#🍥|anime
1 messages · Page 103 of 1
was just running pop-culture wildcards
i'm just building a confyui set up
it's funny to set up
now i'm trying to set up a control net
im gonna do that someday
at least half of it is
working for a sci fi would be the funnest job
lol
theyre gonna make the klingons sing arent they
oh just the whole ship
that was good lol
are u supposed to use multiple negative embeddigns at once or are they meant to be like a replacement for a standard negaative prompt
i put unspeakable-horrors ti in positive prompt and got this
hey, random, but can ya guys rate my new site?
sure
its in very early stages btw
just kind of a prototype(i was going for a neocities/2000's anime site kinda look)
Lets Love Lain
maybe expand the colour pallet but i like the purples, im not sure what colour id ad
how did u make it
if u dotn mind me asking
used this as a base: https://sadgrl.online/projects/layout-builder/
A tool to help you make a responsive layout on Neocities or any other web host.
then i just added random ass shit
now it looks cool
cool
I'm having trouble getting consistent MILFs. Any tips or good prompts outside of the obvious ones?
well mature female comes to mind
add loli,young face,small breasts to the negative
I have that prompt cranked
Thank you!
guys, someone know a stand alone ai to convert images in canny?
an edge detector, canny generator or something like this
i'm missing so much the extension in auto lol
preprocessor ?
yeah, that function
oh, found the note in comfyui
ty
let's see if it work how i want
it's pretty slow the preprocessor in comfyui
on sd is instant
but the node works as i expected
i suppose it had to load the model for the first time
now i have controlnet on comfyui, i'm crying 
holy moly, my gpu is at 75° C
is still safe, but it's very close to the limit
now i need an img imput lol
If I had to guess, Comfy is probably unloading the preprocessor and controlnet model after each pass. Auto keeps it loaded until you process a job without it.
comfy just jump the nodes you didn't change. so, usually, if you don't change the image or the model, you don't face this problem
now i'm trying to abuse the fast upscaling the comfy make very big and detailed images using control net and imput
let's see if it works as i hope
unless i just try to assemble a tiled upscale like in the extension of automatic1111
but it's not funy like this lol
just edit it lol
angry
very angri 
Hi very nice anime wrinkles, I really like them.
Hi very nice anime wrinkles, I really like them.
not sure what is anime wrinkles, but thank you 

What did she mean by that
taylor swift anime.
AI only knows how to speak Monster Hunter language.
She spilled some ketchup
mornin'
love the wide ratio... glad to see someone else doin' em 😄
I guess this is a character people know. I don't know too many characters, but models seem to know this one pretty well. 🙂
what's everyone up to today?
Sorry I scrolled up a bunch and wanted to say I fucking love this website lma
Looks awesome
Yeah bro that has mad personality
links out to other sites that are similar.. didn't know so many people were making these haha
@keen hollow @stone merlin that you for your feedback guys, any suggestions?
add content 😄
yea, i rather meant design wise
i can write as always
bc i wanna design first, then rework the html+css structure(because its a f mess), and then i can add content
thought this image was funny...
not really, I think it's good now
ok then, what content could i add(i am not really creative so yea, my current idea is a little blog thingie, and some buttons linking to my socials)?
I have no idea. 🙂
think about the sites you visit and why
I don't think too many people read blogs anymore.
ik, i wont do much there, its just for documenting what i am doing
Today, I present to you, a most terrifying accomplishment, with the help of @sweet bramble Dorito hair.
how do you guys think npc's in games like skyrim will look in 10-20 years?
not sure the human race will be around to enjoy them
same here 😂
what, why?
have ya taken a look a the air streams and the ocean streams, the ozone, meglomaniacs, crazy people, religious nutjobs (of many differnt one's) and the reasons go on but eh I'm GEN x. been here before, it is what it is. life goes on. or doesn't shrugs
😄
Published a new bugfix to my image viewer: https://github.com/GarlicCookie/PNG-SD-Info-Viewer
Fixes the reordered column headers in detailed view.
some of you used to use it, not sure if anyone still does 😛
What model is that? :o
the model is M4RV3LS & DUNGEONS
I'll have to check it out :o
yea it's one of my favorites.
now.. to make it more anime.. sorta
well.. ok then
I think it lost the plot.. kinda
it's weird, on comfyui, from gen to gen, it get slower
is there a leak of memory maybe?
