#🍥|anime
1 messages · Page 42 of 1
yis
horror
I think you should go for 2 batch size, and uhhh
How many pics do you have in your dataset?
23 in total
Thats not very many 
decent gen
Maybe you could try to pull it off though
How many should i get?
Going for a style lora too
If that makes a difference
I mean, I usually try to do like about 80+
as much as possible, but quality over quantity
True too 
The first one
oki pog thought so
Though also what is your repeats?
this is what they recommend in the eli5 guide they got for this section
So thats, 1e4-5 and 5e-5, that seems fine 
And I've never used warmup, so I'm not sure how it works
im not seeing anything by repeats, can you elaborate pls
But I think with warmup you need to set how much of a % of your training is going to be warmup
you dont need it
oh ya they said 500 warmup in the eli5 guide
Well, at least in Kohya repeats are usually put on the folder name for example "4_Thing" means 4 repeats
Yeah I've heard that 
I mean, I know DAdaptation uses it, but thats it I think
You should set clip skip to 2
100 epochs jesus christ
thats the default lol
lol
And my max token length is set to 150 
my max epochs was like 6
my standard is 14
so something between 5-20 epochs?
I've done like 40, but das cause I always try to go for low repeats 
also prolly why it took like 115 steps or w.e per image
an hour for each., woulda taken 24hr for it to run completely
for all 23 images lol
pls keep in mind im 90% not using the right words this is still new to me and im struggling weith words 
I usually do 3k steps, bunch of epoch, and each epoch is like 100ish steps 
Make sense xD
I think you'd need to figure out the repeats though 
im also not sure if im suppose to extract the weights first, i have been doing that cos it wont let me access the restr of settings without it
and idk if its kitchen sink being weird or not
like until i hit create model i cant see the rest of the settings
New LoRA once more
You should use [REDACTED] as the base model 
redacted?
And I'm not sure what you mean with the weights
Be different and use ACertainty for the base lol
Yes, the [REDACTED] model
The forbbiden one
since idk what that is and im trying to learn no lol
Idk what that is :/
Very cute
ig u can dm if u need
Fair xD
Yeee xD
Now I wanna try cooking a style lora too
Gotta love random errors
Also cough cough @supple raptor yknow anything bout this?
XD
Hmm... My first guess would be maybe a bad download. I would try downloading again and replacing the copy you have to see if that fixes it.
Maybe, I know at least one of those is a bad download, cause Civit sucks balls xD
Gotta love how my LoRA folder is so full of stuff it takes like 20 seconds to load XD
Well, organize xD
Let this be a lesson to all, not to choose a database system that nobody on the team knows how to use.

