#🍥|anime
1 messages · Page 90 of 1
she's suppose to be punching a donkey but hey ran away it seems
She-HULK
hmm not coming out the way I want.. interesting

sheeesh , ma man thank you
I didn't mean to do NSFW
welcome
I tired, sorta worked
treid
sigh... tried even
princess Peach x Chun li
hmm thats a good one let me try 😄
grins evilly
Suppose to be Queen Mario 😄
ok thats.. different
aww why?
I will send my best warrior(s) to defeat the clowns!
sees more clowns
tries a special move
ooh that went sideways
tries something different
oops
changes script to include more real cats but not actually have full on cat people
well umm...
looks like a potential saints row V ad
attacks
defense mode
oh crap
I'll get it
uh oh, cat must have been messing with the sword
thats better
initiates final form
uWu
uses ultra instinct move
there chibi clown girl
ok fading.. gone.. to.. thud
weird neeko
yoo what's the model/loras on this one?
the average #🍥|anime contributor
ok it is now finally time to get back into stable diffusion
time to download a1111 again
WB
haven't generated an image for like 1-2 months till today
only used it a bit for testing my lora anyways
Prompt pls? 
@slate night I used (((watercolor anime))) (((handmade))), (((hand-drawn illustration))) and a style I have save
saved
also (((white background)))
Tysm!!!!
which model did you use?
It's my own mix, I forget the exact recipe but I know it has some DarkMeinaSushiMix in there if that helps
what’s the best way to upscale images from 512 up to 2k/4k?
by upscale do you mean blow it up to a bigger size, or add more detail?
My process would be to use Hires Fix at 2x upscaling to go from 512x512 to 1024x1024, then I would send the image to the img2img tab and use tiled diffusion upscaling to upscale that result by 2x (for 2048x2048) or 4x (for 4096x4096) with around 0.38 denoising and with the tiled controlnet active. You will may need to use tiled VAE for both the Hires fix and tiled diffusion steps depending on how much vram you have available.
For the tiled diffusion upscaling you will likely also want to change your prompt so that it is basically only quality tags. Remove any descriptions of the character at least.
I mean, both
Thank you! Going to try this for the next one
If you don't have the tiled diffusion extension yet https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111
If you have a bad card it might take a long time to experiment with settings, also, you might have to inpaint a little still afterwards.
I would always do inpainting before doing the tiled diffusion upscale. Less vram requirement to upscale the lower res version, and it allows multidiffusion to help blend any inpainting seams that might be visible after inpainting.
If there is a extension like adetailer, but tiled and applies to the whole image, I would be so happy, my workflow would be cut by like half.
Sounds like this would work well for you. It doesn't happen autmatically, but with GroundingDINO it lets you prompt what you want masked and it does that for you. I only tried installing once, and didn't have any luck, but from what I've heard, it works very well once you get it working. https://github.com/continue-revolution/sd-webui-segment-anything
Oh my, it's like reading a physics research paper, but I'll try my best to understand it.
Yeah, it's definitely pretty complicated.
Is there an improved way to outpaint beyond poor man's outpainting and outpaint mk2? My friend asked to outpaint a picture they liked to use as a wallpaper. That would require two round of outpainting to get to 1920x1088 (and then I would crop the 8 pixels). The trouble is that the second outpaint result is pretty underwhelming
After many pictures, I got one I think it's a good left side, but the right side is always bad, and I think I hit some limit on what I'm using to outpaint
yep #🍥|anime message
Kokkoro hit the gym
Took a while, but finally generated a picture, and the result is way better than before. Thank you
No problem, glad I could help. This process is pretty similar from what I've heard to how Adobe is doing their generative fill, especially if you use Hires Fix along with it. It gives the AI much better context for the rest of the image which allows the outpainted result to match the original image much better.
o.o wha?
andshe fell thru a portal
well.. thats diff
uhh oh the clowns are back
oh my
"hello, wanna play?"
hmm i like that but let's do it with mistoonanime model
hmm seems she's got a space suit malfuction
goes to bed... thud
race.
race!
