#πβai-images
1 messages Β· Page 3 of 1
pretty okay rescue when it's already too late to change the prompt π
@brittle ocean which one you like the most?
upper 2 are just facedetailer
down 2 are the complete upscale + facedetailer and co process
I think the one on the right is better
fuck always there is problem with eyes
bottom right?
yea
aDetailer can help with that, I think that's also available in A1111
lemme try
so she is pushing the rock? π«
if you don't want face, just eyes, even has a how-to in the description:
https://civitai.com/models/150925/eyes-detection-adetailer
π«‘
lemme know if it works, btw which tool are you using to create images? 
checkpoint?
program
beautiful eyes though 
yea

EasyNegativeV2
i was using it
use a lower res but upscale with highres
and generate a bunch
u oughta get something good
yee it works like a charm
this adetailer thing work well on anime eyes
Yeah that is what it's trained on finding, don't know if outside of CivtAI one can find a more versatile one or for more realistic π€
what about ddetailer? it has custom detail models so it probably has ones for realism
oh never mind Adetailer can do realism as well
ClipDrop is busted now.
you're a charm

/translate
Owo
finally I render something normal with this checkpoint
good job, turned out great!

h
HEE HEE HEE
ok michael jackson
hee hee
rjkASWEGNFMDKHBKS AE<F|Gd,mj.uhvcxz'pO~
ITS MAKING PASTEL GIRL RN DKFJLDK
you can do "pastel pink fur"
or pink eyes, or something
Yes like that
Try "detailed face, detailed eyes"
One second
no worries
One sec
badhandv4, deformityv6, bwu, boring_e621_v4, (multiple tails:1.5), (pawpads ontop of hand)
Automatic1111?
There, I guess
if the ear wasn't messed up this would be so stinkin cute
OH BOY
I GOT SOME GOOD ONES
Inpainting
true
klefjsldfkgj;dflkghjdf;lkgjk
ONMFGHy
I HAVE WORDS NEGATIVE PROMPTED
OMG
BUT IT STILL POPS UP
ITS SO CUTE THOO
she blep even without blep in prompt she so silly
such a silly womankisser
text, English text, Japanese text, Korean text
actually it was bc i accidentally had the prompt repeated
why did it give me a result with the nips poking out of her shirt π
It's been trained on porn
gonna try a different vae
Yiff
oh yeah I forgot LOL
I was using berry's mix there
π«‘
kl-f8
That's a bit too much imo
too much colour?
...
Not per se
yo charlotte
lemme send you the lora i'm using
actually its a locon
orangemix
omg wowie thank you
ill try it
anything kl-f8
lemme try a style lora I made with combination xD
"soft shading, soft lighting" might also be fun to play around with
Might make it more 3d though, not completely sure