This one is at least tho, but clearly a broken download, good job civit 
This one I tried to download 4 times 
it's referencing backwardsVirginKiller_V11_3.safetensors so that one would be with the b's not the v's
Ohhhh, you are rite
And 4 more attempted downloads here xD
Gotta love when civit half works, so you gotta just keep downloading until it actually downloads xD
wtf
Support me on Patreon: https://www.patreon.com/soundlibrary
FREE Downloadlink: https://link-to.net/63999/22822
You use Amazon and want to support me? Use this link for shopping: https://amzn.to/2DeVUlj
Get cheap video games: https://www.g2a.com/n/soundlibrary
Good program for editing: https://amzn.to/2BV8LbN
Professional studio headphones: http...
AI going schizo
Stuck in a loop. You'll want to delete that reply so it doesn't poison the context memory. May need to slightly tweak the parameters as well.
i am using sillytavern and the new Pygmalion 7B, what should i adjust?
ok wtf this happend like 4 times now(only dakness is doing dumb shit)
and this happens sometimes
Any small change to temperature or sampler parameters will likely break it out of the loop as long as you've deleted the problem messages from memory. I never found settings that were very stable on Pymalion7B though. I've been running llama-13b-4bit-128g and the NovelAI-Pleasing Results preset works quite well for it.
pygmalion7B(which is based on llama) came out yesterday. i think you mean pygmalion6B(which is based on gpt-j)
Oh, you're right. It was 6B that I was using. Well, since the one you're using now is based on Llama, the same preset might work well for it too. Here are the parameters from Ooba incase you don't have the same preset.
Nice! Glad it worked
works way better than before
I went through all of the presets that Ooba has with a couple of characters that I made and that's been the most coherent that I could find without it becoming really bland.
bad_prompt is on civit now aswell :)
new version or the same one?
Ah, good idea. Seems like a lot of people don't even look at HF anymore.
true
probably because of the interface
yeah HF is kinda dead rn ngl
Yeah, HF sucks for discoverability. Unless you already know what you're looking for, you usually won't find it. Same problem that github has.
can you somehow upload two models on the same page in civit?
like let's say version one and alternative
or you need two different ones
Yeah, you can do versions. I haven't uploaded anything there myself so I don't know exactly how it works, but I've seen models that have multiple versions on the same page.
i couldnt find it while reuploading bad_prompt right now, but months ago you could easily do it :/
damn
I will check it out, maybe I will find something
ou yeah also I updated my model too @queen void
i'll take it
Finally 
My First Lora with my Main she looks soo good owo
Testing starts once more. xD
As much as I don't wanna publish my model to CivitAI, but for visibility, it has to be done 😅
Oh no all Luna's are Smile on this server :help:
Wtf is easynegative doing
idk
Omg
bad seed most liekly
I didn't realized it changes so much
this is full strength embedding though
I mean
not 0.8 etc.
Why don't you try uhhh
(KHFB, AuroraNegative:1.1), (NEG-badartistnegOG:0.6), NEG-bad-image-v2-39000
I'm kinda happy
Thats the four embeddings I use
I use em all the same time, along with sum other negative prompts
Style still there 😄
yeah xD
i need the last two
Lemme send em
One is on civit, the 2nd, one and the 1st one I got from HF
By uhhh, Xenon I think?
Almost XD
Yes xynon
O, fair then, O RITE you got a 3060 with 6 vram
I got a 3060 too but it 12 vram, it was cheaper than 3070 too xD
and I need sleep soon xD
Same
I mean, for me is like
Hires-ed
Like 40-50 secs per pic :o
Looner's negatvis seem pretty OP :o
I mean, I use the combo of sis' embeddings and the other two
but this test was on kenshi so :D
I have tried some others, but I feel like uhhh
EasyNegative, and DeepNegative change the pic too much
Vileprompt I'm unsure, Badhands also unsure, and Badprompt also unsure :o
mhhh
I tried EasyNegative but didn't like it very much too, it goes way too much of my liking
I should try easynegative2
basically the same
Oh.
the maker said its almost the same just trained on V3
Perchance :o
almost
well idk why I even do it xD Thats without highres fix and kinda dumb since it could be seed related or because I test it on kenshi
guys... I took a lot of hopium and copium today and its 2 am but I see that correctly right?
Yes
Luna's neg ti's are trained on Kenshi and aurora, mine is not.
and highres fix is rly heavy aswell
although Luna's negs are rly good
model depended and thats just one grid. If you go alittle higher there are some good examples
hello long time not come here
Finally perfect hands
Like 10 versions down trying to get this style right
Had to crop, but it may be finally good 

Need to do eye check 
Hands are scuffed, but face is not too bad, that forehead gem tho, can definitely notice the Raven fanart in the dataset
What a time to come back... when no one is around. 
Hey Phaff! I don't know you, but welcome back.
Kinda tru, also welcom back
Shit, nvm its absolute garbage like the other ones
Maybe less weight, but if this is how it looks already
It may be better to just like, maybe retrain with different settings, again
Or maybe spend many hours editing the pics to remove everything that could be bad in the dataset
Or train LoHA
Starting doing ai yesterday, this is actually fun af.
Once you get in, u can't get out
0.65 weight, better hands locon, still bad, i am trash, existence is pain
Welcome to the Hotel California.
It’s funny when you think you finally made the perfect picture and then boom, 4 fingers
Only funny the first ten times. 
Or spaghetti fingers
True 
This is... not bad, but again like with the other times I trained, the background is... not right
Also, 0.65 weight, heresy, 1 weight is the way god intended 
Yeah, because then you realize you can't count that high because of missing fingers 
This is beautiful. is this model available??
That’s a cool background
Also gotta love UniPC noise, its actually kinda cool to look at
I dare someone here to suffer trying to train a Koikatsu character or even the 3d style itself and report back to me


Unique perspective
I mean, I want to train a 3D style at some point
Thing is, with koi, it’s like between 3d real and 2d art
Also that is very specific 
Say you do a batch of 4 and really like 1 of the images how can you get variations of that image,
I’m assuming you just save and chuck in img2img,
But is there a better way to do it?
I mean, probably wouldn't be too hard to do
I mean, img2img is the best way to do it
Its bc its all ive tried when attempting to make my ocs into a lora
Fair I suppose 
Can doing that get rid of deformities as well
You take enough pics of em?
Yeah sometimes :o
You could also try locking that seed and either using the seed variation setting or slightly tweaking the prompt. Controlnet is another option depending on how similar you want it to be.

