#🍥|anime
1 messages · Page 43 of 1
D-O I-T
Time to bribe Fruit with Mikoo 
We gonna be Miku posting? 
pick a 5th channel tho 😄
https://github.com/KohakuBlueleaf/a1111-sd-webui-locon It needs this
guys for hands just upscale then inpaint only masked
Yeah I haven't had much luck with locon either but need to just keep playing around with it

The hands are unaffected tho, so it might not be working
New bot doesn't like anime style 😄
i should probably reinstall my webui
is it bad? 
Apparently :O
Maybe?
?
I've always had good luck with the badhands textual inversion
might need to add neg prompt as well
it's not fair that everyone here has stable society role except me
not everyone, no worries
everyone
Lets keep or anime style here 😄
i'm just a member
no no nooo, we are just active here rn 😄
We have no life, thats why
like at least like 90% of the people who have been here with me since the redacted leak
check my message count on this server
Openpose hands is good for hands too apparently. But from what I see, both it and the lycoris model hand and finger proportions don't match the characters
almost 100k 😄
then look at even @magic wave message count
Y u hurting me in the hart ?
you exaggerating
U R not alone 🫂

I have 6.6k messages :o
I don't have many 🤣
Fair :O
blop
Oh dear... Miku from the new bot is cursed 
No LoCon vs with LoCon
Worse tis time, probably cause there was already good hands there tho
( . )( . ) got bigger 😏
Also I got a good mikoo one 
If you click on envelope it sends img to you 😄
Yee, das what I did :O
Yeee :o
apparently the photorealistic style does better anime than the anime style does 
I want a version of this that would not be allowed here
JPG?????????????????????????????????????????????????????
saving space 🤣
minor asking for lewds in dms... sounds like a setup 😅
uuuh
12:19 AM to 2:11 AM, I made so many pics
tru
And none of them got through
Actually they're 28
But I mean, they're not even lewds, Clyde just wants to make me feel pain
You hurt my kokoro
ok time to dm a mod to see if I can send an image
XD
huh?
Sakura timeee!
pose?
sry not sry 😄
leaning forward, hands on hips
no pose cn?
nope, just skill
JPGing go brrr
like I said... just skill 
you always say that
jpg haters 
no, usually I say you have skill issue 
In this video, Stéaviñ shows you how to pronounce and use "JPEG" in a sentence.
Is there a word you'd like Stéaviñ to help you pronounce? Just add your suggestions to the comments below and don't forget to like, share and subscribe!
#EnglishByStéaviñ #EBSTutorials #HowToPronounce #ASMR #Shorts
same same
I pronounce JPEG as "gif" just to troll the "jif" haters.
gif is only valid way
Or pronounce JPEG and JayPeegi
Maybe we should take a cue from french. "Je PEG"
What about "Je Pegging"
This is the process of converting images to Je PEG format.
Make sense 
finally fixed my sound driver
@brittle jay to answer your earlier question. I'm using random seeds, just slightly tweaking the prompt.
so very similar prompts
Yep, I think I just changed the weight of a lora and added fang between those two images.
The pose is just specified in the prompt.
like the model has an inclination to reproduce it
this is probably getting asked a dozen times per day, but what's considered the best anime models today? people were going nuts for anything v3 but that feels like a lifetime ago, surely something better must've come along by now?
Only if you ask it to. It's a good thing when the model does what you ask it to in the prompt. It would be a sign of overfitting if I wasn't asking for the pose and still getting it.
yeah but like at least to me they feel too similar
like yeah they are the same pose but also a very very similar angle
and like all the stuff
That's just because I'm cherry picking the straight on images.
thanks!
this was the same set as that last one.
guys why is all my recommended only full of bocchi racist?
any model that they use on projectai discord server
arms behind back 
Thats for cowards 
embrace the deformation 
Exactly 
you... black... black is no.... ❌
wot?
when is civitai going to wake up
Bocchi speaking normally? No way
--Credits--
Model used:
Bocchi - のがみな [ https://3d.nicovideo.jp/works/td85893 ]
Chair - とりそば [ https://seiga.nicovideo.jp/seiga/im3206150 ]
Audio is from the Bocchi VA herself, Aoyama Yoshino :)
https://www.youtube.com/watch?v=-SUe07y9hNk
My toe still ...
That is pretty funny xD
Original Video : https://www.youtube.com/watch?v=FMQb4wcjk-Q
#bocchitherock #bocchi #hitorigoto #hitoribocchi #hitoribocchinomarumaruseikatsu #anime #animes #animegirl #animegirls #animememes #animememe
I get too many of these videos
Gotta make you suffer as well
You should watch Xavier: renegade angle
Ok last video on my video spam spree https://youtu.be/gjBFc5L6E1g
This a classic one
despair
why is she blindfolded


