#🍥|anime
1 messages · Page 36 of 1
model?
Aurora on 3 CFG
yes
@lyric zealot really good prompt 
ikr
bro yumeshiro is sooo good wtf
can i inpaint it or smt
upscale -> inpaint only masked
idk whatever you want
she got penis fingers
ye
lol fr
just gen again
anybody can tryryyyy (totally not asking)
just inpaint the entire arm off 
gacha until one hits
that's what I do
prompts?
either that or go copy and paste a hand over it and i2i
me when
all of them???
no
u guys should make your own negative ti
that's just almost all I have
same prompt, just increase the mask blur to 8+, denoise to .75+ and it should be fine 
i only use easynegative and badhands, i tried coctailing them once and it turned shit lol
aight thanksss
yes
i wonder what will happen if all of them got mixed together
play around with the settings, sometimes it doesn't work, still rng
put in positive prompt
i did once by accident
just accept shitty hands
i had nightmares
ur life will be much easier
I got cool results actually
generous japanese ppl
I made in total 500$ (without even trying much lol)
??
and
and they just donated
bruh
shrug
brother
forgot yen dropped
this is just from koーfi
im telling you the japanese are very generous
If you would like to see more and better versions of Bras please consider supporting (tryna recoup gpu cost lol) https://ko-fi.com/bankaiplease Tra...
well it is an extremely popular model
and also very good looking
like damn this is good
well yeah but all im saying is that the japanese are generous thats it
its not bad i agree lol
but i put in effort okay
i didnt just randomly merged two models
it was all trained closed to from sd1.5
ok maybe a bit overfitted/overtrained
ofc, it is such a good model after all
fr
I have too low self confidence and I am also too lazy to do stuff like that
I only post troll models on civitai
and some random loras and stuff
Smh
Reddit doesn't think so hahaha
Whenever I post there I get instantly downvoted
yet another reason not to use that accursed website 
twitter idk how to use well
also people in reddit actually try to argue
they write longer texts
I've only ever used it for news about stuff I'm interested in
I wish I knew how

hello everyone:3
excellent

its alr bra V bro
damn i didnt get even 1
same

my instagram is more active
this looks scarily real
you know what I'll upload a new post rn
i think people that browse instagram is a little more against AI art
or at least that's what i think
I can't even tell
done did
the grandma face is too bright
she is blessed
how many followers?
29
this is what happens when you are like me
even though I am a youngster, I feel like a boomer with most social media
good enough for someone who doesnt upload
same man
how old are you tho
17
do you ever get any bad ones or is all of them this good 
yea and eyes
Hehe
yea full body always mess up the eyes
for smoking cigarette, it's a little more rng, but im using no lora for that, only full prompting
i realised using, smoking, cigarette covering mouth works really well

yup for that, might need to either inpaint it or tiled diffusion
tiled?
extension?
some extension that's like hires fix, but using tile
this right?
but i don't really know how to use it, so 
yea seems a bit complicated
Clyde allowed legs 👀
will have to check it later
yup yup, or just try to inpaint it
if people from projectai like it it means it is good
@simple mango I invite you to projectai
is that how that works?
yea i tried inpaint before but took a lot of tries
i forgot how i even got it to work
it's been quite awhile since i browse there 
how old are you?
oh
used it for this one
every day there is something new
ok
i will take that as a compliment
general rule of thumb is that, inpaint area = Only masked, put Width = Height = 768, and prompt accordingly. denoising strength should be really low like 0.1 - 0.3
took a lot of editing
other than denoising strength I agree
what should be the ideal strength?

saved
it depends on how good the original is. if the hands look like rocks or don't even exist then you should do higher
will try that later
well, definitely varies depending on your generation, but it's just a general rule of thumb 
I usually like around 0.6

but it does depend A LOT
for inpainting?
yes

sometimes even higher
actually i almost never do inpaint
because it's a pain to get right
so yeah

anyway got to go
got to go fetch friendo from airport

nice
insane website by a madlad. does weekly and monthly updates on all things sd. here is example from yesterday
updates most days
very useful for general news and also for resources
it is already on #1080946152318443610
but not enough hear of it
@wheat falcon sorry for ping
nice