That's a lot of epochs
That too 
Oooo thanks legends
Well
It isnt. At least it shouldnt be. I trained one of them and most of the attempts came out okay. The artsyle wasnt even that much a burden. But THEN i tried doing one of my more complex ocs, which has several outfits and no matter what i tried, all i got was jumbled garbage
Not really
I think they're mostly 20 something
Some less because of less batch size
I’ve just been using the -1 seed, whoops 
Like the latest one which was 15, because it was 2 batch, cause of 768 res
Well
If it has different oufits
and then there's me with 5 epoch
You may want to try making different folders for each outfit
And different trigger word for each outfit
I…guess?
Maybe it didn't tag it right
Also you should change your PFP to be like, 2 frames ahead into the gif
For the longest time
I thought your PFP was like, the shoulder of a anime girl
Or something
xD
Last day i spent trying to train her it took the entire day, and that was second individual day of trying after a month or two
Now I realize its a gif with a bunch of cool eyes
Btw, I counter with Tomo-chan is a Girl!
It incites curiosity, at the very least, which turns into a tap or click on it
you betcha
I mean, probably won't bother watching it, I don't care much about romance or drama :o
Ooooof
Fair then :o


@untold glacier
tomo chan 
Cute tombyo

I have a LoRA for her too 
Momo is easier to do though
But Akali is way hotter
I want you to know that I sat here for the last 20 minutes looking specifically for that emoji reaction. I knew I had it and had to go through my entire list nearly twice before I found it. 
I do that
XD
Time to steal
From a bank? 





Also I think I should restart my PC, it is begging for the sweet release of death
One day I'm gonna get randomly ban because of these anime server and I won't know why

What you mean 
A mix im working on
You can get ban for being involve in a server that breaks TOS
My main got banned
Define "involved"
Join

Man. 
Your main? 
Yes, it said I got banned for involving in frauding
Oh wow :O
Lol at this model name

Somehow I'm not sure it's fixed... 
Perhaps it is not 
6,192 votes and 807 comments so far on Reddit
That Pika's model? 
hmm I think my controlnet is jacked or something
Has someone posted safetensors downloads for the full new model set?
aka all of these? https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main?doi=true
nope, 3 random loras
I have the smaller set back when it came out
Oh damn, that has the lighting of Pika's model :o
Clyde is bullying

too big for clyde eh
That was going to be my question next 
anyone on the new webui update?
Hero
Face is abit off but style is nice
niji does have good style
woahhhh
@trail belfry
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!
Welp, after literally more than 20 minutes of trying to get it through Clyde
I give up
We're up to about... 10 blocks in a row
11 
This one is really goood, please go through 
12

I wonder how lewd is it
I'm not sure
Hmm. Can anyone else post images?
I censored all the parts I thought would be bad in inpainting and inpainting sketch
But it still didn't aproe
Sorry, Smile.
(Story, WIP) Hakunetsu Ōkami (白熱大神, Hakunetsu means 'white heat' or incandescent). She's a former Oni turned goddess that takes a position similar to Ame-no-Uzume's among the Kami when the old gods disappear. She's known for her frequent scandals and her intense personality.
i may or may not start budgting for a 3060 
I got one, it nice :o
amazon sells used for 450, may aim for that
go big or go home 3090
nah im poor i need to budget lol
Radeon 6800XT
Good card, too bad i can barely use it
XD
rice and beans for a month and you can get it
i could prolly sell off my rx 6600, its only a year old
like sell it cheap to recover some losses
Or grow potatoes in your backyard
Potatoes are the most versatile... plant
bought it for like 350+tax could sell it for 150ish, idk
i zoned out what about potatoes
oh hecc no im on a special diet cos my fugged up intestinal track
Don't worry, I didn't understand either. I think Clyde got to 'em.
Also, 8k by 12k res is sure stressing out my PC
Oh oooof

It did
ITS THE MATRIX
CLYDE IS THE MATRIX

Crohns high five here 
Fair :o
I don't even like alcohol, so I can't relate XD
THOUGH

eh i dont like alcho either
weed was for anxiety but ive been smoking it for so long it doesnt work the same way anymore
Though also I feel like sometimes, I feel energized after drinking em
And sometimes
I just get sleepy
Somehow
das weird
What you mean? :o
It do be, sometimes happens with coffee too
I like the taste of both tho
58 MB image
Scan this Cylde 
build up a tolerance to it tldr
Makes sense then
Time to wait for it to send 
It is very close