Oh she's racing
Better watch out
Lol it looks like a straight screenshot from a manga
You know that feeling when you looking at art before AI Art existed and you see an interesting character and hope they're from an anime or something - nope just a one off OC drawing and you're like "Damn"
AI art giving me that feeling sometimes lol
what type of lora do you guys use to add details?
I just use highly detailed or intricate details in prompt
But... there is just an add detail lora
HF: https://huggingface.co/OedoSoldier/detail-tweaker-lora This is a LoRA for enhancing/diminishing detail while keeping the overall style/characte...
pumpkin girl
what is your model?
Just check the image metadata (If it's still there)
yeah, the metadata is there, you can check it
just mad this lol
lol just did another one
Ever get bored and just make chibis of yourself? 
how?
it has it's moments
hmm a twofer?
1 more then i'm going to bed. so tired
sleep walking
awww
is it adetailer that's good for adding things like facial features (hair, eyes etc) and so on to be consistent?
Don't think so
where can i see a good inpainting tutorial? i have seen only one but i still dont know what the settings do or how to change things in the image.
I have no idea man, it seems to differ from which inpainter you use
i am vevy very new to inpainting stuff so i just use the one in the img2img tab (i use A1111)
okay I don't know how to use that one
what is the one you use?
I tend to either use the one in InvokeAI (another webui) or when I get it to work, AutoPhotoshop
sometimes Krita's SD plugin'
I want to learn how to properly use adetailer
What are you unsure about? I've been using it long enough that I don't really remember what's confusing for a new user.
Basically turn it on and forget that it exists, then cry a little when you gen something with a bunch of faces and it has to inpaint all of them.
so all it does is create a mask?
It automatically detects faces, hands or full bodies depending on which model you select in the extension. Then it inpaints the entire rectangle that the target was detected in.
honestly almost everything, idk if i have to download an specific model and theres a lot in the interface that i dont know like what settings should i put to say, fix a hand or a face and what to use if i want to add something, sorry if i sound dumb im really new to inpainting lol
alrighty step 2 on this face
cmon cmon
when invokeAI for outpainting and inpainting
cause I can't get the extensions for auto to work well
maybe I could use meta-human to make something as addon in img2img, since I've already started with it and it's only a reference point then so I think that's allowed in cow
a face that is'
hmm how do you properly use break?
like separate your prompt into chunks that each have a purpse (theme, layout, design, etc)
and that way give better focus to each?
No need for a specific model, though inpainting models can be helpful at times. You can turn any model into an inpainting model by merging using the Add Difference method with SD1.5-inpainting (A) + *your model * (B) - SD1.5-pruned (C) with the multiplier set to 1. In the custom name field you can name the model whatever you want as long as it ends in .inpainting which tells Auto1111 to load the inpainting config with it.
As for settings, you can leave quite a few of them at default settings.
Mask blur controls how many pixels from the edges of your mask are blurred to allow it to blend into the surrounding area. Default is usually fine for this.
Mask mode controls whether the mask covers the area you want inpainted or the area you don't want inpainted. You usually want this on Inpaint masked.
Masked content controls the type of inpainting you want to do. This should almost always be set to Original.
Inpaint area controls how much of the image is given to the AI to inpaint on. Setting it to Whole picture means that it will be given the full image, while Only masked means it will only be given the area around the mask and it will inpaint that area at the full resolution that you have set on the dimension sliders. Only masked is usually the best option.
Only masked padding, pixels only applies when using the Only masked option above. It controls how much outside of the mask is given to the AI. Usually 24-32 pixels are good for this.
