#🍥|anime
1 messages · Page 20 of 1
The more I use Aurora the more I'm loving it.
same
I love how im making cute chibis ina eating foods while another smile making semirealistic
How?
Mainly prompting
What is the prompt?
((masterpiece, best quality)). a drawing of a man in a suit covering his face with his hands, 1 8, color page, intricate heterochromia sad, kaneki ken, manga front cover, sui ishida art manga, tokyo ghoul, manga cover art, full color manga cover, by Takeshi Obata, official art
I use the replicant model
Color page
Never used it
Probably good
Also yeah I can definitely see the Tokyo ghoul vibe
I saw gabimaru
Well they have resemblance
16!
Also what is "replicant model"
A model I found at civitai
WD1.5-beta based model.(Stable Diffusion V2.1-768) Negative Embedding & more info: https://huggingface.co/gsdf/Replicant-V2.0 Licence: https://...
Oh
Can someone tell me whats the difference with anything 4.5 and anything 4.5 pruned?
Less weight. Only real reason to use non pruned version is to train on it.
Also merging
Did you know there is a cute ass animal named pika?
A pika ( PY-kə; archaically spelled pica) is a small, mountain-dwelling mammal found in Asia and North America. With short limbs, very round body, an even coat of fur, and no external tail, they resemble their close relative, the rabbit, but with short, rounded ears. The large-eared pika of the Himalayas and nearby mountains is found at heights ...
Why are there so many people here but little activity?
Cute
This one is awesome! 
I haven't been generating much, a bit burnt out 
thank you 
I'm almost burnout, wildcard/dynamic prompts helping me alot 🥹
I almost had the Niji v5 style ready
maybe its time to try and create a big wildcard
what model you use to make this pixel style?
(SLE, Detailed Background), (pixelart), masterpiece, best quality, haunted theme park, haunted by chibi ghosts, cute, whimsical, glow, glowing, fun, silly, mystical```
(KHFB, AuroraNegative), (Worst Quality, Low Quality:1.4), skimpy, nsfw, grayscale, realistic, watermark,
Prompt
niji is too good 
jeah it's amazing for sure
no u
I've been meaning to ask, what does SLE stand for and what is the token do?
Without SLE
With SLE
SLE = Eyes Improvement
Stands for SweetLuna Eyes
IDK what to called it

ty 

Where is this absurd wholesomeness coming from
o god
Shes got that Canon tube style camera
yeah it just came out randomly lol
yeah but i love surprise haha
Fair XD
CFG scale 1 for surprise 
more like thing you'd get from after putting it in mixer
i does give more misc. details at lower CFg but not as refine
Tis at 3 CFG :o
burglar woman
soooo cute! looks like ririchiyo from inu x boku 

thank you 
dont ask me to put in low cfg again pls
i will never ask again dw 
the cfg limbo is real...i've been trying to get a crispy chibi
not as in fried
but....sharpened
Unparalleled creativity
1 cfg seems kinda cool on aurora tho

i've been meaning to try aurora
ik 3 is fun
gonna hav nightmare
but not 1
It best model 
dling right meow
Try making a turtle with low cfg
Tis 1, its kinda scuffed, but its kinda funky, I vibe with it 
ok 3cfg look fine for me
Its kinda like, low-res aesthetic pic vibes
Low res low cfg is the new meta
Reminds me of this
#1072013871730131004 message
actually got more weird pic but didn't save it lol
prob because 80% I merge are mostly from artist like WLOP/Nixeu/Guweiz ETC 
@quick hill 
chibis for u, Vil
I just love this image so much
That a penguin? :o
EAT
go have some sweat
this is soooooo cute
Look like dog to me 


🐾
I make peech
Oh god my wifi
Why aren't any pictures here loading 💀💀💀
i hope they're happy 4ever
wifi diff
blame jg
Other channel's fine
too tall for Klee
Mikoo 
its fake miku (stare at boobs)
@finite herald Can you guess her mood?
Is miku flat? 
I think so
She is bing chilling
I forgot to change seed but oh well it was different I guess I did previous image before I disabled xformers
Smaller booba mikoo :o
MikOO > Mikoo
I haven't tested my model on Miku yet
Mik..
Miksus
hum not very diversified on what I can Miku in there. I haven't captioned her enough

Punt it
Loving the are smol

Loving the cyberpunk esc vibe you are using
Too much cute. I died 
does more training images typically mean better results for your loras?
ˢᵐᵒˡ
I was supposed to take quick nap and go grocery shopping

Quality is more important than quantity in most cases. It can depend on how complicated the thing you want to train is though.



