#🍥|anime
1 messages · Page 76 of 1
The real hakos bagel
@magic wave I love that picture of peach in anime version from your profile pic
@magic wave Could you please send me a pic of her?
Uhhhh
Sure if I can find it :o
Is that the name of an anime studio?
Idk
You need to download a LoRA for it https://civitai.com/models/6526/studio-ghibli-style-lora, guide on how to use and install these types of files are in the description.
Studio Ghibli style LoRA Making models can be expensive. Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee...
Theres a v3 now?
Apparently
Extremely beautiful!❤️🔥
Thank you!!!
Wow, you also used the figure in your profile too
Yes it's my oc ! I make loads of oc
I also make league of legends style splash arts
Thats awesome
Ty I also train models the image above was made with my merge called anime punk
I could never see myself creating images with such quality
I always get figures that are disfigured, ugly
Oh I wanted to upload a tutorial on how to achieve this
Which graphics card u using ?
Or vram
I can dm you and help maybe
OK
I dmd you
lool love these look so good how did you get this so be them like that if you can share some info maybe?
not when you are merging
Thought it only affect training
No clue if merging counts as one though
best high contrast LORA atm?
Also, I'm gonna do some experimentations regarding merging. I'mma merge Aurora, DreamShaper 6, and Uhmami
How exactly do you do merges in a way that gives out the best results?
has it been figured out yet how to control the hair colors when using tags like streaked hair or two tone hair?
merging is almost done...
This is the combined power of Aurora, Dreamshaper, and Uhmami.
It can be photorealistic (this is not upscaled nor does it have all the intricate details, so it's rough).
hello everyone
heya
And it can do anime, though now it's semi-realistic at this point.
Here's a better output. Like damn.
While the model is good, problem is that you have to be extremely specific to get to the desired output
you guys know if its possible to merge 2 models(LLM) together and convert it into ggmlv3 q4_0 format with only 8gb of vram?
ask on projectai
Added a bit of shading
ok, so i have 1 problem with gunaco and/or chronos, i cant for the love of got get the right settings, if someone uses 1 of these models can you PLEASE give me your settings +instruct mode settings and so on
nubby writing a novel lol
All the PNG info is in the images, feel free to take a look. These two LoRAs are doing most of the work <lora:loraslayernoiseoffset:0.45> <lora:pen_sketch_v1.3_dim_16:1> the first one is from Closertodeath and is available on Huggingface. The second is on Civitai.
I type slow on mobile. 🧓
👍
i repeat
there was a model for that
I think it was a lora
This is the first SFW LORA I uploaded here.(XD) Using this LORA allows for more controllable generation of different hair colors inside and outside...
seems a bit overfit from reviews
or overtrained
yeah but thats only inner hair
I want a way to at least relatively control the colors that i specify when using such tags
though most of these loras are kinda trash imo
there is the cutoff extension thingy
Thats not really what its for
If anything that maybe does the opposite
yeah
This LoRA was created to produce images of women with 2 different hair colors. It works better with some colors than others, and works best with hi...
Thats another one yeah
technically you could also do latent couple
Eh?
?
For hair colors?
yeah
though it would be hard to do
and not sure how well it would work
but it should theoretically work
I thought latent couple was just mainly for generating wider images or two or more characters
it is mainly for that
it basically allows for different prompts for different regions of the image
so you can specify a specific color in a specific region
Yeah that would be hard to achieve then
My favorite kind of multicolored hair thing is streaked hair
make a lora
Which is way more precise than some of the others
A lora wont help
Well not entirely
At best i could only train the lora to respond and use specific colors, and then add dashes of another color in it, but whatever second color it adds i wont have any control over. The only way i could have control over that is if i made a dataset for every combination of hair colors there could be
yes
just make dataset
easy
😎
scrape danbooru
Most models already know what streaked hair is anyway, its just as far as ik there is still a lack of a way to specify to the ai which colors to add
What i just said disproves that.
what
just get every combination
train lora with good tags
You do realize how many color combinations there are right?
trying to get the ai to allow me to do this through just a lora would be damn near impossible
I'm assuming this isn't AI, because if it is Wow!
who is the artist?