Denoising strength controls basically how much you want to allow the image to be changed. 0 means no change, 1 means almost an entirely new image. for inpainting, you usually want values between about 0.15-0.4 depending on the specific job.
hmmm what to use for preprocessor threshold
There isn't really a correct way to use it. You can try to organize your prompt with it if you want, but that won't always give the best results. The best use is just to try your prompt without any of them and then add them wherever the prompt needs the CLIP embedding to be reset. It can help to separate influence of different colors, ensuring a long word or multiword tag/concept fits completely in 1 block of tokens, or as another tool to help control the relative strength of different tokens.
the what now?
control net
it's called threshold
it has no model called threshold
and thank you on the break info
Oh, you mean the settings that are specific to certain preprocessors. Those are pretty well optimized out of the box. It's rarely beneficial to touch them.
no I mean there is no model for it
there's a preprocessor
but no model
so I don't know what to use it on
Oh... that... Yeah. I can't see a practical use for it on normal images. Maybe very situationally with the scribble model, or to cleanup a canny or lineart preprocessed image that you've modified.
There is no model that actually matches up to it.
hmmmm yeah....
I lost out so much money on hospital bills that I won't be affording that 3090
I still have enough for an alright bicycle though
large to x large frame (depending on manufacturer and heights), 7-9 gears, that shinmaeu or whatever the gear maker is called
btw where did pixel2pixel go in its full functionality? Cause img2img aint working well for that
Her hair reminds me of Sailor Pluto :3
give her some tiny horns and an edge glow to her wings that give off tiny feathers as sparks
yay fever free for 3 weeks now'
hmmm maybe combine in photoshop
then she wouldn't be Aya though
Satania isn't Touhou so she can't be in my emo touhou band.
THERE
goes crazy
ok didn't go as crazy as I thought
might be getting a new pc this week.. maybe
one thats way above what I currently have
this
it's alot bette thant he 3050 I have
I would try the 4070
I suppose that's true
but it's overvalued
way
for the performance it offers
considering the price compared to the one's with 4070 it's what I can get
anyhow bck to art
boom boom
does it have to be a laptop?
I mean if you want portable you could also get a mini-itx in one of those weird fit cases where you do have room for a 3090 or 3080 (second hand)
well I haven't had very good luck with regular desktops living in the country
order parts
with a laptop there's no room for things to get in and mess it up
Image 1:
Little bird: I'm telling my dad on you big bully!
Big Bird: CRAAAW!
Image 2:
Dad: By the might of THE SCEPTER OF THE ELEVEN I SHALL SMITE THEE!
Bully: No! PLEASE! NOOO CRAAAAA-CRRAAKHHH
Image 3:
Little Bird: Hey dad, remember that bully I told you about?
Dad: The one I propellered until his head popped off?
Little Bird: YEAH!
Dad: What about him?
Little Bird: Heh well, I'dbeen bullying the big dumby for weeks...I wanted to see if you would actually do something. It was funny
Dad: Son....
Little Bird: Yeah dad? ...CRAAAKKKKKKKKKKKKH
hmmm? I dont know?
grandpa arrives! He eats dad, he eats his grandson lying there dying on the floor. He digs up the corpse of the not bully and resurrects him. He turns him into a human. That human has now decided to destroy the world, grandpa leaves and says good luck.
well half human
oh going dark now. ok
You must be unfamiliar with me
Hi I'm Cryonicus, I do Gothic and Dark Fantasy 99.9% of the time
I like pitch black stories
after paying all my hospital bills etc I spent the last of my savings on steam sale
I have way to many games as is
games I will actually be playing too
I think I will start with something on game pass though, Ghostwire, Danganronpa v3 Killing Harmony, or Persona 5?
are those branches hooked up to his exhaust=?
no clue as I didn't add that
you could try my dark gemini v3
it's still good I think
but I should probably work on a new one with a better base and reach a point where natural language works but booru still has some function. Then focus on lora's
https://civitai.com/models/6209/dark-gemini-v3 when civitai decides to work
might give it a try, thanks
thats with majicmixhorror
downloadingyours now
ok now lets try it
I'm glad I at least managed to make a real submission and not a lazy one this weeks cow
you like dark gemini 3 or what did you mean?
