#🍥|anime
1 messages · Page 130 of 1


any model
Is that like 10 minutes for you
metadata
i think it's really just prompting

but the aoi lora is doing god's work as well
Oki
Hey everyone, sorry to be the one that has to do this but we're getting lots of reports about nsfw content on this channel. Please keep it sfw and clean, if in doubt please remember the rules of the channel here> #🍥|anime message
noted, will
them if i spot any, me included
if possible, maybe pin this message as well 
I'll ping @elfin geode or @finite herald can we please ping this on top? > #🍥|anime message

@lyric zealot wasn't us, right ? 
I'm innocent
there were some cool images here and there but Discord is getting tougher on nsfw content, so please keep it clean. Don't want to cut the creativity wings out of anybody but there's stuff right on the edge
Very interesting pumpkin soup she got
the string got inside 
horror


(I'm the stupid one)
The Prompt: AND_PERP
electrical pole voyage
AND_SALT small nocturne companion
It then processes the AND_SALT prompt small nocturne companion in the context of electrical pole voyage
about this
i literally have no idea what you said
but cool result 
i have no idea what im looking at 

Unique images there.
This one walks a rather dark path, whether for good, neutral or evil remains to be seen, but it will be forecasted with a sense of violence.
ai is into catgirls real?!?!?!?!?!?
i think she likes subaru
double doors
trolleado
yes
is bing image creator sure your boosts get refilled weekly
because im back at 100
i need what bing is smoking fr
Who is this person?
Microsoft is a US company, in the US Sunday is generally considered the start of the week. It is currently Sunday in the US.
it reset yesterday as well though
Mine reset approximately 17.5 hours ago, which was midnight PST on Sunday morning. That's the only time that it's reset for me.
Careful, don’t go too far 
This is beautiful.
me offering @lyric zealot to the mods so I don't get 

what
I found a way easier to achieve it
use clone stamp tool to add things to make it able to diffuse, then use inpainting model to inpaint(img2img)
doing great, and you? 
im okiee
paint my beloved 
slurrrp
has anyone used the infinite zoom extension
I haven't touched animation yet
i wanna try it but it needs a inpaint model and idk if the inpaint model needs to like be compatible with the original generated image model or what
it should be of a similar style
ok
ill try to find like an anime in paint model or something . i only know of rev-animated inpaint which is a good base model but ive never used an inpaint model
saw the guide somewhere
o interesting
:o
https://www.reddit.com/r/StableDiffusion/comments/zyi24j/how_to_turn_any_model_into_an_inpainting_model/ might be similar to how we did the dark model
i forgot the name of it
offset noise
anything v5 looks like it might work
how come sd2.1 didnt take off
@lyric zealot these are just so random
I don't even know whats controlling these prompts lmao
are you still trying to make it work?
I'm married to failure
Whose pov is that
same prompt
search result for inpaint on civit only return about ten results
make your own using this?
basically the same concept of how we used to make offset noise model
remember those model where we can generate stuff that's super dark?
I'm still sad that our miku-mash got taken away
ok i will check into that
i havent seen that noise model thing
is that just the ddpm sampler?
basically most 1.5 model, that's not merged with offset model, is supposed to generate each picture with the overall brightness above a certain level. As such, you can't generate super dark pic with those model. To do so, you can merge it with offset model like explain above


i didn't read that much tbh
i just understand how it works
and how to merge that's it
it sounds like i might just be confusing terms cuz ddpm is a sampler not a model
i guess so 
i like this idea though i wonder if thats incorporated into newer models
i like high contrast type images
not really
and low light stuff
or at least i don't think so
since it will definitely affects the style a little
but you can do it yourself
since it's really easy
i still like stuff like that too
Miniature Miku Hatsune. 😮
I've really been thinking of using the AI Art for learning and practicing my own art
this is the example used on that page
right uses the offset model
although they both look good lol
bascaily sounds like sd by default generates images with an average brightness of .5 ,in a range of 0 - 1
yup
that's exactly what it is
i don't know what brightness it is tho
but that's the concept of this
go for it

