#🍥|anime
1 messages · Page 18 of 1
(even when not called, I'm such a no life I read the whole discord, sorry)
(I could still not have seen it though, if it was to disapear before I came back :p)
Nooooooo
civit ai down again ?
Guys, have any copyright issues with AI generated images? like ''artistic style, etc ''
pleeeeeeeassseeee
That bad?
yeah? isn't is just... I mean... what are you supposed to do with her in that position with that face tell me ? drink tea ?
I mean, true XD
hmm
XD

thanks but sex yoga is for over discords :p
I got idea 
O hi fun police 😉

Fair, was trying to get creative poses 
WELL
I tried again
And less lewd, but its still lewd xD

it's really creative tho


It do be 
yeah I won't say the overwhelming NSFW you get when opening Civit (even with filters on) has no effect on anyone
but there are servers for that, and servers for thont
I think the jack-o-pose thing is just inherently lewd no matter what outfit xD
Agreed
Unless theyre wearing a skirt and its from the front maybe
I mean, its from the front wearing pants
Idk what the context is here i was just being general lol
jack-o-pose? 
Yeee, lemme post to other server

I used to do this with waifu labs as well
civit ai still down T-T
again??
yup

what is this waifulab?

generate waifu there?

does it cost anything?

wtf it's fast af

that's the model right?
or am i thinking about something else
how do I download big files through github?
ok civit is up here we goo
like I want to download the controlnet 1.1 files in one git clone go
is it even possible?
Catqua
@magic wavewhat is the prompt
Good 
SLE, mksks style, best quality, masterpiece, detailed background, highly detailed, intricate, detailed face, long eyelashes, detailed eyes, (standing:1.1), (paw pose:1.3), (thick thighs:1.0), (medium breasts:1.0), (white background:1.3),
(one girl, 1girl:1.1), (solo:1.3), (aqua (konosuba):1.2), (blue eyes:1.3), (blue hair:1.2), (ponytail:1.1), (choker, neck bell:1.2), (thick lips:1.3), (fake cat ears:1.3),
(blue crop top, blue sport shorts:1.2), (black paw gloves, cat paws:1.1),
(shy, embarassed, flustered:1.3),
I guess no
@grand geyser
XD
Also, sad catqua
midriff oof

is auroraOne a new model?
judging by the immense silence I have concluded nobody knows how to download large files through git clone etc

I don't think I'll be able to post any generations with (groin:1.4) tho, judging from the first few generations
XD
Aurora is really good with the groin tag xD
wait, isn't git clone, trying to download the whole git?

oh shit, there's cow
O shit, I forgor about cow
@magic wave
anyone here who hasn't had a nitro sub before?
start genning your submission 
Im here >:o
And I might give it a shot
Order and chaos is a hard one tho

ok if you dont speak up now, youll miss out on 3 months of nitro
I might have a idea to hit a soft spot tho
who hasn't had a nitro sub before?
@magic wavewhat check point and did you used lora?
I do too
https://huggingface.co/SweetLuna/Aurora by @quick hill
ok so....noone?
And no LoRA
Free nitro pog
nope 
I mean, Nitro is pretty cheap
free nitro
yeah but Ive got a code for 3 months of free nitro
for anyone who hasnt had it before
thhnks
doesnt matter if its cheap, free is cheaper
I'll go ask in another channel in a few seconds
True 
Welcome to the community 

Welcom welcome 

Should probably move that to buildings
well, anything is anime enough tho



forever?!? 

how do i get rid of these purple spots in gens? been getting them in multiple gens
hey buddy, you might not be using VAE

Ye :o
Get that, rename to .vae.pt instead of .ckpt, put it in your models folders, and select it and apply it in your settings
that fixed it 👍
I don't even know you can get weird stuff from not using VAE o.o
jesus there's so many black names submission on COW
not sure who i learned that from, but yup

@wind girder Can I ask if a pic is fine in DMs? 
I learned that from experience xD
go for it mate
so abstract
on your model 
low cfgs?
thats crazy
nope, using that dynamic thresholding
1girl, chaos and order, neon light, trippy perspective, city,
Negative prompt: AuroraNegative, (low quality, worst quality:1.4), (bad anatomy), extra digit, fewer digits, (extra arms:1.2), bad hands, by (bad-artist:0.6), bad-image-v2-39000, loli, nude,
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 22, Seed: 243336760, Size: 768x512, Model hash: 80d5cce368, Denoising strength: 0.53, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 10, Hires upscaler: 4x-AnimeSharp, Dynamic thresholding enabled: True, Mimic scale: 7, Threshold percentile: 96, Mimic mode: Half Cosine Up, Mimic scale minimum: 4, CFG mode: Half Cosine Up, CFG scale minimum: 4
simple prompt tho


i thought this is low cfg at first because of the lineart etc
time to test myself
o.o
Liam Neeson will soon star in "Taken 4: The Home" Where he is Taken and to a home, a place where old people are held as prisoners
ok im suprised 7cfg works, but dynamic thresholding seems to help with generation a lot
1st gen
this is very trippy
this w/o dynamic*