🐢
1h of gens with same seed as earlier hour 
Oooooooof™️
Also I think its time for me to make a new PFP, new role, new pfp me thinks 
dog.
Did you use that as prompt yet?
hahaha


My old merge _ANE. Here's link if you want it
https://huggingface.co/Surficialruby/RandomMixes/tree/main
@quick hill
does it have dogs ?
Why you pinging Luna, you wanna expose me?

I kept on testing my model on miku. I need more mikus in there I think (jk, I need to finish captioning the mikus)
Yes
😭
I was just talking today with Luna that people confusing my old model to new models
Which ones?
Any5 and now Aurora
It does actually look like anyv5, way more than aurora
4 months old merge compared to any5 I'm happy with it
Well anyv5 is just like anyv3
aurora model + schizo prompting + <lora:Polychrome_v1:0.7> <lora:LoconLoraOffsetNoise_locon0501:0.5> <lora:kyoto animation_artstyle_LoCon:0.5> <lora:style-akai_sashimi:0.7> <lora:lineless:0.5>
What is lineless Lora?
kinu_bari makes art without the typical black outlines
That one is from the artist that lora is trained on.

the lora can make stuff that looks basically exactly like that though
polka dots and everything
Yeh, I saw some posts from the person working on it and it looked very close.
I wonder what if I try with chibi prompt
heres a crappy gen but really shows off the dots
anyone ever had good luck with crossed arms? I keep getting stuff like this
O.o
Cool
3/4 turned out pretty good aside from the fingers.
I figured out I had ((stretching)) left over in the prompt and it was causing all my issues 
been there
do you have a prompt for this one?
masterpiece, best quality, high detailed, 1girl, teeth, plant, monster, nature, sharp teeth, glowing, solo, checkered, skull, horror, outdoors, looking at viewer, open mouth, red eyes, skeleton, glowing eyes, palm tree, leaf, feathers, vines, fangs, sky
thanks!!
looks very friendly for me 
Familial Love
that definitely is a vibe tho
Maybe I should go over to the horror channel 
Tiki monstergirl 
looks a bit like lain
Since your input is already processed you need to select None under Preprocessor. The preprocessor is for processing a regular image.
Okey thanks
Bro using god mode
I think ai it's just getting insane
i'm struggling hard with hands, what models you guys are using?
No controlnet, just regular text2img. Adding claws to the negative prompt helps improve hands quite a bit. Since I was using claw pose in the prompt I used (claws:1.4) in the negative. Model is AuroraONE.
oh yeah i didnt know theres claw
gonna remove that 

I'm jealous

😢

🐢


How about this one? 

OMG DADDY? MOMMY?
idc please kill me
bouncy !
I'm captioning the PoW pictures of the last 14 events right now
And what don't some of you make me write ... (REGRET) :
a painting of an anime burger girl, vivid color, masterpiece
wut xd
Any New So-Bin models out?

oh ahah 🤟
What is going on
bounce party
😮


loved this one (not from me, RegretGT)
This is going into my most hilarious chats archive
Lol
League of Legends
That too
One more for you 😄
Has to check if he had a pokeball earring
Naaaah bro
Chibi asuna?
You cant just drop the cutest fucking Asuna chibi ive ever seen
How am i mentally supposed to handle this...

