#🍥|anime
1 messages · Page 112 of 1
Stable Diffusion web UI. Contribute to AUTOMATIC1111/stable-diffusion-webui development by creating an account on GitHub.
look for solution for Green or Black screen
i have to sell my vr headset but it doesnt have a hand controller and ive never used it
guys which is the best of the two?
this:
https://github.com/imrayya/stable-diffusion-webui-Prompt_Generator
if anyone tried it
i installed both earlier and
im gonna keep whichever one lets me download the models manually and place them manually
ptomptgen seems to have more options and sliders
they seem to be identical for the most part they both use the same models
promptgen lets u run the model on cpu though. i noticed i maxed put my gpu memory when using magic prompt earlier
Thank you. I'll try the last one
let me know if u find an easy way t o install the models ><
or if u end up using any of them and have cool results
what do you mean? 

It has to be smol and sus
what does that even mean?
And smol sus

You know
that sounds super sus lmao 
Prompt: 1girl, petite, red hair, looking at viewer, horse, riding horse, helmet, gun, holding weapon, outdoor, clear sky, sunset, cinematic lighting
Pls try
I have no pc
Rn
Cause in japan
my pc is literally dying
Excuses
Such a cool image tho
Where is the weapon
Wdym
starts giving horse red hair,
yup yup
the tail is kinda wacky
Closest one yet
i don't generate stuff without adetailer now
which 
This
It kill me form inside
You tryna murder me fr
What you playing?

Abriss: build to destroy
Whazzat
omg i want abriss
been following that game since i heard about it the other day
i keep forget 🏴☠️ exists
hm I somehow goofed.
each one is SLIGHTLy diff
hmm but so small ya can't tell mos of the time.. sihg
ok I made it into a style
how long does it takes to render a 2048x2048 image with 8Gb vram xD?
don't you just oom then?
Pretty expressive stuff there.
WHAT THE HECK
are you feeling okay xD??
I mean it's not impossibly bad I just take longer to do stuff :p
And some I can't because AMD is not very good or at least my card
It takes like a minute to do 512x768 images :p
OOF
bro you've got lots of patience
if i had to wait 1min to see every of my images i'd die
yesterday i generated 60 pictures before getting a good one lmao, it would've taken one hour
angel save me
Does this woman have a name?
How does this lady keep on being fabulous?
Завали ебало
Her visual Aesthetics kind-of remind me of Rin Tohsaka from Fate/Stay Night, in some ways, she's missing more red clothing though.
The hips on the other hand very much reminds me of Rin's, not sure why.
she don't really remind me of Rin tho 
Personality-wise, anyways correct?
what got banned???
wore the wrong helmet to battlefield
ah, sd webui got banned from google colab
The Rinners have arrived.

not yura... uhh...
ok, that's UmiAI problem. sd-dynamic-prompts works well
yuri zahard
yeah, that one
she's amazing 
true 
Oh, fair then :O


Merch: https://posix.store
Python programming language
Interview with a Senior Python developer in with Dr. Harris Dlacc - aired on © The Python.
Programmer humor
Python humor
Programming jokes
Programming memes
Python
Python memes
python jokes
uwsgi
conda
pip
pip install
venv
easy_install
django
#programming #jokes #python
This vid very good
Very funny
Oh yeah the sequel
I swear you doing it on purpose now

well, just posting whatever im generating, that's all 
But almost everything is really good

glad you are enjoying certain stuff 
Nate Four merch? 
It really struggles with making her hold a weapon
i have no idea what's that 
Me neither 
There is written on the shirt "Nate four"
oh you meant that text on the shirt lmao
i literally google "Nate four"
Yeah xD

Bruh

RVC colab also got banned
rosepetals underwater... ok?
She on her period
she's cold
Who's that person there?
waifu # 1
ayo you using some addons to make those light effects?
no, that's purely the model
is it your model?
i've been having great fun with it
she looking angy
she so angy, her iris starts turning color
Am I the only one here who does 1 batch each time?
nah, i love doing batches 🙂 gives me a feel if the model understands the prompt/concepts
me
I just wanna make very slightly tiny changes each time
And I don't wanna wait even a bit more for more options
you mean, 1 batch 1 image?
Guys what do you think about this?
Which navigation method better?
what about it?
i have no idea how android works so....
Oh wait I switch to show you
You can probably guess what they do
Left one opens this
Middle one is home button
Right one is back button
who here has their desktop/phone background of art they made?
Uh
I think I have it on one of my screens on my windows installation which I deleted
well, i guess i will prefer the first option, looks like more apple-y
So practically no
it might show... but i kinda like this prompt 🌹
It does but I hate apple
for both my phone, and computer, yes
But the navigation is cool ig
well, im bought into their ecosystem, so yeahh....
Has a bit more features but it's also a lot less convenient to go back (like the back button)
lel
It is
But when you have a phone case
It is more difficult
Since you have to swipe from the literal end of the screen
And also on apps like whatsapp
They have a feature where when you swipe it replies to a message
So you can like imagine that
imma just run this partially randomized prompt until something really cool pops up, and then do 8 of those 😄 restart sampler at 10 steps is amazing
i don't think it will really affects that, since you usually swipe the message from right to left if you are replying, right?
The apple ecosystem is trash
Like fr
Only the hardware is like fine ig
But software
Brug
well, if you are bought into it, then it's like you are bound into it, can't get out
Like they do like fine ux aswell