@magic wavenegative prompt?
nope, yours
/shy

AuroraNegative, (low quality, worst quality:1.4), (bad anatomy), extra digit, fewer digits, (extra arms:1.2), bad hands, by (bad-artist:0.6), bad-image-v2-39000, loli, nude,
this is what im currently using

civitai is back i think
🥹
xd
time to break it again
(KHFB, AuroraNegative:1.1), (low quality, worst quality:1.4), (bad anatomy), extra digit, fewer digits, (extra arms:1.2), bad hands, by (NEG-badartistnegOG:0.6), NEG-bad-image-v2-39000, monochrome, greyscale
easynegative?
yes
i use the 84000
kl anime?
always kl-f8-anime2 for me too
there's blessed2 vae but I personally find kl more better imo


xd
it gets so hot on that damn line, I keep on sweating
Jesus, those hips.
damn.. you don't help x) I'l go for a second, I'm sure it won't be worse
I've never used easynegative. For those accustomed to its uses and synchronicities, can you explain to me how it affects the end resulting gen? It's supposed to curtail things we don't want obviously, but what exactly does it take away?
I don't really test that much, only really following author's prompt. If they are using easyneg, i'll follow

No idea. I just use it because it's way more convenient

no
i just upscale
oh
well you can read about it here right?
@lyric zealot
Lol. In short after following the link, "the effectiveness is not certain"
@lyric zealotthere is a way to do high res faster?
well, there's nothing certain about any negative embedding right?
With as often as I see it in neg prompts, you'd think there'd be a collective R&D, heh.
maybe try on lower settings first? But when hires fix is enabled, it's always gonna take longer
not sure tbh, because even if there's deep research about it, there's still flaw about it. That's why there's no a holy god send negative embedding right?
try 1.5 upscale?
and your hires fix steps can be lowered
like 10
f
Welcome to Clyde's capriciousness. 🤝
Can anyone suggest me prompts?
Be more specific pls 😔
positive: 8k, RAW photo, portrait, best quality, ultra high res, photorealistic, cinematic lighting, night,
1girl, pink hair, high ponytail, freckles, looking at viewer,
night sky, oni horns, oversized school uniform, long sleeves,
classroom, (dark, night:1.1), head tilt, happy, sharp fangs, holy halo, backpack,```

Actually good
whaat is it anime sharp?
Ive never heard of that one
let me find the file
ok
that's a upscaling model
heard a lot of people using it
I'll try, thanks for the link 🫶
Is it general purpose or just anime
oh nice it's back up
yup yup!
no idea tbh, i downloaded that, but haven't really use it yet
was it ever down? 
yes
anything V5[prt-RE] - is out
didn't know about that
yes

ok
Any so-bin style models?

what denoising strength are you using for your hires fix?
Look pretty much the same for me
Am I blind?
wtf
Looks the same
different upscaling model?
no?
Taken 4!
@lyric zealottakes 15minns wtf
xd
what it does
my bad i did 15 batch size
ftw
xd
denoising strength is like, after your generation, how much should it jumble up before trying to "fix" it again. With denoising at 1, it's basically saying make a new photo. With denoising at 0, it's basically saying, don't do anything. As such, it's wise to take your time to find the perfect balance for the model you are using
It would make the perfect action movie
I just use 0.5 in pretty much every model

🫡
@lyric zealotCRAZYYYYYYYYYYYYY
i mean, if it's nsfw, then don't post it

this is for me nsfw
i made pretty things
A-
it's close 
I want pretty things 
Yeah... There's no way clyde will let me post this

you saw all the pretty things made today


true, those very pretty things

wanna know something funny?
the guy before with the error, the guy who refused to tell us the error and kept whining
I have that error now
what error? 
`Can't find a default Python.
Couldn't launch python
exit code: 9009
stderr:
'python' is not recognized as an internal or external command,
operable program or batch file.`
🗿
Could try to delete the venv and rebuild it :o
yeah Im gonna do that
same error wont even build venv, gonna try reinstalling python then
Oh damn, yeah

what tts is that?
replica studios
I have accumulated 3 hours + of voice credit
it's the best one out there
so yeah
?
I didn't
how much hours of sleep do you have?
Too lazy to fix hands 

squirtle
dog.
dog
I may have to draw this character myself and then run it through img2img