I bet they could probably get asuna in a teacup or Ramen bowl
😭
what should i use with control_depth
AuroraONE can make harpies 
She's not willing to get inside the tea cup
Tea cup too smol. Gotta make it beeg.
I asked for massive tea cup but got something else
massive c cup
Didn't end up finishing this last week like I planned, but.
Inspired by this song https://www.youtube.com/watch?v=ZOa-VO2_SLg
KOKIA's official channel → https://www.youtube.com/channel/UC7Nj69kzQ3ARDEuetOibj6A
フクロウ ~フクロウが知らせる客が来たと~ (Búho ~El búho informa que ha llegado un visitante~)
Album: tontonton doubutsu no ongakkai [2018]
Artista: KOKIA
Letra Japonés: http://kokiasky.lofter.com/tag/どうぶつの音楽会
Traducción: DuendeMadrefoca
Image...
I guess maybe I should start using that clip fix
aurora clip if I applied it right earlier
beeg wings
woa
Tea cup is becoming with her!
hugging with the cup now
guys, i need help, my gens' eyes keep being asymmetrical, any tips?
For me usually upscale or a bit of inpainting can fix this
Previous was aurora clip this one is just clip fix
With 0.3 - 0.4 denoising strength
styles change too
Yeah
😄
Give me one you greedy lady
that's how i abuse it
And this is once again that broken clip
Clipfix is much better here
Broken vs fix
Broken
Goofy finger face
Broken as broken clip
Broken as broken better
Yes this is masterpiece
Smash
Oh my gah
Looks like Zoe from League.
yesj
?
Power
anime community, oh how you never fail our sacred community. I humbly approach you all today to ask a simple favor
a gentle wish...
does anyone know where this girl/song is from
well the girl is ceres fauna, an hololive vtuber, don't know the song sorry
TYSM!!!
@finite herald https://www.youtube.com/watch?v=0RMVJTLZOzQ
Let Me Stay Here
༻━━━༺ .⋅ ✧ ⋅. ༻━━━༺
Composition/Arrangement: Shibayan
Lyrics: kidlit
Vocals: Ceres Fauna
Logo: Spiritsnare (https://twitter.com/Spiritsnare)
Illustrator: a20 (https://twitter.com/atwomaru)
Music Video: awa (https://twitter.com/siinatsu)
Translation & Captions: Gomadare, arc_, T-chan
༻━━━༺ .⋅ ✧ ⋅. ༻━━━༺
me rn
tysmmmmm
!!
now i can bop about
pretty much any hololive song is a bop 
now i can stop overplaying bao's songs! mwehe
Wtf why i never heard
has no idea what hololive is and cannot name a single vtuber
ggwp
Anime pow game night when
No I just did reverse image search for nothing
@finite herald you should learn to do that yourself
u should fite me
omg


wen im not goldfish
Just wanted to use this gif tbh
Bro
it jus had to be me huh
Yuki installs gentoo, a shitpost made by me
video taken from The Melancholy of Haruhi Suzumiya episode 13 "The Day of Sagittarius"
₿💰💵💲Help Support the Channel by Donating Crypto💲💵💰₿
Bitcoin
3MMKHXPQrGHEsmdHaAGD59FWhKFGeUsAxV
Ethereum
0xeA4DA3F9BAb091Eb86921CA6E41712438f4E5079
Litecoin
MBfrxLJMuw26hbVi2MjCVDFkkExz8rYvUF
Dash
Xh9PXPEy5Ro...
all i have to say to that is bloopbloopbloop
Sus quote
Sus indeed. Wonder who wrote it 
🐟
🐠

sus

real
I should do that
fruit does as fruit does

how are you making such beautiful mikus before my very eyes rn
Just need to find that NSAKEY
these are so cute
what's the prompt/model?
Woah... These are beautiful! Especially the last two. Also, welcome to the community!
Beautiful.
I've been trying to get a more experimental LoRA to gen close stylistically to original artist-ish. It's a lot of fun--it and Holokuki.
Hey i got a lot of those
someone got some negative prompt suggestion for me ?
What are you looking for? Big impact, little impact, nuke? What style are you going for?
Sakura
(lowres), (cropped), (worst quality), ((((low quality)))), (((normal quality))), ((jpeg artifacts)), ((bad anatomy)), ((bad hands)), error, ((bad text)), ((missing text)), ((missing fingers)), ((missing anatomy)), ((missing hands)), ((bad fingers)), (extra digit), (fewer digits), ((bad detailed)), ((missing hand)), ((bad finger)), (((missing finger))), ((bad hand)), (bad legs), ((missing legs)), ((bad anatomy legs)), ((bad objects)), ((missing objects)), ((bad items)), ((missing items)), (bad environment), (missing environment), ((bad building)), ((missing building)), ((bad reflection)), ((missing reflection)), ((bad background)), ((missing background)), ((missing eyes)), ((bad eyes)), (bad geometry), ((missing geometry)), ((bad line)), ((missing line)), (bad light), (missing light), (bad shadows), (missing shadows), (bad wrist), (missing wrist), ((bad metal)), ((missing metal)), ((bad wood)), ((missing wood)), lowres, text, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, ((panties)), ((nude)), (((nsfw))), blurry, (more than 5 fingers), (less than 5 fingers)
That'd fall under 'nuke' category, lol.
thanks i'll get going with that 🙂
my neg prompt stayed home >.>