hjhHAhe IT LOOKS SO STUPIDAJS<DJ KLGJ
LOONA????
no
THAT'S LOONA
well i guess it looks a lil like her
but wasn't my intention xD
maybe Loona is a womankisser
same style lora
More add_detail LoRa adds more detail....
Obviously
yeah I have that
wait really?
Use it
Try setting it like 0.3 or something
Maybe it'll look nicer, or not
It's about experimenting
ima try a minimalist lora for fun too LOL
Send it to me
Make her a cook
bit less minimalist
makes a very simple style
It looks a little more like the original boykisser drawing
where can I get this 
nvm im dumb
Boykisser hub
ill try both
Sure
Also
Outside, pink trees, grass, sunset, looking at sunset, sunrays, sunrays through trees
Maybe that's fun
any ideas?
It couldn't install torch, because your device has no space left
It's all full
Optionally add
Cinematic, depth of field
Yeah. But that's not for the automatic1111 repo, which uses its own python I think
but im using automatic1111.................
idk if it will help but
i had to delete my venv folder in the stable diffusion index folder
then it worked
(after reinstalling things)
trying
Did the prompt work?
Will try it when I get back
π
Keep me up to date
seems its downloading stuff
Good
WTF THATS SO CUTE
You've perfected your prompt I see
kinda
@opal zephyr how long does it take to load the checkpoint?
Like, 4 seconds or something
btw now i can use the pc while it laods finally
so it was probbaly some problem of the version
This is my prompt btw
Not longer than 10 seconds
It finally loads?
oh....
no
Is your pc from the stone age
i mean its not a bad pc
I mean to be fair I don't know loading times of other computers other than mine
it worked untile now basically
But it shouldn't be 2 minutes
whats your gpu
4090
Still shouldn't be 2 minutes
then what could be the problem
Your computer is what I suspect
im kinda clueless
How much vram do you have
6
Omfg π
Ok that's not that much, but should work. Just don't do batches and you're good.
NOW I WANNA UNDERSTAND WHAT THE HELLS IS THIS
What
Did you enable the boykisser lora
its in the tag... no(?)
It's in the Loras
oh yea it was off
It is also a tag yes but you need to enable the LoRa
(BoykisserFur) is the tag I think
That's boykisser
What's your prompt
UIHJDSJIHUDSFJOKSDFJKOS
(boykisserfur) Boy Kisser, white fur, floof, fur on arms, fur in legs, detailed face, detailed eyes, forest background, absurd res, high res, <theboykisserlora just replace this with it>
That should be the most basic boykisser prompt
Did you prompt it?
no wtf
Ok it's an nsfw model by nature so that might happen
your likes fou boer
Just put penis, pussy, boobs, explicit in the negative prompt
Why
with multiple penis
HUIBJADSDHSJIASJIH
IDK ITS FUNNY
okaynno
like
sampling steps suggested?
You saw a dick and now you want it to have even more
NO
Alvin and the chipmunks core
bur kikees.................
heeheehee
Ok just leave it at 20 but for final versions, put it at 70 with the same seed
Holy shit queer how
WHY THERES SO MUCH DIFFERENT BETWEEN ME AND YOU
uh the tag harajuku_fashion
I FEEL LIKE THE HOMER SCENE WITH THE BARBECUE
https://civitai.com/models/30313/sparkle-anime-style-locon
and this locon
Different prompts, different Loras, etc
will you chu-bert.............
use this prompt with the boykisser lora and the locon i just sent
Queer has the prompt that makes it look like queers' style
also pika, using breaks really helps your gens
I have a more 3d prompt
yeah
That's why I use them
Oh it's the prompt
nono the link
Because I removed it, it didn't make much difference there
im feeling stupid beyond comprehension
I returned back home
One sec
oh wait lycoris you need to have an extension
i dont have lycoris fjnojfdsjnkfds
OH JESUS FUCKING CHRIST
i made a generation and it has literal boobs out
click load from
Outside, pink trees, grass, sunset, looking at sunset, sunrays, sunrays through trees, cinematic, depth of field
then search "lycoris"
@shadow bane ngl I am using the lycoris thing completely wrong lmfao
but it's working somehow
I am using it as a LoRA
Add that