yes
cool
it's fairly versatile
hmm think it's hte cartoon effect thats causing the wired red outline
Or overprompted
weird
what vae are you using?
trying diff one's
the red outline happens sometimes, I'm not sure why the generations burn. Most of the time they don't
are you using a noise offset?
oh I know what it is, it's the color enhance
ok maybe i'm over doing it on the prompt and styles
dials it back
hmm bettter
sorta
could be sampler.. or other things.. shrugs
also modding cyberpunk 2077 as I'm doing this 😄
To use inpainting model, I place it in the same folder as the other models?
changed to a merge I made and not bad
hmm good question. not sure
yoo thank you so much for this! i will try messing around with it
keep in mind that DG3 responds best to natural language prompts
and often wants a directive on the style
like 2.5d digital illustration, digital illustration, acrylic artwork, dark fantasy art style, gothic art style and so on
in order to give best quality on them
also, it doesn't like it when you use masterpiece, bets quality
what's the prompt?
one sec
Demon 1girl solo, white canvas, dark graphite style, Gothic red-black leather Dress, twin tail hair, standing in a dungeon , grim undertones, detailed shadows, art by makoto shinkai
Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, an6, kkw-extreme-neg
prob not the best
ok the art by probably wont work with mine
it's almost stripped of all that
on the neg prompts, which are embeds?
the last two. which I just added
alright I'm gonna run it first with your unedited prompt, then try editing it to what usually works on dg3
what resolution you use? sampler? steps?
how many clip skips?
@bitter latch
Steps: 40, Sampler: UniPC, CFG scale: 8, Seed: 1690514200, Size: 768x512, Model hash: cca7de46ce, Model: darkGeminiV3_darkGeminiV3, Clip skip: 2, Color Enhance: 0.5, Version: v1.4.0
and vae?
yeah prob not the best over all , I just kinda put stuff in. i'm random
i've tried different one's
that kinda sounds like someone took a vae and renamed it
but I also use some anime and color enhancing one's. just sorta trying them out
where is it for download?
civitai
ahh
color enhance
which I think is part of settings
but yeah i'm all over the place
😄
any other vae?
shrugs
vae-ft-mse-840000-ema-pruned, vae-ft-ema-560000-ema-pruned, kl-f8-anime, kl-f8-anime2, WD1-4-kl-f8-anime2-bless08 ?
sorry don't really wnna look all that up right now
goes back to modding cyberpunk 2077
https://huggingface.co/stabilityai/sd-vae-ft-ema here if you scroll down instead of going to files, you will find a zip with both kl anime and then links to the other two
easy does it
whenever you feel like it I mean
just there you have it
unipc pisses me off, it keeps saying raise NansException(message) modules.devices.NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check. etc
no matter the vae'''
so I guess its made for sd 2+
damn it @bitter latch she keeps walking away from me
could try adding, POV Facing
yeah gonna run my own version
go for it
i'm still downloading mods.. alot of mods for this pack wow
interesting
yeah but I suck at remembering all that language and like to be crateive
IE.. random as fu... boi
sometimes it still gets a few burn lines but I think its cause the model has been merged to its last legs
I'm very creative, but easierfor me to be creative with natural language
nods. I suck at grammer
ok adetailer doing some weird shit
overall I like adetailer though
it makes sure faces don't get fubared
I did forget to tell the prompt solo though
young demon woman with medium long black hair twin tails held up by tiny cute skull hair ties, her eyes are a highly detailed amber color and her skin is soft and white, beauty filter, she wears a nu-gothic black leather dress with red lace frilled edges,twin tail hair, the boss of her own dungeon and proud of it, proud expression, evil smile, seductive eyes, 2.5d digital illustration, digital art, dark fantasy art style, nu-gothic art infusion, dim lighting, hard lighting, backlit, full body view, HDR, 8K <lora:locon-noise-offset:0.85>
Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, neg_hands, EasyNegative
