting? (never used it)
i gotchu ❤️
start with these settings
thx, idk why my webui isnt starting
yeah mine had some trouble last night >.>
the only masked means only the area you mask will be included in the generation phase, so as long as you don't inpaint large enough to cover a whole side then it will "zoom in" while generating and increase level of detail
the masked padding artificially zoomes out the generation, this is good for inpaininting hands when you dont want to inpaint the forearm as well
oh thx man

My head hurted
the denoise strength of 0.1-0.3 will increase detail with minor changes, higher will start to do major changes to whats masked
depending on how lazy you are you can prompt for what you are masking only, which gives more denoise freedom, but for me i usually don't so i have to keep it lower
to make inpait do what you want you realy have alot of exprience with it...
and it sometimes still wont do what you want
there is no on button option for it
I usually just frontload my existing prompt with what I'm inpainting. If I'm inpainting a hand, for example, I would add hand, clenched hand in front of the rest of my prompt. obviously either don't include or change "clenched hand" to match the position you want the hand to be in. This is usually enough to let me get away with up to 0.4 denoising.
That's where editing comes in 
most important is the prompt!
This whole Picture, i changed just by Inpainting... this was a pain :x
Developing the skill to roughly paint in at least the shape of what you want in PS, Krita, GIMP, or whatever else you use will help a tremendous amount with inpainting too so that you can use that as the base for your inpainting. These days, I do basically all of my inpainting using the Krita extension so I can do manual edits and inpaint all in the same interface.
Accurately masking is also a lot easier in Krita.
and is it compatible with vlads fork?
Gn chat gonna sleep
I haven't tried, but probably as long as he didn't change the API since, that's what it uses https://github.com/Interpause/auto-sd-paint-ext.git
idk what i did, but i did it right(pic1=before,pic2=after)
not perfect but last time i tried it was like..
shit
really shit
In the Krita extension, to achieve the same as Eppi showed with the "Only masked" option, you would simply use the rectangle selection tool to select only the area around your inpaint mask.
You can also use this method to create your own inpainting models out of any model which often work better than normal models when you want more than a subtle change. https://www.reddit.com/r/StableDiffusion/comments/zyi24j/how_to_turn_any_model_into_an_inpainting_model/
final
played around with autophotoshop (sd plugin for photoshop) - image quality isn't all that great but the tool is getting better
the background elf and small dragon are inpainted, so it half the forest
also layered
more elves
quadruplets
work in progress, what do ya guys think?
Sexy! I love it
Dynamic pose + subtle expression = 
Dang, I'd been getting mostly nice hands all morning too lmao
what model?
It's a mix I made of like... 7-8 different models at this point, but I'm at the end result that I like the best
can ya uplodad it?
I might throw it on civit, haven't committed yet
please do, it looks great!
Tysm!
dm me if you do please
he angry, he hide face
I saw husbando, I;m here!
lol

What models (checkpoints) are you using?
Working on a new model based on my Dark Gemini 3, I call it Hyper Variant
4k version
Maybe a strange question, but why are you making this all? I.e. for what usage
i wanna have it
Just because
https://civitai.com/models/6209/dark-gemini-v3 for now there's this 😛
true artists gather here
vampire!
When excited, these vampires eyes begin to leak blood eventually dyeing their eyes red
Modern solution to a problem
aye
i. cannot. get. enough of this prompt!
did you guys never run into something similar, that you just want to press generate all the time on the same prompt because it gives such pleasing pictures?
now i just need to figure out how to get this style transfered onto a selfie
img2img control net reference
and maybe canny or something as a secondary with lower weight
then your selfie as the img
potentially also a third control net with line art anime
yeah, i tried them all a bit already, the bad part is they can't seem to stick to the metal face parts
or the unskinned nose
this prompt srsly would srlsly make the sickest selfies
model/lora?