If it works, it works 
What model did you use?
aom3
Yeah, I've seen that the AOM models often need to be nuked to get away from the trained style. It's a great series of models, but I generally prefer more flexibility or just switching to a model that's closer to the style I'm looking for with a lesser negative prompt influence. I seem to get less artifacts and sharper images that way.
wow dragon looks cool
any recommendation?
These days I'm leaning more towards Aurora or Based64mix-v3. But for specific things sometimes I use 7thC, AnyPastel/Pastelmix, or a custom mix.
i never heard of any of these, damn
Aurora is popular within certain circles here atm. I'd start with that one. Very versatile.
yes 
The LoRA is stan brakhage with the Holokuki model, believe it or not. I finally found the sweet spot of usage for it. It's amazing how versatile it can be depending on the CFG from 5-11.
HI!!!
i took a bit of a break

hm, anime models dont know what "among us" is but whatever it thinks it is is kinda cool
doing well! actually excited to use sd again. i got burnt out for a while and been busy all the time. how have you been?
yay me too, @quick hill 's aurora1 model bringing me back
same here 🙂
working a lot since january, came back for a week in april
now i'm in the USA and i brought SD with me on the trip 🙂
Welcome to the US 🇺🇸
What's in the box?
thanksu :3
well i was here in january for a week
Canada > USA > S.Korea > Canada > USA
Yae Miko is a pretty waifu 
cactus waifu
she has returned 🙏
I missed her 
💦
Boy this inpainting is not happening. It's been how many months and we still don't have a properly working inpainting system? 
its like half pixel art
maybe try this as a pass afterward: https://github.com/AUTOMATIC1111/stable-diffusion-webui-pixelization
see if it does some good
So much for my going for a meme response. Inpainting just plain doesn't work. Not sure what's going on.
seems like the same box XD
THAT'S THE PROBLEM. GAAAAH
now show hands Kappa
makes me think of dune
apparently redo of healer
just looked that up sheeeeeeeesh
rest in peperonis 
im into it
what did you use to make this?🥺
Here's the PNG info: (chibi,pastel_color), (1girl),(blonde hair, short hair,blue eyes, golden_crescent_(ornament_hair), opened_yellow cardigan, white collared shirt, buttons, black pants, black necktie,) blush, looking at viewer,smile, closemouth, :3, CLOSE UP, facing viewer, straight-on, leaning forward, mountain background, grass, field of flowers, sunset <lora:cutescrap05v_cutescrap3:0.7>
Negative prompt: (KHFB, AuroraNegative), (Worst Quality, Low Quality:1.4),(claws:1.4), nsfw, grayscale, realistic, watermark, grain,
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 6067843, Size: 600x800, Model hash: b36d9ac70e, Model: AuroraONE, Denoising strength: 0.51, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: 4x_fatal_Anime_500000_G, Hashes: {"vae": "df3c506e51", "embed:AuroraNegative": "e5b5173e82", "embed:KHFB": "adac8597da", "lora:cutescrap05v_cutescrap3": "0c43df6211", "model": "b36d9ac70e"}
don't know the model still but thanks love ❤️
"zero two"
r u using EasyNegative? i find that EasyNegative works better with AOM than other models
also, ur probably missing flat color or something that triggers that style for aurora
oh it might be chibi
I gotchu there too https://civitai.com/models/40199/aurora
i'm using someone's negative from here
Yeah, if I used the same nuke neg prompt on different models, they'd likely act very differently. You've nicely compensated for AOM2's synchronicities with your prompting. That much sadly won't work on other models as well. It's always best to prompt for the model/extensions you're using.
i don't remember who
What's the prompt here? I've noticed that Aurora generally does need or like as high of emphasis on tokens and have gotten some weird gens when I have emphasis too high on parts.