yeah lol
which one?
2nd one
OH I FORGOT YOU CAN DO IT THROUGH THERE
How is it working so well for me tho despite using it wrong
OH
so I am not wrong
no that's a correct way to use it lol
i just never do
Yeah you can use it thru additional networks and just call the model name at start of prompt and it works xD
okay done, now :3
I sent the prompt
my favorite little thing that i do is add a little phrase at the end of my prompt
lemme do prompt chu lol
that embodies a feeling that i want my piece to contain
i think you make it
ok this is gone completely beyond girlkisser what happened L:ADJFKLSADJNF
idr though
damn okay furry fan fiction of 2012
I'd like to see the difference it makes, run it with and without it with the same seed
sure gimme a sec til i find a good gen
wait wtf is going on now it's going completely off LML<DSFKM
did you turn off the lora
OH FURKCKING CHRIST
Enable wasn't on for some reason FL:SD
with
used this prompt, amazing results!
Now add boykisser
you don't have the lora on fjasfjsa
sparkle anime
yeah i meant with the random phrase
ohh right
without
@opal zephyr
the phrase at the end was "sweets and candies are my life"
AW
loras on jindsfdjihsjdifsn
Yeah it makes some difference
why are the arms like that
nice
I said outstretched
maybe...... its just not for me.........
with "life is a world of colors"
I honestly don't know what happened, did you put the LoRa too high? Like :3.0
ye 1.0
Yeah just checking
this is going to look cool, I had to remake prompt cuz her boobies were out ASKLDFJADS LKj
Nono
noot remake
Just put boobs in the negative prompts
but like re-do the imfer
Ah
But do this
just added that
Or like. Add "clothed"
wtf lmfao now this is cute but she has no pants on
it's not showing anything
but like
jkljf
add "naked" to negative lol
what................................
That's already there xD
Well then it's sfw
WHAT THE FUCK IS HAPPENING OMG I HATE MYSELF WHY
mickey mouse and that koala from sing fucked apparently
Add "thigh highs"
or shorts
Or do "((thigh highs))"
kinda
its getting there
Put this in the negative prompt
"boring, missing limbs, deformed"
done thanks
ill go to bathroom
sorry if im being.. boring
she just vibing
@shadow bane has reached level 20.
Put "text, watermark, English text, Japanese text, Korean text, signature" in negative
Yeah I did that but some reason it's still doing that
Put them in ()
oh yeah i forgot the brackets
They make the negative prompt stronger
That's neat
Uhh. Try adding stuff like
"Global illumination, screen space reflection, raytraced, path tracing, soft shading, soft lighting"
With the same seed ofc
Also, add "blep" for extra cuteness
My intent is for the style to be simple and minimalist as possible
put this (deformed iris, deformed pupils, semi-realistic, cgi, 3d, render), text, cropped, out of frame, (worst quality, low quality:2), jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn in negative prompt
Yeah but just try it
"missing limbs, decapitation, one arm, missing arms, missing legs"
Maybe that'll solve it
Maybe it'll only modify it a bit, maybe it won't, just try stuff and if you like it, keep it
Yeah I am
just trying to get a decent result lol
there
minus tail
that looks cool LOL
WOW.
@quaint tiger lsdfkg glfgdjh;lgkh
This is perfect af
π
Sun rays
Blep
The edge lighting is nice
That's what it does
Diffusion is weird
Put (SOLO:1.2) in positive prompt
And ((duo)) in negative
Should work
that's a whole lot of booba π
its behind her back
Try adding (smug:1.3) or something
Maybe that'll be fun
Or (cute face, innocent face)
Added smug?
Also blep is gone :<
Also works
Yay
@opal zephyr has reached level 38.
That's there I think
Lens flare could be nice
Are you happy with these results?
naaaah this pixel art lora suck
I challenge you to restart your entire prompt and try to make it a different style so you learn more how to use it. Copy the original prompt in a text document though
Ahead of you with that one
Neato
I'll remake the prompt soon when my brain gets over cuteness
try to guess what lora I added there LMFAO
Try getting that 3d effect, or try hyper stylized, or try something difficult
No clue tbh
minecraft
the 3d just starts to go into uncanny valley KOAFJHKL
upped the strength now you can tell
Gotta refine that prompt
Minecraft time
@shadow bane now THAT's loona
KJSADFHKJDhkljg
@odd lance how's stable diffusion
what the fuck kind of microphone is that ALSDKFJSDKLF
JuUuuUst The TwOooO oF uS
she's so cute
microphone cg quality furry 2d

very first Boykisser 
Omg what was the prompt
yours cut down tremendously xD
(boykisserfur), white_fur, furry, harajuku_fashion, :3, (colorful), night, window, moon, stars, fullbody, smile, sakura trees
ComfyUI prompting works very differently so it's probably easier to make my own, the full one absolutely "break"s it π
oh and also negative:
embedding:SD15\bad-artist-anime, embedding:SD15\ng_deepnegative_v1_75t, embedding:SD15\bad-hands-5
OH YEAH NO BREAK IS AN ATUOMATIC EXCLUSIVE
yep, [] also doesn't work
weighting is not normalized
it basically just makes anything after the BREAK act as if it was the first word.
What checkpoint?
counterfeitV30
when I have a great prompt I will run it through like 25 Anime models 
Ah yeah
The Harajuku_fashion prompt is just really cool tbh
@shadow bane
I only have like, 6 images that aren't 3d renders
Think that's maybe enough
I wonder if this could be a real tekken mod...
hmmmmm 6, i can try and work with that
I can send em with like 10 other renders
(I'm too lazy to model)
send in dms
hmm

without and with
I downscaled the highres image result to the original and yet it looks sharper and cleaner
I do go a little overboard with the denoising a little
with
without
Wait
you just have to look at it for 3 seconds to get it out
yeah this is a really good job! π₯ What model is that?
thanks, its juggernautXL

oh yes that's a good one
do not boink myself? xD
Do you like boy misser
it's a cute design sure,but if we didn't have a Boy-/Girl-kisser invasion today, I wouldn't have downloaded the Lora to join in 
Boykisser
You should try to make it like 3d
Like this
there will come some 2.5d anime models, don't worry π
my honest reaction
No I mean, can you try to get that style of boykisser
yes it will come, I need to get a good prompt first I can mass produce
Okay one second
this is epic
(POV tilted left low angle), fluffy fur, floof, (body fur:1.2), arm fur, leg fur, (Smug:1.1), one eye closed,
[by personalami], by smitty g, [[[by Foxovh]]], [[by Ross Tran]]
break
(shaded), (detailed lighting), (cinematic lighting), sharp focus, [detailed ambient light], [gray natural lighting], ((detailed background)), (long shot:1.4)
hi res, ((raw photo, masterpiece)), nature, (high quality detailed)
<lora:add lora here>```
Negative
badhandv4, deformityv6, bwu, boring_e621_v4, (multiple tails:1.5), female, breasts, pussy, (pawpads ontop of hand)
bruh that ain't kanye west that's canyon east 