SDXL, nothing else
damn, sdxl becoming crezy 😮
fav prompt + "chicken, bird"
done
and you don't even have to punish ur gpu for it 😄
gotta love sudden perfect hands for no reason
Yep,
but I thought this time the bot was gone for good.
The knight chicken + bird one? brb
can any anime models create something like this one ?
trying to get a model that can do something like this
I want to try chicken knight now lmao
chicken rampage studio ghibli style? 🙂
mecha to the rescue
((anime style)) a chicken, blurred bokeh background farm
what model is this please ?
was searching for something like this one 🙂
Might be the ghibli lora
Thank you! The trick is the prompt, less the model:
((anime studio ghibli flat 2d cel shaded)) a giant chicken rampaging in a modern city
No LoRA, no embedding needed
Stealing that when I'm done with chicken knight lol
will try to do something like this one with different models and see 🙂
It's like Goofy and Pluto, you've got your self-aware anthro chicken and his pet regular animal chicken
Some more chicken knight
Zombie Edition
chicken everywhere...
domo arigato, chicken roboto
An extra straw for sharing, how sweet 
LLMs are cheat codes
Oh really? You started prompting on your own now? 😛
no, i just copy pasted Path of Exile divinition card quotes 😄
me, making my own prompts? RIDICULOUS!
lol
oh. i left artist references in.
that's cheatcodes too
or are they?
my personal favorite blend -> yoji shinkawa+todd mcfarlane+kentaro miura
she's saying something muted?
interesting nuke. everthing is still dead tho
nice prompt ❤️
not sure if anime. but tokyo got nuked so many times -> nuked cities = anime
not a nuke, more like a dimensional craft arriving in our dimension, devouring all energy and living particles to recharge itselves. Unfortunately resulting in a Tree of Life rooting it in our dimension, sealing its destructive powers
so you mean, still destroyed city -> nuked city -> anime
I remember when nuked cities meant hollywood alien invasion
dude, that's so 1995 XD
out of curiosity, do you usually just use some set of words to invoke your character or did you train something extra for it?
we were so close, 1996
Interesting hair
not sen this combo before, cool ^^
For some reason AI sometimes think White = Black
Im sure we can find something from 1995
I mean we're old, practically grandparents in this channel
but you have to agree, city wide destruction was independance day
strange as it was, i don't think i saw t2
because i even thought at the time t1 was boring too
one of the best movies of all time that still holds up today
it is to T1 what Aliens was to Alien
nonstop action
'this is so cliche' -> well, couldn't formulate it as a kid, but that's what i thought
Any negative for skin tight clothes?
skindentation
catsuit
I'm already using leotard in neg
Started to work somewhat I'll jsut try increasing weight
damn she has skills, back wheel floating in the air going up hill
hi
what are the best settings for wizard lm 7b for roleplay?
don't question my intentions
anyone got prompts for nice film grain? or maybe can refer to a lora for me, would be nice
try filmic, caustic
I missed Terminator chat?
I've actually been rocking a Terminator marathon over the past few days.
I mean, only the first two count as masterpieces.
But I find them all fun in their own cooky ways.
Both 1 and 2 are favorite movies of mine in general, and T2 is probably my second favorite movie of all time.
James Cameron knew his stuff.
that said, I've not been able to get decent "terminatory" images generated 😛
I just make hot mommies 
neat
They good 
i make cute girls
Both is good :O
HOW DO YOU MAKE THIS AI ANIME THINGY☠️
@true latch
Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware - check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝|tech-support for any problems you encounter while installing it locally!
@wispy canopy I need help fr. how can I make myself go sleep earlier
idk anymore
I just get too invested in stuff
and it's like I need someone from outside to tell me to stop for me to do so
but I don't have that all the time and I wouldn't want to rely on other people
Go sleep, nowe 
That good?
xD
even if I sleep now I will have very little time
I feel really happy with this
quite some artifacts though
Well, some sleep is better than no sleep :o
what model did you use?
at least now I have a fully customized firefox browser (just switched today)
literally published a theme
changed a ton of settings
ez
Oh I didn't know it was customizable :o
but at what cost
one of the most
it's literally open source
Oh damn, thats fair then :o
no?