I call this piece of art
playing pinkball in the dark
<lora:chapu128dim-epoch-000009:1:1,0,0,0,0,1,0,0,1,0,1,0,0,0,0,0,0> <lora:mmix_v7:1:1,1,1,1,1,0,1,1,0,1,0,1,1,1,1,1,1>
be like
How does this seem?
Pretty good 👍
Using 3D to make lines and depth for use with Multi-Controlnet to try making multiple characters in custom poses:
:DDD
However, outputs are not detailed. I also cannot specify colors to each character. I think I need to use regional prompter somehow
I also need to use segmentation so I can choose specific parts of an image to change color/modify details
ah got my room now. my own personal room
YEEEEEEEEEES
now you can prompt fishgirls all day
no I have no access to that
and I'll only be pseudo-conscious anyway
most of the time I'll be half asleep or asleep
hmm
what about
use remote desktop
that way you could proooompt while you're taking bus or whatever
how can I possibly install that back home? and what bus? I'm in the hospital for cancer surgery. not on a bus
I'm not dead yet
I could generate fish for you
uh no... but do download my model and spread the word on how amazing it is! Dark Gemini 3
what have you made on it so far @mild tree
looking very vibrant
If you can give me a few minutes, ill start up my and send you the results so far, keep in mind i only make portraits pretty much tho
Also
I already posted a review with the gens from my first day with it
My name on civit is Yaghrujhu
Nvm no review
Just a gen collection
man, merging pastel models and getting something that doesn't look like shit is hard.
Have a good one!
really
i was sure it got lost while uploading
I made the prompt simple with just saying two girls in front of a stone gray building. I was hoping it would add various colors to the characters but I guess not.
hm
I made other ones like this but it looks like this
Ah, I increased the steps to 30 now, and added more color prompts. Much better:
The girl's pants on the right at the seam is messed up because the 3d clothing was not meshed properly from the skin weights. I could fix it in 3d or just erase the generated lines, it only takes a min. This was more of an experiment. I'd like to get a workflow down so it could be possible to make webtoons/comics really fast and easily, compared to the crazy 70 hour workweeks a lot of artists have to get through
I love it
nice result!
good luck you got it
i think control net can help a lot if you are not using it
clip skip 12 new meta
i want to know who released this model so i can have a few words eith them
1girl, blonde, elf
whacky
Hmm, you know what? Ma-maybe I too could be neenja! [Chorus] I want to be neenja I want to be neenja I want to chop, chop, chop Chow down Take Chow down to Chinatown I want to be neenja I started training hard, I think I could go far I am a natural, that's matter-of-factual I learn to chop real hard, I fight with credit card I use my nunchucks, even while feeding ducks I throw my neenja star (wah!) oh so very far I do karate moves with a GoGirl groove! I learn my neenja kicks while watching Netflix I try do everything like on the big screen And you guys, you know what? I might do it! I might be a neenja soon!
lyrics of my fav song

looks like you have no VAE
really weird, i had the same problem with NyanMix
idk that one
idk where i got these, was drunk last night and found a bunch of models in my folder

like its just named model
smh
Sounds like the start of a horror movie
Anyone know if regional prompter extension will work for Vlad?
file on the desktop that says play me and its just 10hrs of danny devitto screaming
why do I get same vibe from these?
first is from chainsawman mangaka second is very old prompt of mine (and bad)
how figurine?

Anybody know how to get faster results, i'm running SD with RTX 2060TI (8GB)
what should i optimize
it takes around 35 seconds to generate a Text2img and Highres fix: from 448x640 to 896x1280 using X4 ANIME/SHARP upscaler
https://github.com/lshqqytiger/ComfyUI-directml
found this. finally gonna see whats the hype about comfyui
resize via hires fix?
40 steps is too much
is it really?
yes
Kind of
for 2m yes
I'm always lost concerning the steps
I usually prefer 28
i usually stay between 25-30
20 to 40 is like very fine tuning, so its better to go for 20 for speed, and it still look pretty much the same
yes

seams?
i need to learn english 😂
True 
and also ye learning english will also be helpful probably
i can read, but i miss a lot of vocabulary
I researched a bit on how to acquire language. if you watch anime with english subtitles, you will just become good over time
literally backed by research
i read a lot of novels entirely in english
so it helps a lot, but sometimes translations are trash
I played games in english when I was like 6 year old, and I just
Learned
From like associating words
I think
xD
Actually learned a lot with minecraft
Yeee, games are OP
Does SD works with SLI graphic cards? edit: found the answer
frrrr
mostly listening for me
podcasts and radio shows





this is the best Image I've generated so far
S M U G

this is already dark skin 
im trying to achieve what you did
😭😭
maybe more weight? 
i was expecting this, but just didnt hope for it....
definitely
i don't think i can achieve that kind of dark skin on this model
Not with that attitude you won't 
maby?
idk
but this isnt dark, just lightly tanned
scheming
dannnnggg mommy
im trying

AHHHH GUYS
dark gemini somehow synergises with arcane lora really well
tf was cryonicus smoking when making this

I disagree with this. Often times things don't fully solidify until the 40-50 range for DPM++ 2M Karras. The most fine detail often shows up in that range as well. In my original testing for the model when it came out, and again just now with varied prompts, negative embeds, and sizes it shows this. Take a look at my testing to see for yourself. 

oh shit
I ping bwm for this
@supple raptor
I am pure evil
hahahhaa, I'm joking as well. I just really like having forehead shown for my waifu
it depends on model aswell

You can do what you like, but don't let your presuppositions interfere with your observations. 
for DG specifically 25-30 seems to be optimal, except for landscape stuff
cursed @mild tree dark enough?