I'm so happy my hands look normal!!!
smoking bad 
There's something wrong with my cats
a very human thing to say
Indeed
Not the worse, half decent at least
nope 
thank you RNG god 
very cool
i forgot how much i loved the hades embedding 
forgive the fingers
@lyric zealot same prompt and seed but neutral cfg scale at maximum

most normal ai finger
are you playing with cfg extension or the neutral extension?
seems most of my favorite models do have an inpaint version o just had to go look at the page
wat extension r u talkm about
under the normal cfg, open the extension options and theres stuff

@gusty palm https://github.com/ljleb/sd-webui-neutral-prompt good luck if you try it
dynamic thresholding?
ooh i was lookin at this the other day
very confusing
ill incorporate it into my slot machine randomness
I hate LoRAs that generates the character on the same pose 99% of the time 
i like thos big horns

Oni Horns are always a good choice 👌
I am anger
I have a strong feeling that Stable Diffusion doesn't like cat ears from behind
Lmao
From where AI fails, Human Creativity takes it's place
Or is it the other way around lol
Lmao and awwww
If only eyes and the head can look in the right direction lol


(no horny please:999)
kinda scary

(knowledge of how this extension works please:9999)
♾️
@lyric zealot I'm gonna use these distorted fingers to punch you
how you can get money from anime + AI ?
with godly luck
horror
hi, there's no bot in this channel, you can go over to #1100170312106127410 for SDXL
I'll prompt it out of curiosity
it's gone 
Create an image representing 'One' as a number with personality, eyes, and a mouth, looking curiously at another number that came before him. Make it TikTok-sized with a playful and colorful style.

(this worked, I might be on the right path)
@lyric zealot I was reading some extra explanations on the discussion in github, some more examples showed that the prompt kinda of flows back to itself after the first, like it adds to it
and at the end makes one thing that tries to do it right
The prompt for that one, very simple
a girl with fox ears
AND_SALT [ golden, metallic ]
AND_PERP [ wings on her back ]
now I get ass on my face 
this model
please eli5 if you understand the extension later

~~why do i not get that? ~~
okay, time for lunch

hope you have a great day at work 
Thank you!
What is that show?
The student-filled Academy City is at the forefront of scientific advancement and home to the esper development program. The seven "Level 5" espers are the most powerful in Academy City, and ranked third among them is middle schooler Mikoto Misaka, an electricity manipulator known as "The Railgun." When strange incidents begin occurring througho...
better than the main story imo
it's the real luna 
Ooo, das really cool :O
The infamous bag
That looks suspiciously like that Pulp Fiction meme
it is ^^ controlnetted it 🙂
thx 🙂 loved making that effect
Just like how I loved making things in my current style
Last post is here, got some more above
I can't really post anything right now, at work
i'm doing sysadmin stuff tho
damn, this one came out great
kinda digging this artstyle tho
really digging the roughness of these images, gives them charm, imho
good morning
man when I was in japan I swear every single student had that exact bag
This ai?
probably dall-e 3
what's the prompt? face from hell?
@wispy canopy