Lol


floyd is nice

How do I actually send the pic 
MAYBE
Turn it into binary
And send it encrypted in morse code, then encrypted in another type of code
Like a puzzle
Lmao
i mean u can dm it to me 
just be careful... I'd hate to see you
from rule #7
I could :O
I mean
True
I just gotta tinker with the pic in SD until it works
Maybe inpaint everything below the neck into a big beefy man
Same, probably like, 125 GB model
O
YA
Aight, I'm gonna sleep. I leave you all with part of the preview card I gen'd for deepnegative. G'night folks.

Gnite tho 

Wait why me 
No, only Yumeshiro xD
FKU CLYDE
I will destroy this robot if I ever find it on the street
14 blocks in a row so far

20+ blocks so far
Maybe I should make a sexy skeletons LoRA
Not having much luck today are we
Today, is a day
of tragedy
I hope my lora test goes well
There is no further pain than existence
LMAO
this has given me an inspiration on what to generate next
I am glad my pain feeds your creativity 
Pain is eternal after all, at least you can get something good from it 
And some of the are SO good too
Of the rainbow ones the 4th, 5th, 6th, 7th, 9th, 10th, 14th, 15th
Those all are so good
But none of them send 
That is N°25 of all the ones in total I've made so far
Rest in peace rainbow soldier

And you'd think it is the cleavage, I've made the cleavage disappear, an it is not the cleavage 
Clyde will remember that
do clyde learn from user patterns? that seems scary
I've made it disappear completely with inpainting tho 

Cylde is just schizo
Its not that deep
Just old man clyde
I mean last time discord official said they were going to integrate gpt to clyde or some shit
gpt in skyrim is scary
GPT in skyrim looks pretty dope
AYYYY
After removing all cleavage and skin, one finally sent
Its
Mid kinda
But it sent at laest
NOW this one is cooler, worse hands, but background is pretty great
Bad hand, but this one is amazing :O
Maybe I should add fashiongirl LoRA tho, face looks a lil round at times 
I think it just made her paler, which would make sense
Added my Chelodoy LoRA, I think that LoRA is like just OP
The dataset should be a mess, in the AI's eyes
I don't know why or how it works so well
I feel like it makes image quality and hands better, but its supposed to be a style LoRA
do not question our ai overlords
Fair :O
ALunarDreamV2.0
mokunoki matcha Hiten fajyobore
I definitely see Hiten's style there
Looks nice
That is a cool style tho

With what model can i do this
Any model I assume
Just try from behind, squatting
Seems like someone just using style LoRAs
Oh, well, if you mean the poses, then yeah, any
If you meant the style then that's kinda tricky to guess

Is to just look for a LoRA on civit that is matching
Or use a few loras at once that may be able to recreate that style
Or make your own merged style
That's what I did anyway 😵💫
Man, it's a shame about that hand.

AI trying its best
Missing will smith
Someone needs to teach these waifus to stop putting their hands in the blender 
They wanna be silly
Hello there
General Kenobi
I haven't used an artist as reference in a prompt since like 6 months ago
I reference styles, but not people
Mhhh
Which models should I test the new neg ti on? I have dreamshaper, macaron-mix on the list, anything else?
got a new negative embedding?
if so, i was testing some i found on civitai, i'd love to test some more
@wispy canopy
It's still WIP but it seems to be a "breakthrough" in terms of not changing the style to much
you did a lot of training, or had an eureka moment?
A lot of training, changing dataset, manual tagging, changing models. Alot of stuff
you can try my model, it is pretty creative, only... a bit horny xD
Lol
but, creative it is!
@queen void confident enough your embedding can fix this?
Maybe
a couple kissing in front of a heart shaped star filled sky, stuning fantasy 3 d render, profile picture 1024px, connectedness, a person standing in front of a, hands touching light drops, radiant flares, utopia profile, highly upvoted, interconnected, new release, song
this is my LLM going loco on a prompt, dunno how this prompt will do in a normal, sane model
Eating breakfast rn will start testing in a bit
Hmmm what is this saying to me xD
Well, lets just say, I also think she looks very good

Fix the what? :o
check the hands xD
funny, how different negative embeddings have clear preferences on sampler/prompt, but for another it's totally different.
dreamshaper works decent