Novelai with a custom model I made, was just a test and very poorly trained, hence why there are artifacts, small amount of images and some of them were blurry once resized, as well as a lot of them being improperly tagged
novelai is very very old now
True
for it's time it was the best
“Explore the New Horizons.... Don't let indecision limit your potential, The world is full of possibilities.” Aurora is a Stable Diffusion model, s...

and that's why most anime models are based on it as a base
but now we have just better models
less artifacts
better quality
I see
it would work better for me as a base to make my own models off of?
there are quite a few models I can suggest
but I would just go to projectai and look what people use there
The one I sent is currently one of the best 
I'm open to all ideas
Probably 
I got the space, maybe not so much the vram, but I will see what I can do
they have this thing where you can react with an emoji to images and it tells you the data of the image (what model was used and most settings)
Aurora is 2 GBs, so its good :o
I'll send you an invite to projectai
kk
I personally haven't downloaded any models except default and novelai
I forgot NAI even existed 😭
Same XD
models unless different architecture (if their base is sd 1.x or 2.x for example) or different pruning/precision (this matters much much less), have the same usage of vram
I'm a 6 gb vram kinda girl
on the right settings haha
but first I pet dog
ofc
but you can do pretty well
are you using xformers?
I mean, I used Aurora with 4 GB of VRAM, so you'll be good xD
no
I actually restarted with a fresh install today
mid
might do some boy posting later
oh wow what is this amazing amazing theme I found for firefox?? https://addons.mozilla.org/en-US/firefox/addon/my-theme-totally-not-stolen/
so good wow
time to pet my dog then
GN
holy pizza
My hand not fat it's just camera I swear
Blind
O, cute :0
Nice. Is that Emilia?
Noelle from Black Clover
Thanks.
Gm chat
Like closed the cmd?
yes
Hmm try and delete the control net from your files
In tmp
okay I will re-install
Good luck
I still get this error
well that's different
ham buh buh 🍔
still I don't see it
wait what version is your stable diffusion?
for some reason yours doesnt show on the bottom
msg in tech support
how to know ?
quiet tonight
beautiful pic
POLCE
Ok thatll be a $20 subscription to my patreon for support tips
so cute
thankyou for making it
i cant get the dog T-T
at last dog
the prompts
hyper realistic lifelike texture dramatic lighting unrealengine trending on artstation,award winning photo,nikon RAW photo,8 k,Fujifilm XT3,masterpiece, best quality, realistic, photorealistic,ultra detailed,extremely detailed face,1boy, animal, autumn, bangs, brown_hair, closed_eyes, daisy, dandelion, day, (dog:1.2), field, flower, flower_field, ginkgo_leaf, grass, holding, holding_flower, hood, hood_down, hooded_jacket, hoodie, hug, jacket, leaf, long_sleeves, male_focus, on_grass, on_ground, orange_flower, orange_sky, outdoors, plant, shirt, smile, sunflower, sunflower_hair_ornament, sunset, tulip, wheat, white_shirt, yellow_flower, yellow_rose, yellow_sky, yellow_theme
cute doggo
gm
i was already asleep at the time 😄
Yeah I guessed
i assume you haven't slept yet again?
hello could I get the model ?
hi, all of my recent generation is using ExpMix_LineV2 model
Oh
Thanks
absurdres, epic fantasy anime concept art, Gordon Ramsay embarking on a grand adventure to locate the lost sauce of the lamb
did u use Lowra ?
what's a Lowra
low light lora it give u dark theme
@grave snow try this
same image with dark removed from prompt and LowRA on 0.6 strength as they suggest
so nice
Could someone attempt to create an anime figure based on this character?