do you have any idea how to use neutral prompt extension
My brain is mush from trying
yes, for a bit
@nova remnant it's about replacing the "AND" with the other AND types neutral prrompt gives
with each of AND_SALT AND_PERP AND_TOPK having different effects
AND_PERP is an activivation, that will give back regional control to different parts of the prompt the moment an area comes into contest
for eample, you have girrl + cat, the one that's behind AND_SALT will be given way less priority, it will have to cede it's pixels to whatever else would try to go there
AND_SALT is the opposite -> it tries to make sure whatever is prompted in the AND_SALT will get a region
AND_TOPK tries to take the top 5% most likely output of the all the previous things before
so topk would go well with style or details, as per the documentation
Have you ever watched or read Letter Bee?
nope
It has a similar aesthetic
I highly recommend
I'm kind of done with work I just need to head back to the factory to sign off for the night
Tegami Bachi is great man, sadly I never did get to finish reading it
@wispy canopy ill DM you
On top of this, you can always use X/Y to see the actual effects each one does
Experimentation is key to success, especially in this field
I need merch with catppuccin color scheme
asking for opinions... which one of the 3 looks better?
no underglow hair
i hate it.
hmm, this is pretty cool, actually 😮
im kinda leaning towards bit more hyper realism
ello 
idk if i'm in the goond channel (i hope if not i will delete and repost in the good one!)
But i'm starting using Stable Diffusion for the past days (3 days i think), and yes, i have some question 
so : How to avoid this "double scene" if i can say ? (2 camera angles / etc ?) (in "gif" format?)
And. There's is a "good website" where "a lot of reference" about Stable Diffusion (extension/prompt/settings) exist ? (i'm looking a lot of video on YT/ checking on Reddit/google but yah, can't find something really good for now sooo, i ask!)
aaand : do the negative prompt "are" really "usefull" ?
Or do we need to be really like "negative/opposite" of what we type in the "normal prompt" ?
- size only : 512x512?
But Stable Diffusion is really nice to use, now i need to learn more now "how" it's work 
Thank you if u can answer! :v (and again if im in the wrong channel tell me i will switch
)
(sorry my bad english!)
She is very beautiful.
#📝|prompting-help would be better
Idk how the gif stuff works 
alright thank you
gonna check that 
no prob for the gif!
@earnest juniper https://stable-diffusion-art.com/tutorials/
bunch of stuff there
It's like sitting on air
reminds me of when I gen'd a girl sitting on the sun
she was supposed to be on a bike looking at the sunset
Lmao
Right now I'm trying to think of a prompt that uses Regional Prompting and Neutral Prompting
ok
I care about how people are feeling?
my cookie
@wispy canopy playing around with and_salt monochrome stuff its pretty cool >.>
👍
also someone was very angry at that watermelon in the back
I will give it a pointy/10

she doesn't know magic well yet
@gloomy scroll lmao no booli
I can't get this extension to work well with negatives for some reason ree
its 50% chance booba 
What extension?
neutral prompt
it works differently than regional
but they can be used together
its super interesting