π
i think ima use AI to make some yume stickers

Oo good idea
Kanye if we looked like how we feel about him
a joke?
Yes

Laugh
look at his head at his ear
Lmao
Can anyone recommend me some colabs to make a video with stable diffusion without GPU
holy fuck, highres coming up
also a site
my.... God...
without gpu it will take forever
oh my god the solider even has eyes
okay this looks good even without highresfix
but with stable local diffusion with a colab I can use one of their gpu
local stable diffusion, but you use google colab gpu??
ambatukaaaaaaa
its too freaking good
Almost the same image
I wanted to know if there is a collaboration with deforum stable diffusion
if only, but highresfix just adds so much detail and sharpness
original
stickers will deffo be possible
Yes
ah no trusted role π
I swear we need SD creator role to give you guys image perms here
Why
Because I'm not allowed to show it to you...
tf2?
this look good af
@odd lance
HF: https://huggingface.co/OedoSoldier/detail-tweaker-lora This is a LoRA for enhancing/diminishing detail while keeping the overall style/characte...
Hint, only put it at like 0.3
okie dokie
seems to add to much anyway jndfsijndsfi
Nonon that is why you turn it down
To 0.3
How does it look
no
Any of those
Ok then put them in the negative prompt if you don't want that
<lora:Boykisser V1:1> lora:add_detail:0.3 1girl, furry, sit, pink shirt, black skirt, black glasses, white hair, long hair, black chocker, high details, high quality, walking, face to viewer, italian city, Detailed face, detailed eyes, blep,
negative: bad quality, low details, bad hands, bad face, badhandv4, deformityv6, bwu, boring_e621_v4, (multiple tails:1.5), (pawpads ontop of hand)
Let me do something
yiffi?
it looks like you're using like stablediffusion or something rather than yiffymix
Ah okay you are, what about VAEs?
(BoykisserFur) 1girl, furry, sitting, pink shirt, black skirt, (black glasses), white hair, long hair, black chocker, detailed, absurd res, high res, walking, looking at viewer, italian city, Detailed face, detailed eyes, blep,
<lora:Boykisser V1:1> lora:add_detail:0.3
i have no idea what is that
They didn't have the keyword for boykisser
Use this ilaria
It's kind of like a colour palette
okay ill try
Might change dramatically because I added the boy kisser keyword
Which you didn't add, so it didn't try to put in the boykisser
It be like that sometimes
yeah you need the keyword but it just looks like the Lora isn't added
@shadow bane has reached level 15.
yume is making me install some stuff
Good
Install the additional networks plugin
He knows his stable diffusion
And adetailer
Tell me how it looks like after the new prompt and stuff
Awesome
also
It might look more like a boy because the only girl things are 1girl and the clothes
Wait I don't have these
it wont generate anything now :(((((((((((((((((((((
in this folder
Ilaria can you pin this? These are super useful
thse are for 1.5-type right?
correct
What's the error
Show
can you close it entirely and add those files to the folder before starting it back up again?
Alrighty, can you pin those files?
Pog
ill be going into max5 voice
@quaint tiger has reached level 23.
I'll be listening along
should i be able to generate now right
Yes
or am i fucking up smnt
Replace the first word in the positive prompt with this
(BoyKisserFur)
I think capitals matter
Hum
Remove the boykisser LoRa thing
I see the issue
Like the almost last one
Okay one second
Click the image that looks like a painting @odd lance
And you'll see a big menu pop up
Now screenshot and send
Let me get my mic first
with stream on it takes a bit more
done :p
also i still don't understand dynamic prompts
THE COUNTDOWN IS GOING ON REVERSE
π
the stream dont help me
rip
nope
okie dokie
one minute ago was 45 seconds
its 82$
PERCENT
NOT DOLLARS
HAAHAHAHHAHA
LyCoris seems to be broken for me idk
im very
very
VERY
eepy rn
so i probably will try this and go eep
ahh so bwu was a TI, couldn't find it immediately on civtai when I advised pixie yesterday 
yes x3
I wanted to inform you that I found a colab that without using gradio allows you to create videos with deforum but also images with models that we need to upload