aw ur so sweet tysmm
is this where the cool kids hangout now?
Read? On the internet?? 
lol
Guess I have been missing out haha
Well, welcome to the party. Good to have you
iv been playing with pastelmix and its got a super cool style
you say "nicely compensated" but really, i've no idea what i'm doing
the prompt is the following:
(1girl:1.25), (absurdres:1.25), (anime:1.25), best quality, (high contrast:1.25), (oil painting:1.25), (RAW photo:1.25), (ultra-detail:1.25),
by artstation, by midjourney,
(beautiful girl contrapposto:1.35),
(cinematic ligthing:1.25), (cowboy shot:1.25), fantasy, (magical), mystical, (vanishing point:1.25),
(crystalline orange hair:1.15),
(bangs:0.85), short hair,
(colorgradient yellow eyes:1.25),
(beautiful eyes:1.25), eyeshadow, (looking at viewer:0.85), large eyelash,
(gorgeous appearance:1.35),
(bare shoulder:1.25), (collarbone:0.85), (tonemapping:1.15), wasp waist,
(playful expression:0.85),
(defiant:1.15), confident, closing mouth, smirk, (spellcasting:1.25),
(colorful:1.25), gothic lolita, (lolita fashion:1.25),
eldritch, (full-of-detail:1.15), (majestic:1.15), (mystical), royalty, supernatural,
(wearing a cosmic gown:1.35), (wearing a cosmic skirt:1.35),
(bridal glove:1.25), dress bow, (dress embellishment:1.15), dress flower, (frills:1.25), gold trim, (voluminous frills:1.05),
(wearing a colorful shawl:1.35),
eldritch, majestic, royalty,
(capelet:1.25), lace trim, mosaic gold-pattern,
(wearing a head veil:1.25),
gold headdress,
(elegant jewelry:1.35), (intricate-pattern accessory:1.25),
magical appearance, shimmering,
gemstone, (gold accessory:1.25), (hair flower:1.25), intricate necklace,
(surrounded-by-fire:1.35), (space:1.25),
desolation, discovery, freedom, (journey:1.15), lively, (mystery:1.15),
beautiful glow, (burning sun:1.35), fire print, high contrast, meteor shower, (moonlight:1.25), (pastel hue:1.25), pitch black,
sorry for the text wall
😅
bro i just use one sentence
Early days straight from NAID central. Haven't used their service in a while though. So much goodness here.
my prompts usually aren' that long
is just that is interesting to see how far you can go
flying dude
Never heard of that airline
this is a selfie from a multiplayer lucid dream
I don't see anything in particular on a quick read through that looks problematic, but it could be having some problems with some of the repeated words, especially some of the ones that are repeated within different blocks of increased emphasis. In general Aurora will listen to prompts better than AOM in my experience, but that also means that some of the things that you might be used to doing on AOM will make images look worse on Aurora. If you prefer AOM, that's perfectly fine too. Just trying to offer tips on how you might be able to get Aurora to work better for you if you want to use it.
Would be nice to know how to use Aurora, I've the same use with kenshi as well so learning different checkpoints would be really good
I'll try less tokens
What Nubbs said. In essence, different strokes for different folks, different prompting methods for different models.
Yep, I try to start simple when I'm using a new model and build off of it so that I can figure out how the model responds to different things. Long prompts make it difficult to figure out where problems are coming from because there are just so many things that it could be.
This is the way.
In fact that's a great way for testing all networks.
how do i deselect a hypernetwork or lora in the quicksettings drop downs? there isn't any none option
I'm not sure if you can. I don't use the quicksettings for those. I only use the cards menu for them.
i'm helping one of my friends get into stable diffusion and she didn't know about the cards menu
i showed to her now though
worst case, refreshing the web page should clear the selection for them I think.