I want apple pie
you have to ask her to do that, she's got all the apples
regional prompt + neutral prompt
That looks really nice
neutral prompt to get a specific style, regional proompt to make the dragon
how did you place the dragon in her cradle without masking option?
(Don't use eletric magic near water)
@nova remnant
using the prompt-ex method
basically, using a keyword, you can have it create it's own mask
hmm, im yet to try out the promp ex
@wispy canopy Do you know if or how I can use quality prompts with neutral prompt ?
AND_PERP
black and white lineart manga, drawn by (__manga_artists__:1.4) , hyper detailed manga, retro 80's manga , simple crude style, professional drawing,:1.5
BREAK dragon, small pet
-> small pet becomes the mask, which is then prompted with "dragon"
this way dragon only applies to where "small pet" should be
takes some trial and error to get the treshhold right -> too little and and the mask consumes the whole image, too much and the mask doesn't apply at all
i'd suggest against quality prompts in neutral
interesting, i wonder if regional prompter will implement drawable masking
it does
you can see it happening in the preview actually
The Mask tab?
it does like X amount of steps, and the determines where "small pet" is active, creates the mask for it, and then applies that prompt
no, the prompt that is more akin to "dynamic mask"
so it tries to determine where the subject is you specified
lemme see if i can do a generation which explains it
it hath been a while since i posted here lol.
i was trying out a different extension for latent couple and that allows drawing mask which creates a prompt box .. so you can just draw then enter prompt in its that specific box for latent couple effects
These are amazing
yea; i've been impressed with all the new tech coming out
these are from dalle ?
yup
i hope local stuff can surpass that at some point
@nova remnant this is the same see, with and without the dragon prompt
so basically what happens, a couple of steps inside the generation, it determines where "small pet" it, and then simply makes a mask for that area, and then keeps applying it. so i can now replace dragon with any other creature, and it'll just replace wwhatever creature is in her hands
ahh ok
i have to experiment with it to get a practical understanding
btw is it possible to draw multiple masks and use prompts specific to each of those massk in regional prompter?
replacing dragon with "chibi dragon"
yes it is 🙂
really? hmm .. can i have an example if you dont mind?
like a screenshot?
anime ADDCOMM girl, magic circle behind her, standing on beach, BREAK warm, bright day, sky, clean sky, outdoors, scenery, beach BREAK magic spell, runes, transparent, , circle BREAK (shorts, short white hoodie, sneakers, modern clothes, small breasts, open hair, long hair, straight hair, white hair, hime cut, blunt bangs, azure eyes, pointy ears, dark, choker, fangs, blush,(full body), standing, facing viewer, happy, :D:1.3), girl
here's one i tried for vil a few days back
but the gui?
you mean this part?
eface.. look at this https://github.com/omerbt/MultiDiffusion
think its called multi diffusion regional prompter
its like drawing on the canvas which then creates a prompt box for that mask and you enter prompts in that box to target that mask location
i think i've tried it's predecessor, and i didn't like it
it generates multiple images and then stitches them together
how come?
and then re-diffuses it once more
but global lighting doesn't mmatch ever
this finally ported this to SD?! gotta test right now!
good luck katiekat, hope you don't pull out your hair trying to understasnd
its too good
it's basically a cheat at this point how good it is
im guessing as usual aha
like you cant do that in sd without alot of inpainting
who... who would do that
can't get it explained -> having a huge llm network for contextual understanding and having that translate stuff into tokens is a godsend
not you... right?
oh damn that looks so interesting
isnt that wat magic prompt supposed to do
not sus at all
i forgot i have a llm nowi can make it make prompts for me
@supple raptor dalle3?
yeah
the drawback is that it is soon will not be free (i cant imagine them giving cutting edge art generation for free when gpu aint growing from a tree)
so better abuse it while it last
it won't be, i'm pretty sure there's a llm acting as a tokenizer in front of the image generation, that's 2 very heavy ai models
@elfin bolt can i ask you to dump in a prompt of mine, (one of my song lyrics) and see what it comes up with?
sure
Beautiful girl with guns.
no imma go run them!
indeed
Microsoft put 10billion in OpenAI for Bing/Windows/Office, they have the datacenters to run it and they are collecting data from you so it's not technically "free" 😜 OpenAI's version will be kept paid by subscriptions
@elfin bolt ```Light the fuse; see the sign Unbalanced heart; I'm in my prime Get out of here while there is still time I'm reaching overload; I'm going right on top While you abide my dynamite BREAK But I see you, all of you, clearer than I ever have Stone cold, with no soul, loneliness your only friend Get it in your head, you're never gonna change me (No, I only wanna set you free) BREAK Can I take another step? I've done everything I can. All the people that I see, they will never understand If I find a way to change, if I step into the light Then I'll never be the same and it all will fade to white
BREAK official art, promotional art, surrealism, inspired by James Humbert Craig, oil painting landscape concept ```
at this point i'm confident that big tech know more about me than myself so i dont give af about my image fetish data
lmao
they took your lyric and make a lyric
but its too much token i guess
Shes hot
2 last bottom is only taking consideration of the first two words
nah, it's taking the prompt pretty well, actually 😮
im dead
i have to say, i'm EXTREMELY inpressed
remove from the last break : official art, promotional art
i think that should improve it
if only someone in openai would leak it
and get another 'redacted' situation on our hands
not like anyone here can run it in their home
you got a few dozen h100's lying around to run it XD
i know some sysadmin who has access to 8x a100 xD
i'll rent it runpod just to make 1 image for 10 bucks
im still curious about its NSFW potential
okay I'm gonna become a google engineer
and steal the AI for us
see you guys in 10 years
in 10 years google wont have human anymore at that point
i'd like to see humanity still alive in 10 years
i got this feeling we heading to a clusterfck
@eface here is a GUI for regional prompting... https://youtu.be/sEjMPmCiWJ8
An interesting feature of Multidiffusion / Tiled Diffusion extension is the ability to control specific regions of your image at the same time. This provides another way to get greater control over image creation. In this video I'll cover the basic ways to use this so you can start experimenting.
Link to Github Multidiffusion extension for ...
i downloaded a inpaint model.
and also neural prompt seems to slow generation down a bit
yeah, that's the one i dumped
oh boy... why did you dump it lol
hi
because the regions have no correlation with each other
I sent them to a friend like a month ago or something 
I'm your fan
(exaggerating)
glad i have an actual fan (exaggerating)
that odd.. im gonna give it a test run.. already have it installed and lets see what i get :p
coulnd't get it to wwork, regional prompterr maskmy new friend 😄