No its not lol just using a reference
I was lowkey alarmed lol
what's going on with the ribbon I wonder
https://twitter.com/FLprayer/media u decide if thats enough to train off of
👌
i can try
I've used prompts like streaked hair, white hair, red hair, black hair with reasonable levels of success.
huge heart
No special symbols or anything? Just list them?
Yep, just chuck them in the prompt and send it. Some color combinations seem to be more reliable than others. You might need to experiement with different levels of emphasis on the various hair colors.
huh I got this instead
o different prompt btw lol, thought it was interesting
1girl, close-up, white background, streaked hair, blue hair, (red hair:1.4), red eyes, blue eyes, heterochromia, white shirt
The heterochromia probably influenced that.
1girl, close-up, white background, streaked hair, white hair, red hair, black hair
It all depends on the model ig
Hmmm
Yeah, so with this example, I would try lowering the emphasis a little bit on black hair if the majority of them are coming out with mostly black hair.
wtf is this

Wtf

uh
thanks, first thing I see
when I decide to open discord
I curse you @compact warren

With intel fixing firmware etc enabling their GPUs to process tensorflow etc
I think they will be a go-to for ai/scientific computing in the future
with whatever their next line-up is
OpenVINO or something
or I might have it backwards
what model
im the type to swap places lmao
Anything v5
ugh anything :<
Anything is fine
Not really a fan on how the eyes look though
😳
Lol there are tons of tags you can use to get a different style, plus thousands of LoRAs at this point.
usually when people ask for model, you also mention if you used any loras
no one mentioned loras so I would assume this is straight anyv5
which doesn't exactly look like it
idk
maybe some tags can make such a big difference
anime screencap will get you the flat colors and shading, and one of the many studio tags can probably get you the rest of the artstyle. It seems like people have forgotten how to prompt since things have gotten so easy with having a LoRA or model for basically anything you can imagine.
I know anime screencap exists
but like does it actually do such a big difference?
bruh
ctd is way too good
fr no cap
(image not mine, made by ctd)
best stable diffuser
Closertodeath. A mod in the Touhou server. She makes nice models and loras, and also some really nice images.
Here's her Huggingface @arctic nova https://huggingface.co/closertodeath
Should I screencap and send over there?
dats a big hat
I want a second season of Mieruko-chan
and I think I'm done with these competitions, waste of time as long as the amount of people you know or popularity you have is a more deciding factor than the images themself. I have no evidence in how often, but I often see amazing renders that get nowhere near winning. Yeah it's in the eye of the beholder, but still. Makes me a bit suspicious. I will stick to what I enjoy the most. Prompts and ´model making ;D
should I watch?
also @cyan burrow I have been wandering, do you mind telling me what your job is? if you have one.
I usually don't like shows where the main focus is horny for horny sake
If it's just here and there then fine
I mean, monogatari is my favorite series
i really like mieruko-chan
Its only here and there tbh
did you create these imagee
ok I go slep now gn
what's the best sampler for anime?
Euler A 🙂
I thought we moved past euler A
Lul
I like 2M Karras and 2M SDE Karras the best. SDE and 2S a are also good
++ sde karras works well for me too
Any suggestions to get higher quality images?
For some reason my images are coming out BAD out of nowhere with the same prompts
Nice❤️🔥
She should marry Natsu Dragneel🤣
I'd try messing with the sampling method and sampling steps and see what happens
Are you using the bot on discord?
I am using my own stable diffusion setup
Then try a1111
Everything but Euler a looks awful now 😢
Huh... have you changed models recently? Or like, accidentally had restore faces turned on with an anime/3d model (I do that a lot lol)
I've tried lowering the steps and making them super high and nothing really works now
I havent messed with my model in a while actually
I might just give my PC a break and try tomorrow...
Might just be worth a restart, or checking for updates. Not sure what else you could try... maybe someone else will chime in with suggestions
can you generate an example and drag it directly from the folder on your pc into discord so I can look at PNG info to try to see what's going on?
Yea let me make one quick
Idk if I am just being TOO picky or what but they seem way more blurry and of bad quality than before
Yeah, it's pretty good. It combines black comedy and horror rather well with mystery etc
Its a little NSFW... lol Should I dm it to you instead? No nudity ofc
I am currently on disability. Previously I've worked as both a copywriter and contentwriter for both internal and public articles
Feel free to DM me whatever generated stuff you like and if bot stops you, then just add me as friend ^^ If you want
This is just for trying to figure out why stuff is being weird lol
Yeah you can DM it.