Also, @lyric zealot, theres uhhh, orthographic perspective, whats the other one?
The one thats like, a lil box?
why does everyone end their negs with a trailing comma?
Why not?,
What's the problem?,
True,

CIVIT IS BACK!!!
Is it? 
no idea 
Fair 

i'm trying aurora for the first time and i'm not impressed
i often get stuff like this

are you using the same neg prompt?
(nsfw, nude:1.4), (KHFB, AuroraNegative), (worst quality, low quality:1.4), border, grayscale, cropped
close enough (?)
just to check, you have those embedding right?
yes
I swear there is a name for the other one tho
might be because of your prompting style then
@minor olive
i got rid of full body from my prompt and i'm getting better gens now
oblique projection?
the eyes are nice
Actually nvm I think its orthographic
This kinda thing
Though the AI doesn't recognize orthographic perspective
Which is weird 
isometric might work as well
THAT WORD

xD
starting to get good gens now
https://youtu.be/abETp8Q6H8U?list=RDW9lsWI7zhTY Also cool song for yall 
Provided to YouTube by TuneCore
Venger (feat. Greta Link) · Perturbator
The Uncanny Valley
℗ 2016 Blood Music
Released on: 2016-05-06
Auto-generated by YouTube.
https://youtu.be/WbnHutA1u_0 Here with a music video, cause its cooler
Movie/Anime Name: BLADE RUNNER 2049 - "Black Out 2022" Anime Short
Music name: Perturbator - Venger [feat. Greta Link]
Inspired by ("Jack Box") Perturbator - Death Squad. Check it out Below!
https://www.youtube.com/watch?v=_26VnMsLg9U
*Disclaimer
This upload is only for entertainment purposes only. I try to promote the artists and their great ...
Also the girl in that video has like some of the coolest animation I've seen
wtf I just noticed, since when do you have that pfp?
since like yesterday or the day before

took me a while to notice lol

copying prompts really is the key to success...
Tru
How do you make black edges?
black edges? 
you meant like outline? 
And scratchy style
nice
Its called outline?

this is not anime!!!!
Thx
I used anime model + anime lora
i guess that will be outline, but the style you are talking about is more due to the model im using
for the love of god, someone here please be reading 7th prince
that hand is perfect wtf
just like octane render unreal engine stuff
(focus on eyes), absurdres, masterpiece, highres, cinematic, full HD, 8k, intricate detail, ultra detailed, illustration, portrait, landscape, (1girl), (((lucy elfen lied))), ((bandaged arm)), bandages, long hair, ((naked bandage)), (sarashi:1.4), solo, long hair, closed eyes, bandages, solo, pink hair, (((very long hair, floating hair))), (looking at viewer:1.4), ((upper body)), broken, lips, red hair, (((dark background))), black background, grey background, simple background, fog, Backlighting, light rays, (((eye focus))), (atmospheric perspective:1.6), (realistic:1.4), ((unreal engine, octane render)), covered nipples, hand on own chest, (wide shot), full body <lora:seal-000001:0.8>
Negative prompt: lowres, ((bad anatomy)), bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, (jpeg artifacts), signature, watermark, username, blurry, deformed, (bad eyes), (misaligned eyes), (bad_prompt_version2:0.8), squashed face, pixelated, light
Steps: 28, Sampler: DPM++ 2M Karras, CFG scale: 7.5, Seed: 1986507501, Size: 640x400, Model hash: 1b5f8211ec, Model: aurora_v10, Denoising strength: 0.6, Clip skip: 2, Hires upscale: 1.6, Hires steps: 20, Hires upscaler: 4x_fatal_Anime_500000_G, Discard penultimate sigma: True
yeah also realistic 1.4
this is actually an old prompt