I just figured out how to tell a manga story in Stable Diffusion
"Cat guy sits on a weird floating piece of rock and thinks facing a wall about life"
Well, there can be many interpretations. My take on it is that he was the one who did the graffiti
does neural promps work for negative promps
tactical bitch 😍
i got the sky and field to work out right but something off with other elements in the image
Cancelled this because of bad legs but
Her guns are loaded.
Bing keeps their versions free (gpt-4, dall-e, etc) where you pay on OpenAI. They just turn boosts from daily to weekly and you won't get pictures as fast if you don't redeem rewards for boosts.
they even installed new GPUs yesterday during the high demand
seems like they need more tho
it's all part of Windows Copilot which they're pushing
we never know what the big tech stakeholder decision later down the line, it can be free for now but the money burning is still going on
eventually they have to make a profit
enshittification incoming
they make a profit by the ads you see in copilot & the additional search stake they get
Do they really spend a lot of money for having the service up
yea, but you get served an ad by like every copilot request nowadays
so i'm sure they make it back

is openai ok with that lmao
@wispy canopy this worked out exactly how i prompted by drawing masks on a canvas, 3 columns for 3 different girls in specific outfits and hair colors
electricity alone to run something like dall-e is very expensive, not to mention the opportunity cost of not being able to use that hardware for something else that could be making money.
^^
Microsoft owns OpenAI lol
very nice 😄
chatgpt is rumored to cost 700k a day
meanwhile a user subs for 20 buck a month
with opportunity cost factored in, probably.
3 pair of thighs
Good prompt 
closeai
they are literally bleeding microsoft cash to ensnare as many users, get their data, and who knows what would happen
And then dalle4
yeah :p.. im happy to see how the prompt worked exactly how i wanted them to by drawing masks

only 49% of it. though they get get 75% of openai's profits until their initial investment is recouped.
more than that and my GPU dies with regional prompting
initial investment of 10 bil... ahaha.
API access to the companies is probably the most probable way no?
@nova remnant write me a prompt with a background specification and 3 different people on it?
like users subscription can only do so much
had no idea Reid Hoffman, Reid Hoffman owned part of openai
why me
This is cool
the more corporations control the flow of information the better *
just gonna try one with you
unless you are tied up doing something else
Yeah, I'm sure commercial API access is much bigger than private subs for that type of service. Most normies don't have any reason to use something better than the free Chat-GPT to occasionally ask questions.
offering gpt 4 model that can be finetuned with api access for specific company purposes could be big monopoly in closed AI model
everyone will have to follow it
lest they get left behind with inefficient process
for the neural prompt , are u only supposed to use one of the AND things at a time
or can i just like put each part of my prompt into a ddifferent thing
I just don't have much ideas for regional prompting o:
I like how the neutral one works
so far it seems to be not working how i expect but idk
does and_salt add salt to the image
it shouldn't, those are not sent to tokenization.
that's assuming you have the extension and type it in caps
i may try using the prompt formatter thing the extension gives instead of trying to hand trpe it
mouth with an open mouth with huge fangs with open mouths

Maybe new OC
@lyric zealot 
||step on me|| nice design 


Yes
why is pikachu sad

preying mantis type body layout
They did not win the Pokemon competition
What is Rin Tohsaka doing there?
i got my form from social security i have to fill out and mail in back in like 2 days cuz i forgot to check my mail for a few days ><
Thanke 

Ayyyy 
Looks like boichi
and here dall e said they didn't train on artist (they lied)
do you mean if hiresfix have different prompt than normal generation?
😳
ya like if u use wildcard in the original prompt, does the hirez fix resolve the wildcard again
cuz if u send a t2i image to i2i with a wildcard in the prompt, it sents the unresolved wildcard
its confusing to explain
i don't think so
i think i get what you meant
she kick the wale :<
Can you do this but with a Pokemon instead lmao
what pokemon
make raptorjesus fighting spaghetti monster