I love everything about this lol
I've been posting quite a few in a similar style for about the last day or so.
This one is my favorite #🍥|anime message
man I have so much dust and crap in my keyboard
need to find a good way to clean this thing 😛
dishwasher 
😐
Lol don't actually do that. Just pull the keycaps and clean it with a with a brush or paper towel/microfiber cloth and a small amount of water then let it dry before plugging it back in.
pulling all the keycaps sounds like a lotta work
that's a man's hand ._.
Her clueless mix is pretty cool. Not my preferred style but I dig the kinda painted/pastels look
I sense evil is afoot
the beauty of nightmares warping our shape and form
the more that die in the nightmare realm, including mares themselves. The stronger the remaining get and for each who dies in the realm of nightmares, a dreadspawm is left behind. Waiting to be birthed from the feelings of the dreamer. Until they dream again. Once they become strong enough, they will invade our reality
dont worry about the red stuff, just ketchup. and the purple stuff, he spilled some colored soda
good morning friends, I greet you with a trout
how can i use this
A trout? :O
Use the AI?
it is
I finally found a new place to go after rarbg shut down abruptly
now I just need to go seed myself to crazy
RARBG? :o
a really big and popular torrent site that had every movie/series out there ^^
Ohhhh thats fair then :O
I usually watch movies/series online :o
Also, I found a nice LoRA 
Ah yis mommy purah
Sorry for absence, have a guitar waifu
Indeed 
O, das fair, nice to see you again Eface 
been playing some starsector ^^ fun indie game
right now working in datacenter, why not generate waifus with A100 while i'm at it 😄
not enough ridiculous women in channel
#ytshortsindia #aigeneratedart #trending #midjourney #shorts #tiktok
she's probably cold.
Tru 
You are pretty good at drawing on paper
Interesting
There is this amazing server that has every single anime in RAW version
I can't post discord servers, as it blocks the message
Here it is
Thank you but already got that torrent sites floooooooded with anime
Is it raw though
both
Send me the link pls
@cyan burrow Weird format though
I've tried torrent sites before, but they weren't good
It's one of the oldest repositories of anime online and the tracker most fansubbers etc use for their files
The server I mentioned uses Mega file share
Can you help me?
mega file is awful
I have a problem
there is little information provided that I can answer that correctly
I've tried downloading this but it always doesn't work
colab ?
oh, is it on free colab?
Automatic1111
yes but colab
just use colab
is it the free version?
haahha ok
i have new i make own dm me
if u want
You mean a newer version?
yup with lora and checkpoint all in one
almost time for Mushoku Tensei: Isekai Ittara Honki Dasu Part 2
The document you sent me was confusing, I'll stick with a1111 for now
i make it ez to use and u cant use it
I can, but
I dont want to
🙄
Damn these are some cool jobs
Though AI really hurt them
Especially copywriting
@magic wave when did you change PFP?
Uhhh, around like...
2:28 AM :O
drinking mana potion during raid or smth?
Is it better to download all of the models of ControlNet, since they take up a lot of storage in my drive, so I only downloaded the "Lineart_Anime" model but so far I didn't get good images
the models all have their own uses
has anybody ever thought about making a negative prompt lora but just for guns?
mierukochan is amazing
wait, they made an anime?
ok, lets go with cliche manga was better
this anime is totally unnrealistic
japanese schools don't allow such weird hair colors
i'm not sure if the anime can capture the creepyness tho
like half the students have non-black hair colors
probably no
like halfway through ep 1
that sounds like a good start vs anime protagonist syndrome
no creepiness whatsoever
why do so many people download anime
oof
anyways, the ed was better than the op
"i'm a collector" xD
That's still an oof
@wispy canopy I have actually been considering dual booting windows and manjaro (or some other linux distro)
should I
why not just wsl2?