can you set a default weight for loras when you add them from the cards menu?
like you select a lora and it inserts into the prompt with like 0.7 weight
Yeah it's just like prompt weighting
how much denoising strength is good enough?
on hires fix? i usually use 0.5
Usually 0.3 to 0.4 for me
Since we have a few new names in here, I'd just like to spread the word about the stealth png info extension. Discord removes metadata from images normally. This extension bypasses that which allows you to share your prompts easily with other people who use the extension and also allows you to view their prompts https://github.com/ashen-sensored/sd_webui_stealth_pnginfo.git
Try experimenting which one suits you best
i use meina v9 and counterfeit wothout any upscaler, the results are blurry.
0.35 x2 then 0.4 x2 =x4
what is a good upscaler?
For Hires fix I almost always use 0.51 denoising with either 4x_fatal_Anime_500000_G or 4x-AnimeSharp upscaler.
where can i get that?
I do find that if you're going to use Hires fix, it's important to have live previews turned on so that you can see what the image looks like before being upscaled in order to determine if your denoising is too high because things like bad hands can often be a result of slightly too high denoising.
Yup
For anyone who's system specs struggle with live previews, you can set this to -1 so that it will only update when the low res pass is finished and again when the high res pass is finished.
what's the quicksetting for clip skip again?
CLIP_stop_at_last_layers
thanks
what does clip skip do?
@carmine bobcat More importantly, where can I change it?
CLIP is the part of the model that interprets text and it's split into 12 layers with. Later layers get progressively more complicated. NovelAI found that by skipping the final layers, it usually improved the quality of their generations. They shared this information with the public and since then, pretty much anyone using anime models have done it. TLDR: set it to 2 and you'll probably get better and more coherent images.
settings > stable diffusion
didn't there used to be a "close" button here?
There was, now you just click the same button you used to open the menu.
i liked the big close button, it was bigger so i didn't have to aim for it as much
True, it was also less confusing
Spent like 15 minutes trying to close the menu the first time the change happened
it was also central
I just started using this fork and it's interface is much better https://github.com/anapnoe/stable-diffusion-webui-ux
@supple raptor 12 layer are for all models or only 1.X ?

What that fork got? Besides the nicer UI 
I believe it's only for 1.x models. 2.x uses OpenCLIP I believe.
is there a way to convert to it or do i have to do a whole new installation
There's probably a git command to convert, but I just did a fresh install. I don't know how to git that well.
thx i guess it too ..
i'm gonna stick with the regular ui for now since it's more supported
Mainly just nicer UI I think. I just decided to try it on a whim after Veggie recommended it.
Fair fair, do all/most extensions work fine on it? 
I haven't tried many of them yet, but none of them complained when I installed them. The main benefits I've found so far are a much better interface for the cards (it even saves your location in the list when you collapse it) and the ability to easily add any setting to your quicksettings list by clicking a button on the right side of the setting in the normal settings menu.
do u always use hires fix?
Oh thats pretty neat then, could try install it and then symlink everything :o
Yeah. If I had a slower card I would generate batches without it then rerun the seeds I like with it turned on, but my 3080 12GB goes fast enough that I just leave it on all the time.
That's what I did. I had to reinstall the extensions though because I got errors when I tried to symlink them.
Fair fair :o
do you get blurry result without hires fix checkmarked??
@supple raptor do you get blurry result without hires fix checkmarked??
It's definitely more blurry without it than with it checked. Giving it more pixels to work with resolves a lot of blurriness problems.
i guess its not just me problem then.
i tried using older model elysium anime v3 without hires fix. it doesnt have blurry problem
WebUI also makes it look like more of a problem than it really is because it scales images to fill most of the screen so small images often get stetched while large images often get shrunk
is there webui alternatives?
There are a few. There's a ton of different forks of Automatic1111's WebUI with varying differences between them and the main fork. There's also InvokeAI but that is missing some really big features that make it not worth using in my opinion. ComfyUI is another really good option, it's a node based UI but it is pretty complicated.
what upscaler should i use for hires fix? i've been just using latent
Either R-ESRGAN 4x+ Anime6B or 4x_fatal_Anime_500000_G
Latent is the worst for upscaling imo
I like 4x_fatal_Anime or 4x-AnimeSharp. Realistically the differences between the GAN upscalers is pretty small in Hires fix because of the denoising that happens afterwards.
Latent is usually pretty bad, but Latent (anitialiased) can be good with the right settings sometimes.
some models do better with latent during hires fix
this looks good what model and prompts for did you use for
this?
that looks like meinapastel or som similar mix
this one is really good https://u.pcloud.link/publink/show?code=kZgSLsXZ0M1fT3kFGfRXg2tNtoUgbSI4kcSy
These are different dimensions too, which may have played a role in the missing mouth.
Unless you just cropped them and didn't happen to crop at the same AR. In which case you can desregard that.
where do i put the downloaded upscaler?
they were the same parameters, just cropped
.\models\esrgan
im getting good results from R-ESRGAN 4x+ Anime6B:
left is input (which was also generated), right is upscaled and outpainted
Unless you got one of the other upscaler architectures in which case it would be .\models\whateverarchitectureyoudownloaded
what upscalers are good for doing anime?
Yeah, that one looks good. R-ESRGAN upscalers are just a lot slower than ESRGAN upscalers for not very much benefit when used with Hires fix, due to the denoising changing the image afterwards.
Some of us just listed the ones we use starting here #🍥|anime message
first image i got right with aurora
can't seem to find 4x_fatal_Anime anyone got the link?
what of this do i download, and where do i put it
just the .pth file and it goes in .\models\esrgan
Thanks for reminding me of this one. I lost it a while ago and couldn't remember the name of it.
thanks
2b
@pulsar token what is that stable diffusers role?
basically server admins not officially part of the "dedicated community team" but we help out with stuff out here sometimes
duo monitor desktop backgroun
I sensed my girl was posted
There will probably be more
kinda a leftover from times past though lol. We used to only have 1 dedicated admin role way back but then the community team was made official in a sense so the rest of us up there got to keep the old role and the actual community team took the sleek new red one haha
I kinda like the purple tbh though, fav role color