Steps: 30, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 308183130, Size: 1024x680, Model hash: 021972d7ff, Model: Current_WIP_DG3_FP32_Full, Clip skip: 2, ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 23.7.5, CFG Rescale phi: 0, Lora hashes: "locon-noise-offset: 354e87dbb8e0", Discard penultimate sigma: True, Eta: 0.66, Version: v1.4.0
Used embeddings: neg_hands [b740], EasyNegative [119b]
oops I forgot that second twin tail her
I should remove that
merging faces wasn't a good idea on adetailer
sorry for other ping but i still have some small questions, should the prompt be the same as the txt2img or should it be just what i want like "hands" or "face", and my image only starts changing after like 0.6 denoising
demon succubi twin sisters
Why do my images get messed up like this? (I'm using img2img but the oner ones were okay, just not this one). I added cfg fix and also add detail lora, but adjusting them in all variations doesn't fix the problem. When I took them out and only used the model (mistoon anime) then it worked better but less details
Main issue that can be seen is the car; it doesn't have an even color
What does your prompt look like?
It depends, for very low denoising, you may want to just use something like hand. Most of the time you should prompt for the full area that will be visible to the AI.
how am i supposed to us a Lora art style?
- anime, cartoon, ghibli, stylized, toon, light, brown eyes. - (embedding:easynegative), female, girl, woman, feminine. I pretty much use simple prompts without composition, just style prompts (for img2img)
I would add a lot more to the prompt, the more specific details the better- like 1person, leaning against car, red sports car, parking lot, wood building, stuff like that
Assuming you're using A1111, just download the lora you want to use and put it in the stable-diffusion-webui > models > Lora folder
yeah, it took me some hours to figure out but i finally managed to do it, thanks
can i get some more help?
i downloaded this lora but it says i need a specific checkpoint for it to work
but idk where to put the checkpoint
anylora
That one goes in stable-diffusion-webui > models > Stable-diffusion
Think of it like, the "checkpoint" is the model itself, and the "lora" is an add-on that makes the checkpoint produce a very specific thing
tysm
thanks, will try again when i get home
hi is it possible to use high res fix on an image u have already generated
i have put into png info, sent to txt to img with same seed and everything, but still does a different image, is it cuz i generated 2 images in a batch or somethign?
if anyone knoiws thankyou
You can always upscale using img2img.
yea but i generated an image with high res fix and it looked insanely better quality than when i use img2img to upscale
i assume there is no way then
Use png.info tab and use send to txt2img then check box hires upscale, it should make the same image and then also do hires upscale ....as long as you save meta data to png in settings otherwise idk
Otherwise you can img2img upscale then inpaint but that's a bit more work
hmmm...
trying to get sdxl figured out
yeah better not
ahha there we go
last one. and bed... thud
sneaks one in
After lurking in silence for quite a while, here is one of my latest image ^-^
hmm I think I managed to finally make an ok cow submission for the first time in ages
hope so 😮 I wants black name
how damn dark is your dark mode?
its standard dark mode in discord
Unless you want to get your account hijacked like @cloud wolf probably did, don't click the link. 2fa will not protect you when your token gets stolen from clicking the scary link.
@dull jackal we got a "free nitro" spammer
do we?
Back to my dream about cooking hard boiled 🥚 and having a timer shaped like a 🐧
Lol
Thanks for the ping!
Hahaha maybe I should have some stuff delivered lmao
That timer was adorable. Unlike that scammer
Lol, poor guy probably just tried to get some free nitro and instead they lost their account
wait, sabra has breakfast avavado toast
YOU'RE DAMN RIGHT THEY DID (rip)
Dammit, now I want hummus, too
So, how are you, nubby?
Hi this is a very nice picture anime.
Okey i will try
Not working
I don't get that black tone, it makes it whiter.