is cirnoposting allowed here
😹
in 10-20 years, video game studios wont be a thing
in 10 20 years, art in general will not be a thing
i wonder wat kind of games the singularity will enjoy
can you guys stop saying the world will end
lol
people really want it to, because their lives arent in a good position, i used to be exctly like that
who said the world will end? i'm just saying thy are flattening every form of art and culture, favoring mediocrity.
true
some guy before you said that
oh, still to young to know how the world goes. things never ends, just change
like water, with enough time, things will find their stability
ponders .. hmmmmm
do people actually buy nfts of hungry buirds
very cool
Model?
ahh id like this so much if it were a little different
i blew it up anyway
for inpainting , does the size of the whole image matter or just the size of the inpainting area, like can u inpaint on a huge like 4000x4000 image if its just a small area
let me check
ahh the merge I made
A merge of several different models, Elegant Entropy, aZovyaRPGArtistTools, mistoonAnime to create a new Merged mixed style. works pretty well I th...
holy crap someoen has been making alot of art with that merge
sorta cgi anime realistic.. kinda
well it tried
Is that niji? Kind of reminded me of old ones I did in that.
its niji
Ah cool, after it went out of beta and started requiring the subscription I haven't used it. It was fun while I used it though.
i remember when dall-e came out this was the best you could create with it,one year later and we have sdxl masterpieces already,cant imagine what will happen in next year
it realy is nice soon there will be a app for it i cant wait
is that reimu
is a rtx 3060 12gb good enough for using sdxl?
Could i have the parameters for that?
It's a lora
ponders an anime tarot
Yeah, especially with comfyui
Could you send it to me please?
well it's upside down but not the hanged man . sigh
hmm or maybe it is... shrugs
tries again
sigh, it will not do it
awesome
I got a really cool one cooking right now
cringe normie...
ty tho, I'll use this as my pc wallpapers
lol
it's an interesting feeling seeing this on someone's desktop
about to see if the 8x upscale worked
2x, 4x
the 8x might be considered NSFW
Guys, is there a way to use 2 character loras in a single pormpt?
yeah, just use them
look up the "latent couple" extension and "composable lora" though
you can segment the image into parts and write a prompt for each part
the weird hand feet anomaly drains my energy level
hands usually just need a little fix
🥵
cute
give that guy the 4090
Wonderful
https://civitai.com/models/123036/animorphs-cover-sdxl nostalgia, but wtf 430mb lora
i don't know why, but when i set controlnet, the image generated lose sharpness
What is the prompt for this o-o
Same for something like this
you asking for the prompt? 
metadata here
!
if someone made such a monstrosity in real life, i would die without a doubt. what a nightmare fuel
I think this one was using the vaporwave lora on civitai
The day tech automatically conveys our mood through every message we send, we're all doomed.
he is not, but i'm upset because he is not upset
i am upset,angry even
yeah, right?
i did use clearvae,i just wanted a washed dead horror kinda look
some SDXL results directly from the bot
PANC
does anyone know the correct way to inpainting, for example I want to take off the cape for this image, how do I do that?
I tried but couldn't take off her gown to reveal her bare shoulders and collarbone
in my opinion it will be easier to initially indicate this in the prompt
oh, so there is no way to inpainting from img2img correctly except controlnet inpaint-only.
I think the result for such precise work will not be good
rip fingers
How!?
If i make a merge using darkalfa and another model will the prompting stay the same?
The short answer is no.
oh ok ty
I should make a comic out of these 😄
hmm have to figure out narritive , and text to use
ponders
and ejecting from the crashing craft
not sure where she would fit in
and from this one
These using SD or just Niji?
Any good Yusuke Murata LORA’s or models?
has anyone had issues with using ram trim type apps or razer gamebooster
ive had issue with sygil and auto111 in the past
assuming its shutting down some dependant windows service but havent tested enough to figure out which one
and process lasso memory trim also caused issues i think
devil girl for ur thoughts
i made this on the xl beta months ago somehow i think i had a dreambooth credit
she cute i like them blue too
it sould have all the meta dets but here is the prompt ((reference sheet)), (concept art), (a watercolor painting of a demon girl with red skin and horns, (Mab Graves)), ((different perspectives)) masterpiece, extremely detailed, page border.