smh my head
Which one?
same as 3d. looking at you, 2005 3d anime things...
I'm using expmix_line_v2
whuuu, spirits within was 2001?!
Thx
@finite herald your master has summoned you!
omg
fruit x miku colab
Dis so cute
Lowkey makes me think of Malenia from elden
Yeah?
that's the full prompt copied?


fruitmancer, cute chibi, market , lovely colors,inviting atmosphere , hatsune miku
No maidens?
my model isn't the best when it comes to known characters tho, and if i respect my own beliefs, i would want to delete them completely from my model, only to become existing when loras are used
with my quality prompts
Table hair
@finite herald there's mikus handling fruit over here!
helloo
sobbing at the club rn
fruitmancer club edition?!
haters will say its FAKE
fUUUU i pressed generate but left upscale on... now i have to wait for the oom 😦
thoughts and prayers
already pressed stop, still have to wait for oom 😦
be a chad and ctrl c spam your terminal and relaunch
but muh prompt!
you dont have the uhhhhh whats the fucking extension called where it saves your instance
reee cant think brain fried for fighting kohya all day getting it work on cpu
should I get a 160cm or a 180cm body pillow?
also what i do is i just keep the browser open, then when its fully relaunched i just turn the res up and down and hit gen, hasnt given me any issue slol
depends on the character's height
yeah, but still. i hate having to relaunch (but it IS the best way)
if megumin i'd suggest 160cm
I just want a body pillow first
yeah no 100% understand
I need it because it will help me sleep
takes a minute for the venv every time myself
like rn I just hug my cover
if you're not going to put waifu on it, get a big one
I may put waifu on it some time though
I think I'll get the smaller one just in case
NOOO out of 8 gens, i get 1 decent hatsune miku -> she has cameltoe 
what's the problem
the part where i can't paste here
does the bot block FEET?
No u 
depending on contrast and stuff
it is not meant to block feet
the bot watched too many quentin tarantino' movies
man
I didn't read anything other than that message right there
y u buly me?
y not?
i'm tired of getting good landscapes and having faces like this, what do i do???
higher res
send to img2img > tiled diffusion upscale + tiled vae
this is what i usually do, but they are not cooperating
buy me a 4090 bitch boy
got a spare 3080, come and get it!
ship me
now!
i am actually willing to give my address for this fair cause
sussy baka
i have 6

damn, my model doesn't like the idea of dancing i suppose
Or clothes, it would appear.
the modelname is alfahentai. that should say enough about the preferences it has...
when did you switch to jpg?
here?
left: ✋
right: chaos
thx crazykazoo... i did think it couple of times, but never really looked into it
I mean, can still train stuff with 10 gb
when you have more, you start using more
now i have an LLM loaded, stable diffusion, suno-bark at times
Right!? Lmao 😂
LLM's are real memory hogs though
hmmm.... seems my model is kinda lost when it comes to dancing, or hatsune miku... it's kinda lost in each and every category, unless your prompting style is random chaos... which is just my style
I mean, I don't need that much :o
yet.
i didn't need 10gb+ too
until i did
Wellll, fair :O
i wish i at least had 8gb but stuff is EXPENSIVE ASF
https://civitai.com/models/49865?modelVersionId=60356
pls go tryyyyy
Mixed Down is a WIP. I spend 3-4hrs a day on it merging and tweaking. V1.1 has over 50 iterations. Please enjoy and share your results
i was too lazy to upload more than 1 pic atm
wuuuut
?
Good c:
anyone else names their lora file names like this?
haha
go for used cards
got my 3090 for €750
im looking at a used 3060 for 450CAD
and i thought canada was bad
Miku feet time
wtf...
better than a €1900 4090
Rookie numbers
6 days ago, USD is 497 pesos xD
where is the feet