This character I could get it done, not how I want, but it is relatively perhaps close 
But this one the AI just aint doing it 
this is so good
so so good
Thanke 
i love the kind of woman that will actually just kill me
XD
when i woke up today i was thinking: "damn, i really hope some hot chick blows my brains all over some fucking hallway"
Though, mouse girl uses knives, bob cut girl uses guns 

yk? As In the rapper?
you know

O, well, they're my OCs, so ye I know xD
literally my will to live


Sm famile
da funi... 

nobody got it

XD



e
e
e
They told me the devil would be attractive
Well, they were right 

Model? Looks beautiful
Aurora
Oh right thanks
my friend wants to know if aurora is capable of NSFW? hes so dumb...
Of course, it's mixed with anime model so it can sometimes be quite NSFW on its own..
Is it capable of fingers
gacha xD
I mean, its way more likely to be good
meru has grown up
PERFECT
It's a personal model I mixed
Didn't save it though
Colab went poof
civitai dead again?
CAT!
civitai rushing into things to make a quick buck when they see their high trafic, without thinking about how to run their new side of their website. It would have worked much better for them to have a patreon where they get donations to keep the site running and more
ok these images are disgusting
strange silent hill like creatures
gut morning my friends
Gmornin 
for some reason control net doesnt create a proper representation of my pose
Ooooof
😂
well the closest I've come
Does anyone have tips getting different colour thighhighs like this more consistently?
i think your best bet is to train an embedding or lora
civ is down?
Yup, down.
smexy
good morning 🙂
@wind girder, are you going to make supermerge at one point of your PoW competition models?
I don't go about it as a merge, because each model isn't really trained on useful tokens to merge together
But I'm going at it as a larger finetuning dataset on multi concepts. While not being full finetune at all, we are growing the dataset each week enough for it to be quite the project already
I'm around 25% done with captionning everything in a more coherent manner, but this is be a mega train, not a mega merge
and i'm actually very amzed by your models... at times it seems the trained model just simply outperforms the "original" images
can't wait to get my hands that completed model 😄
And i'm sure not being trained on useful tokens is that important -> the model just needs those nudges for extra creativity 😄
Thanks, it does take quite some time and efforts, but people download and use them, me included, and the new users are usually super happy when they get their ping, so it does seem to be nice for everyone as a week-to-week thing already, but I'm also awaiting on that bigger one, with tokens that mean things and not just "bnp or vedi"
well the only concept that is correctly trained and controlable usually is the main one, "SDArt"
so a little more control for people is still nice ^^
if i saw the hugging page right, it's a 2.1 model?
both
I train both now
I've been doing it for 3 models
there is a 1.5 and a 2.1 base
... you have an A100 for yourself now? 😮 that takes ages!
If i created a character lora that I like, is there a way to continue training off that?
takes around 2 hours to train both models, and a full day to upload the 43GB that means to HF and civit
I'm on 3090TI
https://huggingface.co/Guizmus/SDArt_AliceInDiffusionLand
https://huggingface.co/Guizmus/SDArt_AliceInDiffusionLand768
because of how HF works, I make it as 2 models though
Looking for a used 3090 myself now
i just need the more vram, 10gb is coming to bite me so much 😦
❤️ much love ❤️
I've been on it for 2 weeks already but it's quite some work
uhoh. the hatsune miku sakura song is blasting in my headset again, am i forced to do couples holding hands again? 😮
4 seasons?

if done with combinatorial generation, fixed seed, and prompt editing [season:0.3] you can get the same base image 😄
I'm using wildcard 
if too much work, can i have the prompt @quick hill ?

sorry, scrubbed 😦
my prompts sure do love... love 🙂
well there IS snippets which have "heart to heart" and "face to face"
I KNEW IT
adding some extra random shit for extra flavour makes my random lyrics prompting even better!
as said. my model has a REAL unhaelthy fascination with doors.
Greetings chat
good mornign ❤️
nya good morning
OH SHIT this prompt 😮
oof... prompting that on sd 1.4 (base model 1.4) is... rough
ezz
it... just fails
#1 tho. I can't believe they didn't just flatout merge (knowing SD shenanigans)
retried the prompt, ti's a real lucky hit
ah making a character from nothing and having it be consistent and then creating datasets for an embed is fun
also making him smexy
hailo
yeah