Negative prompt: multiple girls, lowres, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, jpeg artifacts, signature, watermark, out of frame, extra fingers, mutated hands, (poorly drawn hands), (poorly drawn face), (mutation), (deformed) breasts), (ugly), blurry, (bad anatomy), (bad proportions), (extra limbs), cloned face, flat color, sketch
Steps: 25, Sampler: DPM++ SDE Karras, CFG scale: 4, Seed: 1905688386, Size: 1536x1152, Model hash: 423dc55b3f, Model: flat2DAnimerge_v30, Denoising strength: 0.41, Version: v1.3.2
hmm... ok
puts in , no other changes, just default. let's see
oh nice but i'm I can't post htat here
oops
uses most of the settings except negitive and a diff model
adds adetailer prompts and upscale. tries again
oh and my own negatives
there we go
hmm the hands tho
grins evilly as he tweaks the prompt
nice
here ya go...
tho I wanted a map as a background drawn but did get added. odd
(map:1.3)
i'll try that next
hmm removed seed and just gone random , interesting result. tho it's not done yet
and she has a 2nd form. wow
Demon horse
waittt how u generate that
and thats suppose to be A demon betty white
o nvm i see prompt
ponders foodage
Just "random prompt"
ooh boy these dan dan noodles are spicey
I can't feel my lips
which keywords are responsible in the prompt line for concept style generation?
anemone my beloved ❤️
this ... tiny multi-colored chromatic neon outlined repeating stacked randomly cuboids, partial darkness BREAK RTX, SHADED, Unreal 8K High Resolution, BREAK Hyper detailed, Conceptual abstract style
tho didn't really give me what I wanted
shrugs
👍 but I'm not talking about the cube. I asked about character concept illustration.
oh that was someone elses
this
tho I kept tweaking it
umm yeah thats bordereline my dude.. be careful
heck I would suggest removing it befor eit gets you booted
no
shrugs
hmmm...
just two sisters looking at each other
shrugs
i've been a mod (not here on this server) and my mod senses tingled and went oh no
tho I do not want to be a mod, gave me ulcers
so I'll go back to my corner and be quiet
enjoy my anemone instead 🙏
🙇♂️
can be sold. only immediately claim copyright on the design
nah just doing art
$8 a month works wonders! but the illustrations are amazing!
love the color style and the light
jeah niji is insane the also working on insane inpainting and realtime drawing im so excited
crap! it's so atmospheric! 
Experimentations with the Dark Sphera Style LoRA: https://civitai.com/models/121760/dark-sphera-style?modelVersionId=132505
Love the qualities it brings to things. I'm not a fan of horror though, so I prompt out most of that side of the gens.
#1025467151206854736 a boy
super cool
Nice
Sailor Moon
: https://youtu.be/GHytzQosPP0
Watch girls from #SailorMoon in this extreme high-ressolution render to celebrate the 100th video and 500 subs!!
Please watch full screen in HD and please consider Liking and Subscribing. It really helps spread the content and makes the channel grow.
Popular playlists:
https://www.youtube.com/playlist?list=PLmF5vTIA_9u5ZmMmuq3b3eRF1wWPLMri3
...
FINALLY got SDXL working properly with Auto's build. Relatively decent gen times as well. Now I just gotta relearn prompting and see if any other features work yet. 
base xl?
Yeppers.
damnnn. always feels xl got crazy potential but still waiting for the xl scene to blow up on the sidelines
so far still seems pretty barren, for anime at least since 1.5 scene so developed
Yeah, I getcha. I had been waiting for proper integration with 1111 and more of my foundational extensions. Now that some have been updated, it's worth trying out for me.
also ive built so many 512x512 datasets, idk whether to trash em or upscale em for finetunes 🥲
I wouldn't upscale them. That'll just introduce more issues than the problems it solves.
maybe for characters could be fine but not for styles
the bg looks more coherent than in base 1.5
why not use them as is
Without the Adetailer extension for instance, the first face would've been the end result rather than the more detailed, fixed second face.
oop i replied to the wrong meskage
i meant this one
waifu panda

Any specific settings I need to change, or can I just drop in the checkpoint and vae and use it?
is thos solvable
Go around center bottom. Ez.
I had to ensure my webui-user.bat was set as such:
Also, ensure tiled VAE/multidiffusion is off and that you have the sdxl_vae loaded in the VAE slot. SDE Karras and 2M Karras are the best samplers for it at the moment that I've seen. Additionally, set your refiner in the Refiner area as SDXL_refiner.