needs to watch more quentin movies
i hear Dusk Till Dawn is good


unique, i like this
@lyric zealot you make so much miku art, but do you listen to miku?
not all, but some old school, yeah

which ones?
yeah we have very different styles
I'm not a feet guy, but those feet are really well done.
i'm more of a kikuo fan
the opposite of cute 🙂
ahoge test
and back to totally cute
The heck is going on here
7'11 "humans"
anime, the not-cuddly variant
my fav song:https://www.youtube.com/watch?v=PhTVMDFDRx4
Provided to YouTube by CRYPTON FUTURE MEDIA, INC
きらいきらいきらいって · KIKUO
Kikuo Miku 7
℗ 2023 きくお / CFM inc.
Released on: 2023-03-24
Auto-generated by YouTube.
Illustration : 駒鳥うい様
https://piapro.jp/Sucre_uiui
Music : Clean Tears
Lyric : Clean Tears
Vocal : Vocaloid Hatsune Miku
nico Video : https://www.nicovideo.jp/watch/sm35431143
S.C.X Homepage : http://scx-ct.net/
twitter : http://twitter.com/cleantears
Soundcloud : https://soundcloud.com/cleantears
☕
sakura miku best miku
They do be 
i will try to i2i my "sketch"
hatsune miku BACK in my active playlist. again. she just doesn't leave there...
other come and go, but miku keeps popping back 😄
cherrypicked 🍒
Cuuuute
character is hacchi, protagonist of quq (visual novel)
go buy it
severely underrated
more good results
Can you explain what it's about in a single sentence?
my brother in christ what do you mean single sentence do i look like chatgpt
I'll take that as a no. 
The desert. A girl has been wandering it for days, dragging a giant gun behind her, and now she was completely exhausted and just lay there, acceptant of her fate. Who is that girl? What is her destination and goal?
Features:
- 9 endings
- Art style inspired by retro games
- Varied soundtrack, which the composer put some effort into
- Many ways to die a stupid death ```
game description
This is a story. About a girl. And the Stranger. And... many others. Not many other things, though.The desert. A girl has been wandering it for days, dragging a giant gun behind her, and now she was completely exhausted and just lay there, acceptant of her fate. Who is that girl? What is her destination and goal? Features:- 9 endings- Art style ...
$1.99
335
should've just posted the steam game
it's probably 1 dollar or something for you americans
that's a big ass book
It's part of my AI meme creation series for responses to long posts. :3
Granted, it's been several months since then, but I find they're still accurate and humorous.
the 6gb vram in question:

there is low vram option no?
trust me, i've tried
buy gpu
moneyless
sell family
he will sell me instead
Hello am 48 year man from somalia. Sorry for my bed england. I selled my wife for internet connection for play "conter stirk" and i want to become the goodest player like you I play with 400 ping on brazil server and i am Global elite 2. pls no copy pasterino my story.
is the that the poking stick meme guy?
try anime lineart
looks like she's nutting....
with preprocessor
had anime lineart preprocessor, but normal lineart model
and... it doesn't respond well to start/stop guidance with my installs, i never figured out why
try try'
and i figured it out. It's UniPC. AGAIN
As far as ControlNet goes, are hands consistently better when posed in a way where the fingers are more spread apart from each other? It seems like perhaps when fingers are next to each other, there can be more overlap and confusion and bad hands.
my pc is crying for help
maybe try using multidiffusion ?
i am.
it's just taking a bit
is it still suffering?
I like how she's part of the installation.
OH