Have an appointment now, but after that I would give some people the neg ti for testing, if some is interested in testing it, just dm me
@queen void yeah, it untangled this image too, pretty well even

sometimes dumbest prompts are best prompts!
melonmancer goes shopping for... well. melons.
it just had to bork the hands
sorry nerf, seems melonmancer is destroyer of hands 😄
sometimes when i finish an image i think to myself. how can i surely know the image i have created is not like a trashy duplicate of something that already exists. you never know with these anime checkpoints. 
well, it'll never be an 1 -> 1 image
but my model was made out of trash, refined with the great stuff this community has, and then fouled again with my vile prompts. Unless you're saying melonmancers are an actual thing, doing something "first" isn't bad
we also had a lobsterpunk phase around here
so. just keep experimenting, all the things you're creating are all uniquely you, and the more you experiment the more it's going to be you!
i mean i guess thats true but at the same time im scared that im just copying someones hard work without knowing.
but its not like i make any profit out of ai or something
Experimented with stuff. I can now use multiple characters from character makers in Blender. Then using depth map and color control, the result gets decent hands and same characters consistently:
Initial input image rendered in Blender:
Output:
It gets the 3d look out of the way. Unfortunately, it thinks the shadow is a pattern on her clothes. Shoes are consistently different. But that stuff can be edited manually.
Well, there's a lot of things that went into the models, and by now, i think there's going to be artists who would've liked it that their stuff wouldn't have been trained on
but look at it from the other way, the artists in question cannot possibl draw/create all the things you might want to see envisioned
and in my case, i'm just using their number representation to cause 'glitches' with other numbers, in hope something interesting pops up, totally unrelated to the original artist in question
in the future for certain, but for now, SD would treat it as a whole image
you can "cheat" for a bit with maskings and regional prompts
but one-shotting stuff like that is still pretty hard
a giant collection of 15 melonmancers in various styles and one.. ugh... MELONmancer
#7 and #9 tho, sooo cute ❤️
Yeah pretty much what he said. You could start off with a character as close to what you want as possible, and then use lower denoise settings. There is a custom build of Blender called Goo Engine which is specialized for anime. I've never tried it but if you can get close to what you want on something like that, SD could help it from there
or a really rough sketch and then inpaint away
thats a great thing to keep in mind. i think your reply made me calm down a bit about this stuff^^
i can give you an example -> the lowest right of my melonmancer, the girl with the personal melons
the only thing in the prompt that causes tha tdeviation is George Asakura style
however, when you look up that dudes style... NOTHING like that
the takeaway:
A) george asakura did not. in fact. ever draw melonmancers. neither did picasso
B) It doesn't even resemble his work, but that specific name caused that outcome
now, if you use lora's to imitate someone's style, that might be a different story
next up, carrotmancer 😄
evil carrotmancer lol
lets say i use 2 different style lora and make them weigh both 0.5.
the outcome of that style could not be traced back to artist x or z right? . so would that make it something independent and completely new.
can you trace the images i'm making back to whatever artist would've made them?
tomatomancer in library. case in point.
not really no 😆
of course, if i add artist style lora's, that's a different thing
just straight up ripping a single artists' style is not done, but that's my personal oppinion
but when you start mixing up lora's and all that stuff, to make something that's uniquely you, it's something different
if those two lora's give you a good feeling in terms of style, just go with it 🙂
strawberrymancer!
meh. my model forgot a whole lot of things, including a lot of vegetables... in favour of other... juicy things.
meh
it's not great at anything 
i mean. these are not strawberries. you do get mart tho 😄
geomancer 😄
wow, aren't these just adorable?! ❤️ ok, my model is redeemed, this warms my heart
Could there be a reason why inpaint decides to generate what ive masked AS WELL AS the rest of the image?
Ive checked and rechecked, i am directly on the inpaint mode. But instead doing one area its doing the whole image plus that area
Did i misclick on a setting somewhere?
Correction
It thinks the eyes are masked but actually theyre not, its just the mouth, so i still dunno wtf is going on
inpaint padding going wild?
maybe send a screenshot or smth so we know what youre doing
Nvm. Idk how, but after editing the image in pixlr a little bit, i saved as a png and loaded it into the inpaint box. Strangely, the ai reads the image as if its already been masked
I went back and saved it as jpg, and that fixed it
I dont get how png or jpg filetypes affects the inpaint process but ig thats one way they do?
I felt my description was apt enough, my apologies.
All good, just saying seeing what you can see helps sometimes 
Ai...
chibi streetmancer!
gravemancer. look at this smug face xD she knows who she's making the graves for xD
?
Bang!
id ask in #🤝|tech-support 
I turn off xformers and it's working
then id reinstall xformers if i were you
Bruh
I dunno for how long
But my clip skip has been set to one this entire time