Yes it does. Step detail/convergence points vary wildly based on the model.
i really like charred skin look

Do a Nidalee prompt 😄

would need a lora
actual charred skin





Darnit Smol. Beat me by a millisecond with the blue heart.


whats up with this weird pant
hot tho
is this bad?
Another 2M Karras step comparison. Look at the detail in the eyes, fine elements of clothing, and environments at 40-50 steps. Also look at how sometimes the clothes somehow inexplicably get lesser with the more steps, lol.


nononononononono

please kill me
i just have to genuinely stop...
otherwhise bad things might happen if i continue

hands are messed up but dang

Clyde'd
What model is this ? :0
expmix_line_v2 
Thank you!

Yes
damn
No
no damn?
No ai
“I will now proceed to pleasure myself with this fish”


https://www.youtube.com/watch?v=wAkX1giNels since no posts in 30min so have this funny video
Adorable
Thx bro


I love the line work on that, how it looks a little sketchy but still clean

Ran her closeup through as a base image

Should run her through that lmfao face lora
Hey all, guess the char 😄
reimu from touhou ?
Kyo Soma female version ?
your fantasies can never be quenched

Yeh 
just for that
I'll dedicate my stable life to generating fish
not even girls
I'm done with anime girls
Dan
I'm sorry, I only see legs and attitude of a professional wrestler (Kasumi DOA?) (Kasumi Yoko?) 😮💨
Nami from Full Metal Panic? :o
wow thats pretty similar but shes not from an anime 😮 also dont mind the not 100% accuracy, she has no lora xD
Oh, hmmmmmm
I mean, I wanna say maybe Baiken from Guilty Gear, but that is a pretty big stretch :o
Its Hime-Sama from Crunchyroll 😄
O, I would have never guessed that XD
That crunchyroll character? Oh... I scrolled and saw I was right! I would've gotten it. 
idk wtf the fourth one is but wee
Holy fuck those fingers and toes are good
evenin'
Evenin 
howdy
sup
I be goods, how ya doin?
doin' well
Ayyy, good to hear that :D
is there any good ways to stop text from the image
oh wait i forgot about (watermark)
yo your a star now
congrats
sponsored by crunchy roll
you should have added white streaks too
Thanks discord for tracking everything I say and then giving it back to me as ads
When I use ControlNet, I can get good hands. But when upscaling the resultant image, it ruins the hands and makes everything else better. What to do?
how are you upscaling?
I used the hi-res fix with latent. I don't know too much about upscaling, haven't done much with it
okay yea neither lol
I see, I'll have to research more on upscaling because it's a vast topic.
The upscalers I have consistently recommended to me/have recommended to other people are 4x-AnimeSharp and RESRGAN-4X+Anime6B. I use 4x_NMKD-Superscale-SP_178000_G for everything because it gives it a "painted" look
this gives me big persona vibes
I have the esrgan 4x anime, used it once. Does it always upscale x4 or you can make it upscale x2 if you want it to?
you can set it how you want. I have it do 4x for first gen pics (512x512 or whatever) and 2x for hi-res'd img2img pics
If you want more contrast and richer colors with an excellent anime upscaling, then try 4x_fatal_anime.
I should try superscale to see if I can get a somewhat painted look without the extra contrast boost.
best of luck. i'll check that one too, thanks
Redo of Navakas, the Trickster God of Change and Adaptation. They're genderfluid and as they would say, they "take whatever form makes people the most uncomfortable"
Yeah, its down
Fudge
ayo its kit-9
good android
Gotta love how Stable Civit is

Its such a Stable website, never down, never fails

If civit went more than 5 hours without a crash it wouldn't be civit
True 
Well, atm
It is
"up"
But, it is not working
xD
Like I can see it, its right there, downloads no work tho
I love it. When it's working it's so badly optimized it feels broken, sometimes it flat out refuses to do anything and most of the time its just dead ❤️
It is the prettiest sharing site tho
I mean also the most convenient, huggingface is not as easy to search and find stuff
Truee
They got a lot of functionality, just need to optimize the site and upgrade the servers. Easier said than done
Yehhhh
I mean, upgrade servers first, just make em bigger
And also fix whatever is the thing that is making them crash every 6 hours
XD
Hahaha yea, seems like they're dealing with some issues caused by an attack on their servers, at least that's the announcement from 1½ hours ago
Thats a funny way of saying horniest
75% porn models. But you know, filter button gets most of them
I suddenly discovered i could like and unlike a model and give it i finite likes for as long as i clicked but then it suddenly starting going backwards wtf