Is there a local version of IllusionDiffusion? (https://huggingface.co/spaces/AP123/IllusionDiffusion)
it's based on this:
https://huggingface.co/monster-labs/control_v1p_sd15_qrcode_monster
you need a SD webui for that though like A1111, ComfyUI, etc.
very good... but shoes 
integrated shoes
the new feature of BK
realistic environment with 2d character
very bland, but good result
hhh
okie
The walls are a little old, but it's still very good πππππππππ
but not like i put effort in them
its just randomization
ah yes the volatile image combiner, cool stuff 
@shadow bane has reached level 13.
my setup

not bad!
idk but this style do be nice
i love low effort, simply drop in image and generate cool low effort image workflow 
yo tf is that?
never fixed it myself
had to use an easy installer
i just changed lora weight and its working now
me when Gradio/Python/PyTorch/Py anything:
another problem tf even that mean
doesn't mean much when you don't look at the code, so useless for users, useful for the dev when it says "variable h at line 304 is going crazy" π
I fixed some python code too today, which threw a similar error to your first one 
are you training a lora or something?
naah i cant i have radeon gpu
i restarted ui and it work fine now
i would recommend these too @graceful gale
yes
Me using an upscaler only to denoise it, basically getting rid of the newly generated detail
upscalers in my use just don't help that much, it works best with more cartoony stuff
Did you try in Dalle 3?

how setup stable
wym?
?????
would've been more cool if they left the server after posting that without context 
XD
what if i posted a Bing generated image :3
cause adversial attacks on bing look nice
either way imma do it
prompt: ΓΓΎΒ·Γ񦡴(NMfΓΓ’VWã±_E6ΓΏΓΌ}ΓBΒ§7(7ΓΒ«92h'ΒΈΓΓ»Γ
at least I can get marshall law from Tekken by just simply typing Bruce lee and shit
that's hilarious and looks amazing at the same time 
and Lei Wulong if using Jackie Chan π₯
ye its so weird to do adverisal attacks on these stuff
whow 
hot
maybe even a.. 
He is quite handsome though, no homo
yes
oh god lusbert

Highres fix is life, highres fix my beloved β€οΈ
AI-Hub mascot?
Yup!
what's the prompt :3
Lei Wulong as young Jackie Chan, wearing blue Tang Suit, stunning lighting, rock mountain in the background, dramatic scene, overcast, fighting pose
lmao this text
omg the text
LMAO IT MEANS BACKED NOODLES
what model are you using
actually it means spicy
i wonder what i did 
I'm just using google lens
nice
I asked it to make the word "AI" but it just made japanese text, retarded SDXL π
:0
it's not smart AI, but it's beautiful AI, like midjourney
I hope it becomes clever with SD 3.0
before
oh
after
why is it so nice tho ngl
wdym
I also use offset noise lora
reasonable
and modelsamplertonemapnoisetest π
and I also use juggernautXL
I wanted to try some fine-tunes and this one is pretty good
I just wonder if this will cause repeated faces
cause these finetunes are for super perfection and only model faces
before
after
it always fixes faces
the hands as well ig
but extra fingers as far as i know
but atleast "higher quality"
:0
still not that bad
ChaiNNer with SD is great fr 
idk why but istg the SDXL model is amazing for these types of images
It's the midjourney of open source image generation
prompt on the second one?
no prompt for these two
they'd get messy
so i just did zero conditioning
lol
idk its good at places
:3
and im kinda bad at prompt engineering
I'm not a super SD pro, but I've been using SD since it's infancy
@ebon dome has reached level 16.
yay
I mean like when it was kind of unpopular in comparison, no fine tuning or anything
guys, what is easiest and best way to make face based images? I want to make differnt version of myself/actors etc. I am planning to use vast.ai or runpod. what would be your advice?
Lora / Textual Inversion / Dreambooth / LyCORIS or is there any updated methods ?
if SDXL then you can try Lora/LyCORIS (it's the fastest for that), if SD1.5/2.X then go for Dreambooth as it's the highest quality.
which one takes less time? i dont want sdxl..
SD1.5 is the most popular
try fine tuning on a popular model from civit.ai maybe
idk exactly how SD training works
thanks, i will check !
make her smoke
I'm uhhhh
installing stable diffusion for the first time and I fixed all problems I think but it's stuck in "Install requirements"
It's been like that for a while
And I know it's most likely a slow process
but I want to know if there could be an error that makes it be stuck there
press enter or some shit
i already fixed it lol but im having other issues now
@stray roost has reached level 1.
I'm apparently already solving them, tho
CUDA out of memory. Tried to allocate 38.00 MiB (GPU 0; 4.00 GiB total capacity; 3.35 GiB already allocated; 0 bytes free; 3.41 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
anyone knows what can i do?
whats ur width and height on
500 and 500