them apples
I think this WebUI might have a few issues reading LoCons @supple raptor
I mean, it seems to be working I think?
Thats my Sabu V4, and it seems to be applied
Did you install the locon extension?
I did the LyCoris one, lemme try the LoCon one
to be fair, I haven't tried any locons yet. I also got the lycoris one.
Fair :o
Maybe the LoCon one does it
I think the lycoris one works differently. Not sure if they need to be in a different directory or not.
debatable? seemed to start trending into realism a bit more here
Yeap, LoCon extension fixed it @supple raptor 
It's probably closer to realism than anime, but I for one welcome it here.
True, it looks amazing tho

I prompted SD to make 2b looking in a mirror, and it's showing the backside of 2b and GOOD LAWD (can't show bc of tos probably)
Thats a really good promtp then 
or maybe a good model 
It's something my team is working on, trying to make stuff a bit easier to make pretty haha
SD needs to tell 2b to chill out
@magic wave image browser doesn't work on the ux fork 
Could probably pull an earlier commit of image browser to resolve it.
Miku Smasher 
This 4 second clip took me 4 hours to render on a 2080 😅
Neat.
Pink 
O, fair, good thing I've never used it xD
You must not gen remotely 
Though why does it save into folders like this now? 
I don't xD
ok I think we are back to the right territory haha
I just go into my img folder and scroll around for pics I want
"Computer~Sama" eh? Getting a head start on pleasing our AI overlords? 
It's an option in settings. Something about save in subdirectories. Turn that off and it'll be like auto is by default.
True too! And that is pretty neat 
Yes XD
I've had that name for my PC since I was like 14 tho xD
Oh, my auto wasn't like that by default 
Maybe got added recently 
Also I feel like I should do a egirl pastel pink webui palette, but I wont use it 
Can someone tell me if bikinis are allowed?
From the pin: Rule #4. Absolutely no NSFW
Nudity, excessive gore/blood
Anything that exists to suggest sex including sexualized poses
No illicit imagery - this includes NSFW content involving or depicting underage characters such as schoolgirls, lolicon, etc.
Sexualized poses, or borderline underage sexualized characters in particular cannot be shared in this server. Heavier actions such as time-outs or bans will be considered on repeated offences.
.
Doesn't say no bikinis necessarily.
Dear god her hands
I'm gonna just be safe and not generate 2b in a bikini, I would imagine it would break tos
Instead I'll generate cat ears
Why easynegative is so important...
Bikinis aren't inherently a problem, but it will depend on the image. Cameltoe, nipples visible through the bikini, excessive cleavage, or sexualized poses wouldn't be allowed. If you feel one is safe but aren't completely sure, you can spoiler it and self report by reacting to it with the ⚠️ emote which will flag it for mods to look at and let them know that you reported it yourself. That's not a free pass to post whatever you want, but they are pretty good about just letting you know if one needs to be removed as long as you posted it in good faith.
So not to encourage posting anime girls in skimpy clothes but you should be able to tell when you look at your gen whether its sexual or not
Yeah, given how sexualised 2b is, it's safer not to
I was going to go for a beach day vibe
That's fair, she can be a bit tricky to keep safe.
Clyde is amazing....👀
managed to get a sfw beach day
/faq
This aint doki doki literature club




