I want the OS
itself
isn't wsl like a container or something? idk I am bad at computers
basically, inside wsl2, it'll act as native
that's why i can use linux repositories on my windows machine
it's basically above containered, as far as i know
yeah but that's just a container of environment or whatever they call it
like there is no desktop
i just use windows as i'm very used to the shortcuts and explorer
it responds to low level shit pretty well actually
explorer????
you use the explorer?
i use win+e
brother
- win+r
now why would i use that?
but i already know where everything is.
that's just making sure everything is put on the right place
just try it
try using adetailer
adetailer ?
show me the way
extension
never use it
try
how i nerver instal it haahha
https://github.com/Bing-su/adetailer thisone ?
ye
ok done and how to use it ?
enable
that's about it
where i put my prompt in this area or the top one
I do empty prompt
but the top is positive, bottom is negative
this is the prompt it is gonna use to fix the face
Download transcriptions at https://georgecolliermusic.com. Check out the Discord!: https://discord.gg/3qxNDA3NCy
Original video: https://www.youtube.com/watch?v=ap88ic4BucM
Piece: Joc de Doi
Performed by: Nicusor Tunea & Nicusor Tunea Band
Transcribed by: BA
faq:
Q: how are you related to jacob collier? / A: i'm not, we just have the same last...
man got that tri-brow
Ah yes, my drawing skills are unmatched 
I turned emojis into humans with the help of ai #ai #shorts #midjourney #trending #viral #emoji
i see saxophone, i always think of this rendition of megalovania https://www.youtube.com/watch?v=Z1JHxDYILa4
@wispy canopy that's amazing 🔥
did you see their most viewed video though?
their touhou songs?
i like their (real) orchestral touhou songs
like for real, that's so well played and mixed 😮
@brittle jay nice one bro more better when i use realistic lora
detailedproject very nice
So apparently Clyde hates pink sweaters
Eew funko pop /lh
Thats not a funko pop
I mean it has the funko eyes though. And the face shape
Unless it's a different reference that I'm just completely missing
I think u just need to look at more funkos
I'd rather not 😂
ahoyoyoyo
Hi, i'm a noob so can someone explain me what is the difference between anything v3.0/3.2++/4.0/5.0?
any v3 & v5 are form original creator and v4 & v4.5 are from another person who just wanted to be troll
ok why today is out 3.2++ after 5.0?
Haven't followed that model for long time
Ok last questions why when i generate images are shitty? where am I wrong? Where i can learn?
Sometimes makers of those loras add recommended strengths to download page readme
multiple loras usually requires lowering individual strengths , too many will always overbake an image, the exact amount depends on both the model you are using and the loras
I guess you have experience in this field 

Finally some answer
in general loras that conflict and cause overbaking faster are character loras, having a style/character/concept/lighting will turn out better than character/character
add_detail 3.0
my problem is subject loses definition if i use more loras, quality drop really a lot
lower the weight of the loras below 1.0 , then increase the prompt tag associated with the lora you want the most
Anyway thanks for answer questions that i've search for months
i will return in another month then
The community isn't noob friendly, seems like everyone has an phd in computer science 🤣
is a lot and fast, hard to keep up and harder to provide answers in context without knowing what level of knowledge there is, feels like a few college courses worth of stuff since it launched last ...aug? sep?
I just usually expect that people use google
Or is it just that because I'm programmer I know how to use google 
For some staff ok, but things are so fast that when i find it is already obsolete
damn it I lost the extension that lets you generate a preview of a merge without merging...What's it called again?
Supermerger?
Well that I haven't heard of
It includes mwb, experimental range (going into negatives) and such. You select a model you want to merge with (your primary loaded normally) and then you make the settings and then you generate an image without merging first
it will make an image just like if it was merged. Only you can make small adjustmentsover and over
Pretty much same as supermerger
Ty your advice are great now with new checkpoint it works better
@gloomy scroll there is link just above
Actually I think I tried this once but I didn't rly have enough vram for it
weird, works fine on my 3070 ti
Yeah weird that my 1070 had problems 
Same, coming back with a 3060 
Till then. Lots of diablo and various steamdeck things
soon as I find a reputable seller for 3090 and it's for 700-800 dollars
I'm going for i
then quickly sell my 3070 ti for 400 before it drops
this merge couldn't get her to cause violence but instead..it seems she uh found the refridgerator and grabbed some raw food
yes that's it
depends on which community, as it really is seperated into a couple of ones
like in here I wouldn't say everyone has a phd in cs
but if we are talking about projectai
then maybe
The difficult part is gains reliable informations, and things change really fast so online tutorial gets old in minutes.