you still running five line prompts @wispy canopy ? 😛
no, this one was actually very discrete
nice
yes, it took me a WHILEEEE to get the separation going decently
this was the end result for the COW submission (after some tiny bits of photoshopping)
@cyan burrow this song as prompt always cracks me up. When the line With a crash of thunder a new demon appears Unfurling it's rotten wings It lets out a petrifying scream, the pretty monsters come out
Prompt ? 😶
professional lighting, (cinematic lighting), (nostalgia:1.3), professional (landscape painting) of a (giant mysterious monster) destroying a city, city in burn, building collapsing, crash, (super giant monster), ugly monster, catastrophy, storm, (angry monster), dark clouds, dark sky, giant monster in the horizone, expressive style
With a crash of thunder a new demon appears Unfurling it's rotten wings It lets out a petrifying scream
BREAK Light speed in 3, 2, 1```
Thanks ! Its crazy the monster looks like fenrir 😂
no, it is totally random
, like this
Is this a model or a style?
Bit of both, but mostly style lora
(Model is Cetus-Mix 3.5 with Ligne Claire Anime lora
Do note that this lora destroys hands pretty bad sometimes
Ah I see it in those two
I thought it would look really nice on 3d models so I want to try it
Never it tried on 3d yet
Im too lazy to try and make a shader
I was just going to project it and see how i feel after
let's see if I can train an embed on my character
That’s amazing @molten shadow
training embed, 3 epocs, first try, wonder how it does
why is there dutch language still left in my model? I entered vogel (dutch for bird)
hmmmm...
Vogel im kafig 
when uwu and :3 enters the prompt, you know it's serious business

i <3 u,chibi, outline, uwu, :3, rawr,
i just put "2girls, nightgown, embracing, hugging, noses touching, bed"
ok fr now, what drugs were the makers of Magical Destroyers on?
dwugs? big boom
prompt?
grey hair, red eyes, long hair, flat chest, solo, outdoors```
nothing special
i'm returning after a bit of time
@lyric zealot Soosi, rate COW :O
Ayyy 

Also you not gonna do COW? :O

maybe later if i'm feeling up for it
Fair xD

I mean chaos and order is a tough one :o
model?
I had to use controlnet and latent couple, and I never use those xD
anyone have a model I can use to create male characters? anything creates ridiculous cartoonish stuff when I try to prompt for a middle aged blacksmith dude.
its overtrained for uwu and sex fantasy
Uhhhh
i'll check it out, cheers
any tips on how i could prompt Anything to co-operate with me?
Uhhh, not sure :o
like bruh, really?
oh lol
its my favorite anime
I started it, but then I dropped it, forgor why :o
prob start as slow
I think I remember it being slow at the start ye :o
Maybe ye :O
(sle:1.2, detailed background),mountain fuji, field of red spider lilies,(graveyard), aesthetic, background, wind's howling, smooth, soft, god_rays, volumetric,
86 -eightysix-,vladilena millize, solo,white hair, long hair, officer, white eyes, cute,
<lora:vladilenaMiliz86Eighty_v20:0.5>```
Spider lilies, thats a pretty cool ane 
is there a prompt for the hands to be out of the picture like this one?
I have no idea for the hands neither 🥹
arms behind back with some weight can help
Or hands in pockets
Or arms behind head if you want something spicy 
or arms in negative prompt if you want something different :p
How considerate of them
name?
thanks
oh XD
😂
uhoh. waifu stuck
Why is it bad? Also, how'd you do that?
controlnet: anime line art
how do you use it?
you need update controlnet first
download the thingy
and ez
ah update the ui?
the controlnet extension
where do i put this archieves
archieves?
thank you
oh
what i need do exatly ?
you download the pth file
put in \extensions\sd-webui-controlnet\models
and the yaml file should already be there if you update the extension
Hoshimachi san?
prompt?
You need ControlNet + Openpose + charturner + wildcards(for random)
Openpose pose from civit lekima_turnaround
ooo damn good idea
i was trying to do a sheet for a while
thanks 👍
this is how the prompt looklike can't paste the text here or it will call underline function lol
did you try using double ` around it?
__color__
cool
ahh it works
just throw in the txt file lol
lora:LoconLoraOffsetNoise_locon0501:1.4 < can ignore this line but you won't get white background consistantly
yeah you can also do that
yeah using DPM not Euler A
doesn't need that much steps
also use Hires.fix 10 step so its actually 25 +10 steps
not really
well technically yes
but practically no
25 step for normal cutting, 10 with lower noise for upsacling to keep the detail
hires fix step does less than normal steps
well you can technically have hires.fix with full strength at 1 then it will make your previous step almost unrecognisable
the second look sus
Civitai is down
sadly





