Wtf
That could explain so many things
Christmas Date!
rip pixiv
what happened to pixiv?
cant post pictures that copy someone elses style anymore
bruh

how do you even copy right a style?
or like how do you even determine what style is whos
Wut
Sexy

👋
Now make her smoke weed instead of cigarette

how do i even depict that 

hey ho
By saying smoking the devils lettuce 😂
bloodshot eyes?
that might work
how are you? :O
Maybe throw eyes half closed in there too.
very busy working on master's thesis
if smoking weed make me like that, i don't want to smoke that
Hahahaha
I could be using this time to play games or gen more images
I would generate more images but I am currently at college 😭
her
are getting bigger tho
oh, okay. :O gl
@lyric zealot https://youtu.be/U1ei5rwO7ZI
Snoop Dogg Smoke weed every day (dubstep remix)
Download the music here : http://adf.ly/nspln
Maybe that'll help
hey
Ah yes 10/10

actual good editing
is there anyone create an anime hero using indian mythology
if you want to do it yourself, i would recommend going #1072220168534642768 and reading that
Not gonna lie wish that was an anime lmao
lmao yeah, but there's blend s if you wanna watch that

I’ve already seen it 😂
https://www.youtube.com/watch?v=bhLmC1cqRIQ comparison
Snoops is better ngl 😂😂
yup definitely 
@lyric zealot Can you make her hold a bunny rabbit ?

kinda have to rng a little
Hmm I hope it's safe enough 😅

Haha just ask and make sure
I think it's fine, but don't go any further
^^
I'm not mods tho
I would say self report and ask 😂

Haha but ur a well respected individual from this comunity and people listen to you so …
😂
Haha
but well, I think the above is fine, but just don't go any further than that
or just ask a mod
I figured as much 😂

she stole the vegetables from my veggiemancers!
I think it's fine, just think as if it's a bikini
Haha MOVING ON
it's hers! 
Nah it’s the bunny rabbits food

stole from the carrotmancer!
or purchased at the local fruit/veggiemancer stores? 😮

mine are peaceful 😮
she looks ready to fight
we are adding support but at 1024x1024 it is likely only for 16GB+ cards initially
got time for 1 gen, then i have to go, hope something useable comes out, for the veggiewars...
this is the only valid way to communicate here
he's busy doing his thesis
wait fr?
later tonight, more veggiemancers to fite!
what thesis he doing?
cya
cya
does anyone mind doing that?

oh damn
@lyric zealot do you mind doing that or should I do it myself?
I havent even start training anything yet

So you should probably do that yourself 
good way to start
I'm too lazy tho
yeah alright
I am just too too lazy to do anything except genning
but I'll send this in #🚀|pow-discussion just in case
good luck 
let her cook
Cuuute
jessie
Welcome to the community 
HAPP!
Welcome and what an entrance to the community
ok I'll do that maybe after I finish my current lora
AHAHH thanks, i'm new as heck to all this but having a crap ton of fun
That’s the best way to do it
amazing starter tip from me is: best way to learn how to prompt is to steal borrow the prompts of the images you like. for most images you can simply download them and put them in png info in your webui. if it doesn't show anything then you can ask the one who made the image for the prompt and they will most likely give it to you.
Always ask for permission first is what I would say
Lmfao
I borrowed forcefully from Holokuki, and thats how I prompt now
Works
10/10 reccomend
holokuki very nice
yeah that's pre much how i've been getting most of the loras to work, search up other examples of people using the lora and just apply it. big brain af.
It do be, really organized too
Hands better now?
Kinda
also why am i putting commas after words when i'm typing normally now ffs
What's changed haven't looked in 4 months or more
https://civitai.com/models/40199/aurora Has good hands
https://civitai.com/models/47085/envybetterhands-locon There is a better hands LoCon which works well
And @queen void is working on a negative embedding to improve hands too
What's this civitai
I don't see you prompting on #1100170312106127410 
Here is some good hand
Mmmm some good stuff, wished i had more time doing this
I just woke up 
Good morning 
Gm!
i think it works with text as well
nobody doing anime stuff there 
MIKU
is it a challenge? 😄
That locon thingy is a Lora which needs an extension also?
Wait is this a competition?