No
oh right
it's left to right
well, it works well since you don't really swipe from edge of screen for message replying
but maybe it's just cos im used to it
Does anyone else here do abstract art sometimes?
these came out nice ❤️
Yeah kinda
You know the song polyriddim?
You should listen to it
Cool song, I presume that's riddim? I'm more of a fan of melodic dubstep and early 2010s style.
🧟
I guessed so from the other genres you like
How is it going crazy kazoo
Very well just chilling in japan
Thats where I want to be chilling
Bing chilling
actually that pic just looks very generic tbh
way to say my tastes are generic
They really are tbh
lmao, well you like what you like, not judging 
I was in Akihabara a couple of days ago and there were like just girls standing in the street with like "kawaii" clothes, makeup, and mannerism. They were all I think trying to get me to follow them to a maid cafe, but it was kinda creepy seeing them all almost robotic. Like there was one girl every couple of meters at least in the area near the maid cafe.
I like you

That sounds very weird
Yeah man I think it will be harder for me to enjoy "kawaii" stuff/girls in media now
2d > 3d anywyas
lol jk

It can generate electric guitar consistently though.
LOOK WHAT MY FRIEND MANAGED TO DO IN THE STABLE
lit af
No,
Fabulous, and trustworthy.
Kirby is adorable!!!
me and my stupid brain went for "horror kirby inspired by giger and junji ito". It's about as bad as you expect
halloween tends to get ok text
it's a horror skull with huge anime eyes lol
What are you talking about?
This image is awesome.
whatever came out of mine, isn't 😄
that kirby is awesome tho
dropped mine in #1072015504870494359
oooh... i even made it better 😮
"did you never hear we elves have just 4 fingers?"
Well it makes sense now that I think about it
fembot reporting!
Does anyone know what keywords would describe this art style?
somehow, i feel :smuggered: is a thing with this image
anime title screen?
Poster
bold manga colorful art poster, for fantasy village mining/farming rpg title, aesthetic, watercolor techniques, playful whimsical characters, people and animals
in the style of cover art title screen, anime aesthetic style of rune factory, oshare kei, magical fantasy rpg cover artwork
Thank you very much!
NP! getting pretty good at interpreting what eye sea?!
How would you describe the character’s art style? Like, would it be a sketch with thick outlines?
MJ can describe images you upload with /describe and SD has tools like "clip interrogator" to help define any image or style you some across
I'll give those two a shot.
i would say bold outline because mentioning line art or sketch may turn your output into black and white
Thanks, this is amazing.

Pretty proud of this generation.
now to put it thru my own original style 😄
hmm ok lets try a anime XL model
pretty