Like the same shade as her shirt? Or just like normal deep dark skin tone?
Like the same shade as her shirt
Like this
Hm
That might not be something the model you're using is capable of reproducing, or you could try like jet black skin and literally every other skin shade tone in your negative prompt
I am using aom3
Yeah I'm not familiar with that model sorry 😅
how did you get so much detail
it was inpaint bug, i was generating star shaped pupils and got this, i tried changing the image kipping eye's but failed even with controlnet. Tried generating something simmilar with text but also failed.
A friend wanted to have black as obsidian skin, this LORA acts as an helper to that Please offer me a coffee here : https://ko-fi.com/maidragora Or...
Ill make a test grid and see if any of 1k combinations get the skin right
More or less what you wanted the Grid i made will be done later today and ill see if anything darker comes with it
ty!!
yeah but overall the whole image is very well made
well, hires.fix is good, do good promts
what model is this 😮
comfyUI ain't that comfy. but it's something interesting to play with IG
1, sardonyxREDUX_v20/ 2. CheckpointYesmix_v15/ 3.meinamix_meinaV10/ 4. sardonyxREDUX_v20
is this sdxl ? 😮
this is using SDXL UI to run some 1.5 models. It's working out but I feel like A1111 just does the UI better and includes everything. Nodes are tough to learn but fun to mess with. I have a few renders from actual SDXL 0.9 but they are looking like something I would make about 3 months ago. I might post a few when I figure it out
i hope automatic will release an sdxl friendly update
were almost at mid july
same i mostly cant wait till the training updates to uis will enroll
cuz i train models
nice!
Worst case, Vlad already has SDXL support in alpha and is actively working on improving it.
what model was this?
1,CheckpointYesmix_v15/ 2,anyhentai_20 / 3,CheckpointYesmix_v15
slowly getting closer. made the base image in SDXL and then brought that into A1111 and img2img with flat2D @ .4 denoise with SD upscale.
def weird things happening but overall, fun to try
MVP Nubbers, always got them links
My computer can't even run SDXL I bet
#1071936820121698304 message
The resolution possible is off the charts
at least to me lol
you can really zoom in on that image
you can see the chemtrails in background😹
I downloaded a niji lycoris https://cdn.discordapp.com/emojis/1093686691388198942.gif?size=48&name=Chika_Dance&quality=lossless
:ooo i hope u like it!
ooooo
awww omg so cuteee

Cute
you could do all of that in comfyui y'know?
sorta lol I guess I'm doing the heavy lifting in cpmfy but I'm more familiar with A1111 as far as img 2 img
dangit
I had a great gen and then I accidentally closed SD while it was hung at 100%
ok I got it back
Thic
baraumbababa
I've asked in two other channels but...anyone know if stablestudio running with automatic backend is able to run extensions, lyco, lora, embed, hypernet, models, esrgans, samplers, models, control nets etc?

Perfection from an imperfect model.
SDXL
heh this prompt guess made no sense
i mean, if i have to compare it with sd1.5 or 2.1, it's a huge step actually
0.9 can be downloaded with auto-approve registration from hf
i'm hitting ~12gb i think?
which one out of all is it?
damn
8gb not gonna do it then
If I am to continue doing anything with AI, I need to upgrade
either that or like sod it
wait is it Everyjourney SDXL youre talking about @wispy canopy ?
no, the one by stabilityai
I don't know where that one is
it seems the other ones have been modified to allow lesser vram
and since I have no organization, company, research etc. I can't get the official one
They are the same in terms of perceived quality. Don't hesitate to use them. You can't tell the difference without instrumental measurement, and even if you have VRAM to spend, you'd be much better off using it for highres fixes, controlnets and other advanced techniques.
getting everjourney sdxl and sdxl 0.9base fp16 pruned
yeah but I can't seem to use the official ones anyway
not by vram or by me being a user not a researcher
Oh, so that lora is made for fp32?