did i just get neuro reacted
there added a few more 
Good :o


WAIT
Just two >:o
I take back my pat


Sorry guys, work stuff took longer than expected. Will start making preview images now and then upload the new negative ti!
FIONE MORE INCOMINE
should I stay with the name negative_hand?
ya'll taken the recent auto updates?
He finally updated?
Not a big fan of this
what even are the watermarks
artist watermarks
SD watermark, saying it's AI generated.
Hope we get an extension for that now.
im sure someone will figure that out
I just hope someone will care enough to make it. 
Thats fair ig lol
It was removed because the option never did anything.
i need to use a parenthesis but escaping the character isnt working
What makes you think it isn't working?
red
As in, it never worked to remove the watermark or it never had a watermark to remove?
Nevermind. I can't read.
i generated even though it was red and i think it worked
Yeah, I've been reading it and still don't understand if it never had a watermark to begin with. Lol.
summer mfs
The code meant nothing, I get that. But did/does it have a watermark by SD or not? 
winter homies
Yeah, I'm pretty sure the token counter is just is just checking to see that all parenthesis are paired and not taking into account whether any have been escaped. I'm loading now to check.
There was never a magic watermark to begin with. If there was, then the AI art identifiers might actually get it right more often.
Ok. The actual answer is in this pull discussion https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/7510
No watermark with webui.
Yeah there's been some recent updates
curious if anyone has tested out the new stuff
Will soon
Reading through updates first
This is good. Likely what you were talking about the other day.
wrong pic one sec
wee found the sweet spot for upscales without ooming always, 1.75 to 896
its 10mins but its nice
Other notable changes in the latest Auto1111 release
- "Extension settings backup/restore" is nice.
- "Image viewer scrolling via analog stick" Why??

- "Restore the progress from session lost / tab reload" is very nice.
- "Increase extra networks UI height to fit two rows of cards"
- "Use override to apply checkpoint" Interesting.
Cant wait for the full controller support for my WebUI, and the PS5 port
Lol
ive seen some mangaka use analog stick to scroll in programs
I can only assume the analog stick is because some bored developer out there is using either --share or --listen to generate from their Steamdeck.
are you from Argentina?
Thanks had a good read on that xD what a joke
Yeee 
I like the new x/y/z script value dropdown select
chomptastic
Yeah I didn't entirely understand that. Don't we already have dropdown menus?
Wait. It's for the values? How is that going to work exactly?
me too lol
please show example with SS
I wanna see how it looks like
164 gens, enough for today
I don't give my social security out for people, sorry. 
Ayyy 


Im not at the PC rn but you didnt updated ?
my webui is like super ancient 
tried updating and it broke every part of the webui
If something is broke come to #🤝|tech-support and well fix it
well, have to resinstall a fresh copy when I'm less lazy
just really lazy to go through the motion 
I'm certainly going to try a separate installation for this new release. I like how my build works currently and don't want to risk sudden big changes.
I'm gonna break everything and cry later 
Goodbye Vlad.
I'm sure I'll be fine. I have 3 different forks installed and at least one of them is in good working order XD
Also, apologies for the bright
If you encounter issues or want to know how to upgrade to torch 2.0 and xformers 0.0.17 here is the guide:
#🤝|tech-support message
It looks like torch 2.0 should install by default now. I just deleted my venv and will confirm once it rebuilds.
Let me know then ill update the guide
More Vram usage

If Vlad's is any indication anyway
Not really but for GTX cards also no improvment
Not particularly in my experience unless you have a 40 series GPU or want to use xformers 19+ so that you can have deterministic xformers.
guess just updating for the sake of updating 
one day, just one day i will have a 4000 series card

Yes!
Yep, new venv is building with torch 2.0+cu118
Perfect, thx
But xformers not
And now the installation wait begins...
still waiting to see which xformers version it builds with. Haven't gotten there yet.

xformers is 0.0.17 by default.
Nice as it should be
Is there a problem with 0.0.19 that you know of?
Not that i know, is it dev or stable?
Last time I installed it was dev, but I know they're onto 0.0.20 now.
looks like 0.0.19 is stable now. pip install -U xformers just upgraded me to it.
Now your going all in for performance xD
I just want deterministic xformers XD
Me too but then there is also the cudn Benchmark Option who gives us non deterministic images between sd sessions
@lyric zealot
I'm fine with that. I rarely try to replicate images across sessions, but it's nice to be able to run batches with Hires fix turned off to find good seeds, then lock seed and turn it back on without worrying about RNG ruining the fingers on the second attempt.
Yea true, mabe one day they will fix it 😄 hopefully
Super cute! 
Oh right, now I need to reapply the SAG modifications.
Messing with my styles has made life so much easier
RIP.... the patch doesn't work automatically anymore.
Whats SAG?
Self attention guidance ⬇️
Ouh nice does it ad a slider or how to apply it onto an specific Object ?
Ah nice
It's great when working with low CFG and helps with general togetherness of the gens as well as better contrast. It replaced dynamic thresholding and the like for me.
You still lose some detail and creativity as you would with higher CFG, but the tradeoff is often worth it depending on what you're trying to achieve.
I've used it as high as 8 CFG and it still has noticeable improvements.
They vary per model, but are great for things like this. (no SAG--created before release)
Nice style
Yeah it's a great LoRA: https://civitai.com/models/16055/colorwater
interesting
@lyric zealot Didn't have emote maker lora downloaded
you should get it, it's great for emote like pic