Vroom
Just bought a steamdeck
guys, how do you add text labels on generations? through a prompt, or is it LoRa or some other method?
text and other prints on clothes.... how?
the farm saga thing?
yeah, luck and sdxl model
sdxl occasionally gets words right if you prompt for them
yes, for example
and if not SDXL?
no method or way?
a girl with a text balloon saying "text here", anime
so it's just that -> luck
sdxl models are able to get it right some times
understand...
usually more than 8 letters just fails, unless it's a very common text
Vinland saga?
that too, but i made image earlier spelling farm saga
Farm arc best
dunno how much text you wanted, but getting even a single word right on sdxl is hard 😛
this is your example prompt.
a girl with a text balloon saying "text here", anime
true, but didn’t understand what to write in it
Got a feeling if he didn't have that gun. he would have it stuffed somewhere.. unpleasant
world ending wind chime
almost
Text is notoriously hard for Stable diffusion to output, so in this guide we will look at how we can achieve this.
Step 1: Create Your Plain Text Template
- Create an image with black text on a white background. I have used paint.net to add the text SAND in
controlNet is Power tool!
of course, controlnet is very good at making sure it comes out right
is that my chime prompt?
sorta I just used World ending windchime with of some of my saved styles and sdxl anime model
#3 boat tho
Now I can tell people "I use arch, btw"
like the models for the prompt generator my cmd prompt keeps crashing while trying to download. have u used the extneion yet tough with any cool results?
Ahhh, I’ve never used that extension before, so I can’t help much
i havent read this but i want to try something similar heres my idea
create 3d text or whatever in blender or ms paint
then run it through img2img to try to apply a style or something
i feel like controlnet using a depthmap or normal map or outline or something would be good for this
I think it’s not difficult to check each of the methods. depth or normal
the part i dont know how to do it to take the image from paint/blender and get controlnet usable stuff from
I remember that Blender has the ability to render a certain pass. but you can also use separate software that will create a depth map and a normal map
for example
Online NormalMap Generator FREE! Create a Normalmap directly inside your browser! No Uploads required, completely client-based
suffered a fatal accident
Controlnet has preprocessors for all of the models. You shouldn't need to do anything particularly special ahead of time. Basically just create a png or jpg containing the text in the correct size and location and provide that image to controlnet, selecting the appropriate preprocessor for the controlnet model you're using.
What
so sad 😔
Upscalers aren't on Civitai, you'll find fatal anime here https://openmodeldb.info/
👍
oh... the was talking about this? I didn't understand.
i know a bit about render passes which is what originally gave me the idea to try to use controlnet but also then i am not sure how to give controlnet a normal and use it exactly
so like if u rendered a depth map or normal render pass im not sure how to actually use that in control net im sure theres probably a tutorial though if not i should figure it out and make one
Give this one a try too, I've been using it lately Fatal darkens images a bit and can add some noise in areas that are already dark. https://openmodeldb.info/models/4x-NMKD-YandereNeo-XL
https://stable-diffusion-art.com/controlnet/#Using_ControlNet_8211_a_simple_example probably contrins it i havent finished reading yet
for example
pic by @supple raptor
lovely forest ❤️
In that case you would select the normal map or depth model and set preprocessor to "None" since the image is already processed.
but you can also make an image of the depth map during generation.
Yep
rezult.
with my ControlNet settings before..
very very simple prompt red forest,red flovers
Sorry I don't use SDXL
It took 300+ seconds to generate a single image for me
I don't either. That upscaler doesn't need SDXL. It's just a coincidence that it also has XL in the name.
It's right here #🍥|anime message


Yeap!!! I did it myself!
btw. now i use for text on-line generator
Of course my goal was to add text to the character. prints on clothes or so on.. but this experience is also very good.
did u do that with the depthmap trick thingy
for this I used the following settings
Why?!
Giga finger
with @supple raptor model KakigoriV2 kawaii!