I would have recommended this https://www.sdcompendium.com/doku.php?id=start
but niaki doesn't update it as often anymore
before some explanation i'm using some wrong checkpoint.
first image🔥
I was scam with fake anything (v4/4.5)
Lmao
aye, I'm focusing a lot on getting darker stuff to work this time around, so maybe I can spin this off into another version
That was the whole guys goal lol
Still, cant deny it worked like any normal model
I don't understand how that markets?
No wait no real scam, but lost time
Probably joking
because people thought it was real so people started dowwnloading his model which then made him be more famous so his next models (for example pastelmix) also got popular
and then he sells a merge of one of his models to a service company for an "undisclosed amount"
nice, i'm merging my model too, gonna see if i can add in vtuber knowledge without corrupting semi/hyper realism style
Technically that's fraud and if he lived in another country it would be hard to get away with
We really gonna ruin the vibes over a joke? Cmon now
oh andite not banned from this server I see
haha
it wasn't a joke though?
Crazykazoos comment was
pretending to be the creator of an existing product and that you made the "sequel" to then make another one popular and make money from it
Bruh
Dont help him carry this lol
but like it's not really fraud let's be honest with ourselves
We’ll let u believe that then
Back to art, please
especially this early in the AI game

I pray
dude...that is the worst fucking way to try and defuse a situation "we'l let you believe that then", come on
andite playingn 4d chess while we out here playing 2d chess
you could have said anything, even "well it doesn't matter, let's get back to art"
@brittle jay you can play chess?
Whats wrong with letting you have your beliefs?
I CAN play chess
Everyones allowed to have them
not that I am good at it
so post modern
...you know that the tone of it was what I took issue with, kind of expect better from you
The tone can be taken in various ways, how you took is simply the way you took it. I have no control over that
andite lives in the Phillippines right?
listen, "anything" wasn't trademarked or anything and was released openly into the internet by some chinese people
he can name his model in whatever way he wants
not fraud rly
it was released under a licensing
yeah licensing but no trademark or whatever
I'm also not convinced it wasn't all just a happy accident on his part 😛
rather than 5d chess
he said before or after the fact?
after
yeah...
if he said it before, I wouldn't doubt it
still I think he lives in the phillippines and was in a pretty rough spot on money
so that he made good money now
good for him
yes
man I wish this server added the png info bot already
I never add any info to my png
what? youre not talking meta data?
its not very crazy lol
bowsette, black dress, surprised, crown, horns, blonde hair, cute fang, lora:bowsette:1 lora:迪斯尼小人001_style_mech:1
(worst quality, low quality, normal quality:1.4), KHFB, AuroraNegative
lora i found, and mislabeled so its uh....dunno now
I type in google and get this
what is aurora and why do people so often use it for negative?
embed made for aurora model
aurora is Luna's new model, successor to kenshi kinda
and why on negative?
Luna made two embeddings for the model
KHFB also one

im too lazy to change my negs is the real reason
same buddy same
“Explore the New Horizons.... Don't let indecision limit your potential, The world is full of possibilities.” Aurora is a Stable Diffusion model, s...
...i vaguely remember the style lora now xD should be on civ somewhere
preprocessing 160 seconds
what is fubard
by this time I wouldve completed 5 merges
not sure if it completed it or not since it crashed upon doing so
one way to find out
I swear that's a hamburger he's eating
Lee Hodges was given a 1-stroke penalty for waiting more than 10 seconds. #golf #fyp #pgachamp
such drama
held my breath
I'm not into golf and I did too
is that a vr headset?
this is the new bigscreen vr headset
Bigscreen Beyond
i'm still shocked that it's gonna be $3500 for that headset
similarity is unparalleled
is that vr or ar headset? 
I was thinking it too but thought itw as too obscure lol
vr, no cameras onboard
vr
i think it's the next contender after hte index line
I don't get why a lot of these new ones don't have that strap on top of the head
oculus quest 1?
i think it depends on the weight of the headset right? like if it's light, then there's probably no need for that

I mean like how light could they be???