Couldn't you just convert it to fp16?
That Everyjourney appears to be FP16 Base fine-tune
so anyone has a 3090 they wanna sell cheaply? 😛
So it's pruned, kind of
I dont mind pruned, unless I wanna train on it, it doesnt matter
is sdxl considered 2.1 or higher/its own thing?
I'm also getting a hold of umh....firefly photoshop
Great stuff, been using it to outpaint
damn thats so good, can we have prompts and model
photoshop generative fill isn't bad
Image 1: Just a good ol gen from some months ago
Image 2: Photoshop Firefly Generative Fill
just playing around 😛
it's still low resolution but that can be fixed with SD
wow, so fumé
fumé?
SOGD FOOOT
why is my direct message tab telling me I have a new message and I dont? It's not on pending, it's not on like requests and its not on active...nothing....is a ghost speaking to me?
hah PINK NAME
it might be time to buy Gyre AI plugin for Photoshop https://www.youtube.com/watch?v=7z7S0R1_-PQ and just replace most things even auto does 😛
I guess some of it is in photoshop and some in a webui
or gui
ah yeah there's a free version of Gyre
then there's paid which includes photoshop plugin, interesting
because no one is talking ima just post a meme
bah its not very intuitive, cant find a way to use my own local models
Tried training genshin style, but may need ti train again with more characters in the dataset
Rem ❤️
Coffee anyone?
Don't mind me. I would take that
Nay, but I would like to inform you all that I finally popped a very painful pimple that was on the last place you want a pimple, fyi
thank you for listening

Ouch
😨
thank you for your concern
dance on all night with the vampire sisters, then wake up dead
imagine waking up as a vampire inside a dumpster in the middle of the day
and you have to sit in the filth, stench with things touching you and crawling on you and juicy garbage leaking
cause if you open it, you'll be set on fire and die
why is this channel always active when im not there 😢
dats a big halo 👀
Go on
Eat the cake
ckpt + lora
80's lora? cool style
@wooden coral i generated a few with similar cyberpunk style few days although yours are more realistic
backgrounds really similar tho
Yeah im using breakdomain realistic so that makes sense
Genshin LoCon is surpricingly hard 
supposed to be sorted food
she doesn't look hard to me 😄
hides
oh boy ok I think I used too many prompts to describe violence
wait what? I can't say a certain word that....does nothing to say in here?
what the hell
I mean can I say bifurcate?
apparently
eviscerate, vivisect, gut, devitalize, garrote, exenterate
ok so I can say all the words that the one word describe
but the server blocks that one word
uhuh
its not an image
its a word
what would you say describes those words I mentioned?
if they were to be put into a category
and dear god this prompt is not holding back
check dm
I will make fun of furry stuff til my dying breath, but I accept you. Weirdo. Fellow weirdo
most furries ( the one's I know) are big anime fans too
I'm just a weeb. Not a furry at all and I don't see the appeal. Although catgirls etc with only the ears can be cute
wait.. we already are 😄
Furries aren't that common, just very easily visible
hmm thats different
hehehe.. riiight
😄
however you think there are, x's that by 1000
however many. blah I can't type
ah some of these uh bloody generations are so good, I wish I could show them but I can't and I know I can't
hehe yeah i've had those
nekomimi😘
hmm not the cat cafe I was going for
well...her armor is inefficient but that's what I get for forgetting my negprompt
to battle everyone!
girl... you have something in your eyes...
battle rage
she should consume less red bull lol
she drank another one
can anyone here help with training?
never trained anyhthing, my gp is not strong enough 😦

Impossible, there's no dust on it and it's all quiet
I believe I must make some food, burgers perhaps. Maybe a few fries. Then watch a horror movie but not sure which one
Malum was ok but left me wanting
trying really hard with SDXL
so tough to get something that looks as good as 1.5 for anime at the moment
who is she? lol