This best frfr
going with the times!
Yes but sadly
True and real
my kind of art
I also use both
is not Anime
NMKD is best for general use imo
Fatal I recommend only for anime
But use another NMKD
Not the yandere one
Like it's not bad for general use
But yeha
Yes
My favorite
For non anime
For ultra realism though I don't know
Oh anime sharp is classic
Oh man
You making me remember
I just realize mid journey can't do any free trials no more then i realize SD has the same thing.
sd doesnt need free trial , just install webui and generate unlimited ammount of pics for free
Ik that yet I can't have my PC with me all the times right?
That's why i use mid journey sometimes and i forgot SD has channels for ai generation too
Soo ima leave mid journey now
I extremely love SD
you can set up auto1111 to work remotely and access it tru your phone or tablet
When your phone does better then your pc:
guys tell me there is no chance to use SDXL models in Auto1111 with 6Gb Vram?
with auto1111, probably impossible
with comfyui, it might be possible tho
what about ComfyUI?
yup, that's what i said, it might be possible with comfyui
how?
literally just using comfyui with sdxl model
maybe without refiner since that takes a lot more vram
nope.. all my attempts to load the model in ComfyUI ended with a message about out of memory
and it withour Refiner
then you probably can't support it
another way is to close all other software that might take up vram
When trying to load the model, an error occurred in the loading node.
So very expressive.
MOUE
Is that a Bunny hood she's wearing?
i think it's an extension of the beanie 
@lyric zealot How is she so adorable?
i agree 
once u made tthe canny black and white outline thingy how did u make the final result
i guess thats the part im not getting
Should be possible on Comfy from what I've heard, but not worth it at all. It will be very slow. As it will need to swap parts of the model in and out. The model will need to be fp16, and you'll almost certainly need xformers. Also, you'll get terrible images at 512x512. With XL you have to stay at the trained resolutions.
- choose an image that will be the base
- By specifying the desired preprocessor you get an image outline
3 you write the prompt that you need.
4 you generate an image for which the base will be just the image outline
This pic is still amazing with those expressions of this fabulous lady. 😮
Thank you. Then it’s not worth trying and wasting time.
So very fabulous, how is she doing this and making it look so easy?
she used to work part time at night club as a dj 
@lyric zealot She's a woman of so many Talents.
How is this canon she'd do this, too? 😮
all switched to comfyui or something other as auto1111?
the quality really improves nearly every week
how did you manage to switch the perspective of the character?
that's before reincarnation 
what do you mean by that? 
i usually just stick to auto1111 because im lazy
If her hair is Magically Beautiful, what sorts of Magic does she have?
ig that character is a lora?
she might be able to hop between timeline
here with metadata
Huuummmmm
I'll check it later when I get on my PC
My anime gens never look
This crisp
Like
The colors I see on yours are bright and well defined 
And when I use a model that doesn't have a vae included it just looks bad
I tried a few vaes with them but images just look bad
So she has Spatial powers of the Dimensional persuasions?
1girl, black hair, forehead, high ponytail, red eyes, oni horns, school uniform, sharp teeth, smile,
<lora:aoi_12:0.3> <lora:TAF2022:0.3> <lora:mjty:0.4> mjtyhz
Negative prompt: verybadimagenegative_v1.3, (low quality, worst quality, bad anatomy, bad hands, normal quality:1.5), artist name,
(claws, holding, holding bag, black hair, animal ears, ahoge, twin tails:1.3), (mini skirt:1.2), (leotard:1.3), holding skirt, lifting skirt, clothing pull, animal ears, depth of field
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 22, Seed: 1035270889, Size: 512x768, Model hash: ec37a01a01, Model: anmitsumimimi_v10, Denoising strength: 0.45, Clip skip: 2, ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate/erode: 32, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 23.9.2, Hires upscale: 2, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, Lora hashes: "aoi_12: 1aae14b8ed57, TAF2022: faa401305beb, mjty: 14cf725faac8", Dynamic thresholding enabled: True, Mimic scale: 7, Separate Feature Channels: True, Scaling Startpoint: MEAN, Variability Measure: AD, Interpolate Phi: 1, Threshold percentile: 96, Mimic mode: Half Cosine Up, Mimic scale minimum: 4, CFG mode: Half Cosine Up, CFG scale minimum: 4, TI hashes: "verybadimagenegative_v1.3: d70463f87042", Version: v1.5.1
Template: (masterpiece, best quality:1.2), detailed background, anime, hyper-realistic, highly detailed, mature female, extremely clear, official wallpaper,
1girl, black hair, forehead, high ponytail, red eyes, oni horns, school uniform, sharp teeth, smile,
<lora:aoi_12:0.3> <lora:TAF2022:0.3> <lora:mjty:0.4> mjtyhz
Negative Template: verybadimagenegative_v1.3, (low quality, worst quality, bad anatomy, bad hands, normal quality:1.5), artist name,
(claws, holding, holding bag, black hair, animal ears, ahoge, twin tails:1.3), (mini skirt:1.2), (leotard:1.3), holding skirt, lifting skirt, clothing pull, animal ears, depth of field``` there you go
possibly 
CFG 22

Like what is that accomplishing for you
I never know when I should change it
cfg 22 is because of a very specific extension
Basically this partDynamic thresholding enabled: True, Mimic scale: 7, Separate Feature Channels: True, Scaling Startpoint: MEAN, Variability Measure: AD, Interpolate Phi: 1, Threshold percentile: 96, Mimic mode: Half Cosine Up, Mimic scale minimum: 4, CFG mode: Half Cosine Up, CFG scale minimum: 4
I haven't used dynamic thresholding in so long
dam, that's a long one
but interesting that the character is always so accurate
Well, that might be due to the dynamic thresholding. Cfg set to 22 so that it will follow the prompt more, but mimic the style of cfg 7
So she is blessed.
Hey, I just started on this ai thing and I wanted a tip or two for something I had in mind so if you have the time, could you give me a hand
Anybody's help would be much appreciated
thats beautiful (:
I do be gaming
....ok
sanji
one piece is mid
hi all
Dynamic thresholding enabled: True, Mimic scale: 7, Separate Feature Channels: True, Scaling Startpoint: MEAN, Variability Measure: AD, Interpolate Phi: 1, Threshold percentile: 96, Mimic mode: Half Cosine Up, Mimic scale minimum: 4, CFG mode: Half Cosine Up, CFG scale minimum: 4, TI hashes: "verybadimagenegative_v1.3: d70463f87042", Version: v1.5.1
Template: (masterpiece, best quality:1.2), detailed background, anime, hyper-realistic, highly detailed, mature female, extremely clear, official wallpaper,
What's this plugin
for sure for sure, sorry for the late reply tho
if im not wrong, it's called dynamic thresholding
uhh oh
uhh oh?
Human Dalek hybrid 😄
here, have my latest generation
I mean they look like tanked up waste cans with plungers for weapons
it's the joke bosses that usually are the most dangerous
does she have 4 mecha-boobs? 😮
kinda...? shrugs
this dj XD
last one 😄
11 seconds until my last batch 😮
i found where i switched to the proper format for my lyrics prompts
thuds
it's just... uhh... kinda a lot i generated. i want that all in a single prompt 😮
"your system on" damn, this demented bunny scares me
level boss
tried it on the hf, not sure yet
Can you use this locally?
Whassat?
yet another model type
Oh so Würstchen is a new type of model? From what I can tell is it like uhhh, speedy XL?