Lmao
why is dall e always at capacity
told you so
Now these are nuts.
Where did she get that hammer?
It was inside her dimensional pocket sleeves
This is a Rinner. 😮
i hate this bitch ass dog
are you using boost, and how long does it takes for each generation?
mine is taking forever 
is halloweedn this year
This is a Rinner.
I trained an AI in Trackmania with reinforcement learning, until I couldn't beat it.
I just opened a Patreon page, where you can support this YouTube channel with a donation (link below). Between research, programming and editing, these videos take a long time to produce. Any support will help me to spend more time on that in the future ! This...
monster,ugly,surgery,evisceration,morbid,cut,open,rotten,mutilated,deformed,disfigured,malformed,missing limbs,extra limbs,bloody,slimy,goo,
good set of monster tags
make it urself
also my prompts are in different boxes for each mask i create
@nova remnant they don't show up I think
):
png info doesnt work?
in case the extension is not embedding the entire prompt, i just prompted for 1girl for each of the girls in image, with different hair styles
👍
got that from @wheat falcon
I have tiled diffusion
for the free gpu option 
how is it working on 4gb?
it doesn't
I sometimes can press the free gpu to clean some vram and gen more
instead of having to restart
but the tiled vae and stuff doesn't work for me
uhh.. it wont let you do regional prompting?
i love this tiled diffusion for easy click and apply
exactly how i prompted for this image, their hair, their dresses and colors
I have yet to try the regional from this extension
I thought you were using regional prompter

i was but that felt more textual than this
Bathing in tea?
what is that
one of the extensions for regional prompting
@nova remnant Only one region is working 
now it wants to work 
im using 4 regions for this... 1 the background and three more 1 for each girl
as soon as I say something
@nova remnant what do you do for negatives
do you put on each region or only one in the normal box
negative on top section as usual but for positive i put them in regions
when you want to draw a scene , utilize the background / foreground options too, like background for the image background and foreground for each regions you want to put characters in
sleeves of holding 
I'm back, energy went away for a few minutes
yall so good
io love this style
I've seen that reaaally minimalistic style before
and how to, but didn't get to it

style lora was made by this guy https://civitai.com/models/24833/minimalist-anime-style?modelVersionId=29709
Why do I get battle royale vibes from this
feels like they are about to loot guns and go to war 
i have dat good lora
if this didnt have a floating wing ><
whats the best model for sdxl cuz im fucking tired of the fat ass dog jumpscaring me
did you type anything in the background region
you can either select foreground or background for a region
Mines coming out weird

ok
@nova remnant
can I dm you
deku genderbend? 
Ye, a loose one at least

Just used an image of him in IPAdapter
Some work better than others
refill time
dude ive gotten them 2-3 days in a row
its part of their program, i got it too
couldve sworn they refilled weekly
A Red riding hood. 😮
tahts awesome
im hoping sd will catch up to this level of understanding
what did u type to get that
i wonder what kind of magic they use to make them so good
uh prompt gone in dall e 3 but someting along "digital anime illustration style of scared short white haired girl with white dress running on endless stairs with giant dark monsters made out of jaws, eyes, and giant hand approaching her, dreadful scene, running motion"
capitalism magic
oh wow
we need to steal it
or does that mean their models are just trained on millions of stolen assets lol
we are all the same tho (they are millions times better)
Give prompt
masterpiece, high resolution, highly detailed, best quality, 4k, ((cute vampire girl:1.4)), classical gothic Victorian __devilkkw/attire/attire_uniforms_and_costumes__ outfit, (extended vampire fangs:1.2), open mouth, ((huge fangs:1.5)), short hair, ((diva makeup:1.3)), sleepy, shy, petite, (small breasts:1.2), adult 1girl, (__autowp/fantasy-setting__:1.2), creepy, haunted, midnight,slimy,goo,morbid,rotten,eldritch abomination, saturated colors, high contrast, abstract, foggy,
also a bit of stylepile
but mostly that
sec ill get the actual prompt
👍