Gn 👋
goodnight
Goodnight!

Alright, finally got the two scripts manually patched for SAG. Time to try it. No error this time on loading SD so that's a good sign.
yo what is that style/lora?
It's likely this one. https://huggingface.co/Surficialruby/LoraDump/tree/main
I will eventually figure out that loea colab
Didn't work. I'll just live without SAG while I wait for ashen-sensored to figure it out.
Okay, I'm done. Its uploaded enjoy! https://civitai.com/models/56519/negativehand-negative-embedding
im trying now
turns out skateboards are tough
It hasn't completely defeated this hand blender of a prompt I've made, but 2/4 images with good hands is a lot better than I've had since I started working on the prompt last night.
agree ⬆️
how well have people here managed to get images to be well "controlled"?
for example: can people get a single character to reliably make a set of different facial expressions from different positions?
smug is an example expression i dont think can be done well
quick run.
not too shabby
guess that's going in my recommend lists
counterfiet?
amazing
naw somethin' else... magic maybe? id have to go look again.. ran a grid 😛
magic...."weed"
is like guilty gear
someone fr commented on my model "does it do nsfw" my brother in christ this is civitai it has too
Using img2img and ControlNet with 3d character models, most of the outputs are pretty bad with most models. I can get decent one-shot results with only a few models. It's a bit limiting, but the speed and control is great.
looks pretty smug to me
Looks interesting. I'll try it out tomorrow.
I just noticed @nova remnant is a luna reference
It's an experimental SD checkpoint.
Morning
Gmornin :o
It's lora merge that can be found here and model I used was Aurora
https://huggingface.co/Surficialruby/LoraDump/tree/main
Am I using it right? It doesn't seem to be doing much. Before & After.
@lyric zealot New day new try
Hands look better
How to use it brother?
Use what?
How to create images like them
What's that?
Graphics card

Nah I didn't have these
Literally i don't know how to use all these
Time to look around a bit then
The mighty wall 
you are using it correctly, i think. The thing is, as mentioned on the civit page. Its not a perfect cure for bad anatomy, since the general Idea is that you have a decent neg ti, that is not changing the artstyle of the initial model, while also improving quality
Did you retrain it or is is same as this?
retrained
Hello, quick question, can I take an image like for example a picture I took in real life, and put it into SD to apply a model and enhance details on it ?
I tried img2img but it doesn't add details because no hires :/
what started out as
ended up
well
heh
I need to adjust my latent upscaler a little
What kind of details? Higher res?
This guy went from this to this with SD, I want to do similar things (forget the title its not important)
well it's definitely img-to-img but can work with inpaint too
hires fix helps with small details too and if it doesnt work well for you, you can always use lowRAM option (I do that and it works well). It renders a bit longer but helps a lot
but with img2img there is no hires fix, so I can't use hires fix on an image I already have ?
There's tiled diffusion extension and you can choose to upscale there
oh interesting
will check that thanks
Does someone know how to fix this issue with reddit?
I might go home today!
Can anyone tell me how to find 18+ nsfw in civitai!?
I used to see like allat now I CAN'T FIND THEM
Check your account settings
Hlo
hlo sir
Hiiiiii
welcome
Thanx
wait what is the etymology of welcome?
I dont know
waiting on an opening for a ride home and then I'll be outa this hellpistal
hellpistal now that's a new one
wait so did the thingy go well?
does everyone here have the discordo effect when opening discord?
you mean the cute discord sound at startup?
my youtube recommended is getting out of hand
first vid :)
plz don't hate me for a lame vid.
and yes i know unfitting music. i wanted waka laka but couldn't find it.
so i just did paralyzer(youtube picked it)
disclaimer
i don't own any content in this video nor do i own azumanga daioh
disclaimer
i don't own any content in this video
reminds me of the ronald mcdonald osu map
finally be home to prompt or fall asleep immediately in 2 hours
getting outa the hospital woooh
Can't believe I missed this
lol
I also missed like a major announcement in school and like some messages people pinged me from 29 apr - 2 may
That's what happens when you root your main and only phone
I formatted my phone 3 times
Each day I formatted once
First time it worked perfectly fine, but I ran some script and system corrupted, then next day I formatted again but then realized that I put an outdated rom so then the next day I formatted again and this time is perfect
Fr
hii, new version of my model is finally up. if someone wants to have a look at it:https://civitai.com/models/5374?modelVersionId=60706
Some of the images might use embeddings or Loras. If you want more creative results, go for Clip S kip 1. On the other hand, Clip Skip 2 is your be...
8.5GB....
@fickle moat
Currently, there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝|tech-support for any problems you encounter while installing it locally!
if you want you can trim it down to 2gb with this by selecting f16, no-ema https://github.com/Akegarasu/sd-webui-model-converter
i will, once my slow ass webui starts
omg i can use any model with this
yep
I still use 32 full when available. almost exclusively use my own model though cause it can capture almost any look at this point
15 more mins then I'm home