Github lookalikes are never easy to browse, navigation of those places is for the most savy of techbros
Casuals just need a one and done link
I mean, true
Though I mean
I like to see what the LoRAs I am getting kinda look like
Huggingface LoRAs be like
Mhm
No model card
Exactly
There are definitely some great models on huggingface but they don't got any preview for it
And name of the file looks like its a encrypted nuclear code
gm
Yes that is a whole ass problem for me lol
I recommend it

I'd play the p out of this videogame
NieR:Automata 16-bit retro demake. Yes please!
But madame, a demake would be something akin to
Resident evil 4
damn that's cool
It'd be my POW submission if it weren't for the extra lock of hair by her right elbow. I might have to fix it so I can submit it.
Interesting
Idk if this is the right place to ask this, but I got a question. Does anyone know what model made this image? i love the look of it a lot and wanted to make more in a similar style
1st thing that comes into mind is this
https://civitai.com/models/12597/moxin
ngl, it looks like a generic anythingv3 image with no vae
wow! thats a bit different but thats really REALLY interesting! There are so many different textures and different uses of lines. the display images for the model don't look anything like that image, but that really is an interesting look! the use of lines in random places really feels human even though you can tell from the ai mistakes that its ai. genuinely might be a new favorite of mine, def going to tinker with this one
more from my animepunk lora
vs my own custom model trained on the same style
im still refining it
Interesting
https://civitai.com/models/47085/envybetterhands-locon Anyone try this out? 
how do you use a locon
https://github.com/KohakuBlueleaf/a1111-sd-webui-locon.git Install this extension through "Install from URL"
okay and where should it appear?
A boy eating
You can use it like normal LoRAs
@topaz yacht
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!
ooo oki thank you
Yeee 
I am trying to use this lora https://civitai.com/models/16432/chichiolder-dragon-ball-z or https://civitai.com/models/25868/chi-chi-dragon-ball
Here's the Annoying Bitchy MILF we all love to hate. Who none the less we would tap on a heart beat. Seems there were a couple of LORAs of her pre-...
But i can't
Is not working
how can i fix this?
civitai seems to be down again
guys I forgot how to make good ai art
this is so sad
time to steal prompts and change them a bit
Sup
I remember that's how I got good in the first place

Oh damn, I gotta try it out then :o
Its a big boi tho, 500 MB
totally dont check the filename 
Looks pretty sweet tho 

i'm checkin
Frooti 
Smiili 
Ayy :D
Also Fruit I think you missed the Mikoos I posted a while back, the Mikoo Coolection™️ :O
?!
in #1055215351731851354 or ?
omg bless u..i was scrolling thru ur msgs too hahaha
AAAAAA
Yeee, I was trying to find it too XD
Though I just realized I did the last one wrong, it has long hair 
That one was one of the easiest I think, the 3rd one of the 1st row was so hard, that skirt is so specific xD
I do yes 
Should inpaint eyes for even more detail though 
How was this used? I want to compare the 3 models.
Put all the models on the X axis 
Separated by commas
smile .. 
TYSM!!!
Okey thanks
i get this
You welcome 
After all I love me sum Mikoos too 
(inpainted the eyes)
Hmmm, try using the hash codes then
I never tried with names, I always did it with hash :o
WAIT
Before that
You should try setting Y and Z
to None
WAIT
Y
To like
I'm stupid
Y set it to seed
And set it to a fixed seed
For easy comparing
now is working
Oh that works?
should i do it?
I mean, its better for comparing, since it will use the same seed across pics
Some people like to compare with random seeds, I like to compare with fixed seed
when doing plot xyz with model comparison
it's already auto fixed seed
for that same batch
okey thanks

I get this with seed -1
unless it's different batch
Its a shake weight 

can confirm
whoa, weird... I was creating images very similar to that last night. How random. 😄
Have some Tifas.
easiest way for me is select the model you want then copy paste it from the selection drop down with its hash, or if you dont have 100 models like me you can hit the little clipboard and itll fil the values and you can delete the ones you dont want to use
yeah id say its worth it so far
Damn, gotta get it then
Guess what I like the most on this image?
Where can I get this lora?
ill upload it soon
its a style lora but i also have a mixed model for this so ill just upload both

Bf3 LoRa ?
Where can I get latent upscaler for img2img?
i kinda like this
very cute
noice thats a good sonic
does anyone know how to use this one https://civitai.com/models/7808/easynegative
Original Hugging Face Repository Simply uploaded by me, all credit goes to https://huggingface.co/gsdf . Counterfeit-V2.5 (which has 2.0 as well) o...
?
that's a embedding so download it
put it in embedding folder, and use it in negative prompt

thank you!
if you need more details, ask me again


I thought it was this one that Clyde decided to block




















idk how to inpaint