The no-half-vae is the most important part of the commandline arguments. But the --medvram is important for speed, even with my 10gb gpu.
Anybody need a bucket?
creepy
epic
looks like something coming out of a lovercraftian nightmare
heard theres a lovecraft lora havent tried it yet
yea on civitai
maybe u can combine it with some anime models for some cool visuals
i already found the right model i think, i have 2 in mind
theres a couple lovecraftish type loras ive seen but i barely used loras yet
C o s m ic E ldritch tech - World Morph 🐙 --- [✅ Marked on Pinboard ] --- This LoRA morphs things around it to create cosmic horrors and Eldritch c...
im not sure how theyre supposed to be used best in img2img or t2i
i'll experiment for fun
https://civitai.com/user/navimixu/models
ttheres a bunch of cool loras here
this guy is obsessed with coffee
https://civitai.com/user/konyconi/models
this one similar type but also a good collection
using kakigori btw
gonna try one of these loras and see what it does with the same prompt
the ai is going nuts, i like it
but it's scary...
ok, i'll stop
this is not healty XD
generating such images, it's like peeping in another horrible dimension
this one is rly cool imagine making a video like this
lol
Kemono Friends with your Kakigori V2 look very kawaii! thanks!
i like this one alot!
This gives me Transformers Armada/Energon/Cybertron vibes
Very cute
didn't go as planned
hmmms
runs in circles and falls over
hahah failed succesfully?
oh those hands

i love the 2°
❤️
Magical Rin lol
comfyui is missing a very important component. tiled upscaling
it's necessary to do a good highres fix
any tips for using this?
yoo what style model is that?
also this one is very cool, kinda reminds me of vergil lol
It is vergil
whats the model used for that? Really like the sharp style
hey, what prompts do you use to create characters from anime and they actually look like them, dm or respond to this ty
well, if it's a well-known character, then just their name and maybe describe them might be enough. But if it's not that well-known, then you might need lora for that.
some really well-known character just need their name
or you get something in between...
Aw
Could I please get the model name?
lol
kill it with fire
why?
it's crossing into furry territory
well prob cause I used a furry model. oops
robotic pokemon?
sorry for replying late but here: https://huggingface.co/natsusakiyomi/IrisMix/blob/main/IrisMix-v3.safetensors
yay 10,000 views on Swedish Quora
pretty good little mix. thanks
and back to bed, but first here have this (thud)
Thank you very much!
Smurfette
even smurfies have their puberty
see ya space cowboy!
True
Mommy Smurfette? 
wow, what's this model? it's fire
Model:XXMix_9realistic_v4.0 - v4.0
With Lora: Add More Details - Detail Enhancer / Tweaker (细节调整) LoRA - Add More Details, yuzu detail-rendering-colorful - unreal feel, 【Art Style】Sakimichan Style - 1.6, KoreanDollLikeness (v2.0) - (v2.0)
wow
Mainly because of those Lora its become like that
i must try
Sure. I try that in Tensor Art. You can directly use my prompt there. https://tensor.art/images/625479506389514304?post_id=625486356858177152
undefined
she looks beautiful,ima steal it
No problem bro
comfy will burn my head. i'm trying to upscale with tiled ksempler and controlnet in the same workflow lol
the target is to get an high res image wil the least time consumption possible
Oh XD. I use Hires-fix 8x_NMKD-Superscale_150000_G. And ADetailer face_yolov8s.pt
my gcard is pretty slow
Oh. I use stable diffusion on website
oh
i like pushing the boundaries of ai, day by day
that's the reason why i love comfyui so much. because i can set things without knowing pyton
Oh i see👌
when i code with pyton, things turns always off in some ways or another
Oh i see
being honest, i like coding, but i'm crap at it XD
Well atleast AI chatbot can help you with that. Im currently use Claude2.
what site do u use ?
yeah, that ai it's really a good wonder
dayum could i have the parameters for that please? i really like this style
i mean the image should be ecoded
upscaling the image
oh ty!
Now time to sleep and leave this queque of 7 LoRAs cookin
Mini Milk 
Serana - Skyrim
what software is that?
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts Gigachad Derrian's LoRA training GUI 
Yeee, I mean I train styles mostly, its usually good on the first attempt unless the dataset itself is bad :o
who were you on civit again?
LUUNA :o