that's what i gather from it
Maybe I should try it out, less VRAM? 
so, @nova remnant is right: can we run and train it locally? if not, it's dead as far as i'm concerned
ooh, gotta highres these 😮
O wait, you can't get it locally? 
O, fair 
XD
She looks the same
I've never seen this before like this week
I was watching Omegle videos of people just don't stupid stuff as usual but on Omegle lmao
People like that or something ? I've seen people do that a bunch of times
Its just a funny thing xD
thats awesome (now make it 2D.....)
Untrue and unreal
Ewwww
Why is the second girl using apple products
Disgusting
Like that keyboard and mouse are so bad
It is 2D 
every image on pc is 2d anyway
Technically people can only see in 2d
we live in an anime
Is that a LoRA? :o
Why were they punished?
just art style lora for the sketch look
too much cute is illegal
I need that LoRA then, and to know who made it 
I use it at about 0.7
very good LoRa agree!
I downloaded it on a whim, I didn't expect it to work well on anime models, but was pleasantly surprised.
405 mb kinda makes sense 
Fair :O
I wonder if they did anything fancy to train it 
Ao in the Touhou server resized it and the smaller version sometimes works even better. It's usually at least as good as the original.
Ohhh fair then :O
with/without lora
left 😂
Very black and white
think then top
The one with flat color and more 2d style is with lora.
I'm stealing that more 3d looks cool
Oh, that model is breakdomainrealistic
Could I please get the model name?
Can you send me where you downloaded it from?
Couldn't find it on civitai
I remember the author took it down

You can try to find it on hugging face
well that came out odd
I started using ddim
Works pretty well for me
And it's the one the model uploader recommended! :D
In my PC everything is slow ddim is one of the fastest
Ble
hmm yeah works good
try restart on 10, it's pretty damn good too
changed my saved style I use too
I call it soft anime: fine art, soft colors, black outline style, elegant, very detailed Cel-Shaded With shaded detail's, soft shadows
yoinked, lets try it 😄
go for it
why not 😦
I'll update it on another driver to see if it works
1.5.2 broke for me
1.5.1 works
Just lazyy
i think i reinstalled venv when i switched to 1.6
You can find a gdrive link for anime and realistic on stablediffusion reddit
looks a bit cooked
shrugs
high cfg?
5.5
interesting
resists urge to apply for VGH (voluntary go home) and just make art
<---works from home

Which model is this
More examples of this lora
I want this lora to create strong color contrast in 2d style because recent models are becoming more and more 3d
what lora and/or checkpoint are you using on civitai?
This Lora is trained by me, I didn’t upload it because it is still not good enough.
The model is breakdomainrealistic, you can find it on hugging face, the author took it down from civitai
oh ok thank you,
it looks great btw
Variations using PPF noise by simply doing PPF seed changes:
lies. you just wanna check the booty
miku boba 😄
What's PPF noise?
I definitely want to check the booty.
Is it some how different from seed variation?
checking the booty be like
PPF Noise = Perlin Power Fractal Noise
aight, this prompt goes into the collection too ❤️
currently at 331 lines in my master prompt, and 1k images generated today XD
331 lines in prompt? how many tokens is that?
probably around 11-12k tokens now

but. only 2 or 3 lines get selected
i'm a madman for sure, but still an sensible madman
when you mix it up, and get stuff like this ❤️
I've been thinking about doing some sort of dictionary build where I put together a massive dictionary of proper nowns, nowns, verbs, ajectives, etc., and then just have my randomizer build out of that.
he be smelling ketchup
guess what next thing i'm building
i already got nouns
got movements
got mediums
Nice
our 2 protagonists and the foe they will encounter
working from back to front... i'm getting there
I need mora characters