finally
you mean me?
i can't sleep in hospitals been sleeping like 1-2 hours max every day
dark gemini 3
sorry i dont really understand
my model
ah ok
I haven't used any other in months
damn
just been perfecting it
almost all I can do with model merging now and see as my comp is too low on vram to train a model. my next step is loras merged into the main model
unless some flaw comes to my attention and I can't solve it using loras
how much vram?
then it's enough for lora training which when used with good settings is like practically the same as normal finetuning unless you training on a really big dataset
that's what I said though
and I can train a lora in stages by resuming it with new images
yse
or Train two strategically and merge
but this way is not optimal
no but I have no other way
you can start training from beginning
I do..... but if there's too many images for the VRAM I'd have to do it in stages
oh yeah
You guys sometimes just get like, unwanted p0rn?
you really are extra high
depending on models but not really
no my neg prompt is too strong for that
Not gonna say I get often, but it occurred like 2 times lol
should i do only fp16 OR ema only or should i do both?
even porn models can't generate nudity. sluttiness sure, nudity no
wdym?
they can generate nudity
....
just most aren't really that good at anything that includes acts or motions
or males
lol
do you lack object permanence?
yeah
you should have said then
I did, a few seconds earlier
yeah but I cannot be sure if they are connected
I could always blame my autism anyways
but now I utilize both my adhd and dyslexia, checkmate
I'll send you the new DSM of only me
I love my adhd now though, used to hate it but now it's a tool. buh bye
Crazy how it's bad for males, not much interest anyway. lol
well there are some models for males
Both is good
What's need for such things?
(Story) Fyriel, Fyroris's daughter
what's the opinion of the pros/cons for checkpoints for anime? like WaifuDiffusion vs Anything vs NovelAI vs OrangeMix etc. etc.
https://civitai.com/models/40199/aurora Is the best one, thats my take :O
The first 3 are very outdated by now
lmao thanks I was wondering why it wasn't loading
understatement
Realistic anime!
Saw this in another server kek
/imegine
home....
heyy, any idea how to reproduce this kind of hairstyle ? i don't even know how to name this 😦
cornrows with...a leaf?
XD
the rest is cornrows
it's supposed to be spikey blond hairs
okay thanks
and for flat hairs like this ? '-'
uh that provides very little detail and looks more like a helmet
try to make the lines between the face and the hairs much thinner and the blending much greater
fp16 of my new model version is out
what's new since the 26th of April except Deep Floyd? Of significance that is
'Anime protagonist hairstyle'


Anyone got any ideas to make this character more interesting? 