I get it on like the bigscreen one
but like apple vision pro
light enough for them to ommit that strap
wait
did apple vision pro not have that strap?
no
oh right

might be light enough for that
and tbh, it just looks comfortable enough
i think if you compare that with meta quest 2
then yes
it also depends on the length of the headset
like, if it's kinda heavy, but really slim
then might be alright, due to center of gravity right?
well, im definitely not buying apple vision pro for sure tho
if I wear one without the strap then my nose hurts a ton, or I have to tighten it and then the rest of my face (esspecially the cheeks) hurt instead
nah trust me meta quest pro is hella comfortable even without the top strap
talking about quest 1, and also 2 from what I tried
apple vision pro is apparently like quest 2 weight which is insane
considering they have so much processing power in there
too heavy
can agree with quest 2, it's heavy, and hurts after hours of wearing it
or maybe even hour
though IG if they focused a ton on making it like heavier on the back so it's more equalized or something
true, they also made the battery as a separate entity, so i guess they know they can't just squeeze that much stuff into the headset

it probably can tbh
like you can run sd on cpu
it has the processor the mac thingys use
How to burn out a meaningless 3.5k headset's cpu
I aint impressed with it lol
i bet you can run sd on your new steamdeck
Yea but not gonna bother since i got pc to lol
it mostly comes down to balance, the quest pro is balanced like perfectly or at least it was for me
I got diablo working on my deck
true true
diablo 4?
Yee
nnice
yup, i heard from my friend it's working great on deck
Keep settings lowish and yrah stable 60
9:50pm lol
?? where do u live?
I think I will skip school tomorrow
6:50 for me
you meant later?
pm
end of year anyways
yeah
east coast?
lmao
techniaclly today
it's like 9.50 am for me
Very sexy isn't it?
for sure for sure 
women + blood makes me feel ill
I don't mean like women who are bloody in themselves
Apple choosing to enter the VR market at prosumer level prices is disappointing
like women covered in other people's blood and relishing in it
well, apple is apple
though they did do it majorely cheaper than competition
yep....but I mean,, they'll be outcompeted by Meta and Pico at consumer level and Vaijo at Prosumer and Pro
they outcompete varjo
I don't think so
about same specs but lower price
Varjo has a lot of knowledge and expertise now after years of doing this at that level
uh.....apples is like 3500 euro right?
there's always super apple fans that will buy apple products
instead of like 5000/6000
depends which version
this is also their first step into their ar/vr industry
if you look at their models which are around the same specs
anyways, watch thrillseeker vid
I'm still waiting on sadlyitsbradley vid
true
well, i would say it's more of a ar than a vr, so if you are looking to play games with it, i bet it's gonna suck ass
though for enterprise or whatever you wanna call it IG software matters less
varjo's more expensive ones are both vr and ar integrated
enterprise doesn't need games
they just need their one use case to work good
or maybe a bit more than one
ok I think I wwill go to slep now
goodnight 
just wanted to share, but this is my desktop background
based on this random art I found online
I swear though everytime people come over, they think I am gay
Is there a way to make a pile of dead bodies?
Fair, that makes sense XD
hmmm? wat?
A large farm with human blood and guts being sprayed over the plants, looking over this plantation is a blonde woman in a white dress with a black headband keeping her hair in a bundle, viewer only sees her back as she keeps her eyes peeled over the plantation. Next to her is a half naked brunnete with a spiked leesh. 2.5d digital illustration, digital art, anime style,
Oh yeah no no I see it
\
lmao
yeah I'd probably assume that
necromancer
Okayu :o
Where do you guys post your AI art?
time lady
Yee
Got an extension that can work similarly to this?
https://github.com/ashen-sensored/sd-dynamic-thresholding-rcfg
guys
There's this one but I'm not rly sure are these how similar
https://github.com/mcmonkeyprojects/sd-dynamic-thresholding
Oh I guess Ashen's is just fork of one I linked
But I haven't atleast heard any other extension that does what dynamic thresholding does
@lyric zealot
what skpt model you used for this?
My guess would be Fav_ExpMix_Line_v2
thanks
interesting
did you use highres fix?
i used resolution enhancer since my vram is only 6 gb
boom boom boom boom
the denoise ratio seems to be on the low side
@nova remnant close but it's v3 now 
lol
what is prompt
That is a good question



