#🍥|anime
1 messages · Page 138 of 1
first of all, make the tin foil a fashion statement! now you're not a whacko anymore
what are you using as background prompt?
lol fancy
A girl with a tinfoil hat and dejected solemn look. Behind her is a board with photographs, documents, and maps, linked together with strings. Illustrated in high quality anime style
is that dall3?
... it's my own custom merge
impressed it got the tinfoil!
but yeah, dall3 is crazy good
i mean, i'm really having to put everything i know for prompts to work
you might be able to use gpt to create regional prompt inputs
I haven't really looked into scripting in any sense for SD
nah, the regional prompt is pretty easy to use once you understand how the masking mode works
I mean to have it compose the mask regions for you
to get the dalle experience of things nicely composed
oh, that's not even important in my prompting world XD
regional prompting better than latent coupling thing was?
was that the one?
been a while since I did regions
haven't tried Tiled Diffusion's version yet
yeah, regional prompting WAY better
as it's using attention instead of latent -> it's not significantly slower
and with prompting mode you can define the masks dynamically
so in my case, i put the attention on the headwrap, and then define for that area that it should be crumpled aluminum foil
it deserves an instant lora
hmm does it not take your main prompt at all?
it does, and it doesn't 🙂
nice
if you want more dalle raw pics I can provide for lora
basically, with the prompting mode, after a feww steps it looks which region was activated the hardest on that specific part
sadness
so your main prompt is about a girl, the region is just tinfoil hat?
manga collage, headwrap, flat black and white, multiple (manga:1.4) panels background, Behind her is a board with photographs, documents, and maps, linked together with strings BREAK (crumpled tin foil, crumpled aluminum foil:1.4), tin foil hat, headwrap
so the headwrap is what calls the attention to that prompt
oh it must be
That's multi diffusion
Also you need a background region
Put it first, then whatever foreground
From what I tried with it, regional prompter is better
regional prompter also allows stuff to bleed outside their regions, if the image requires it
Hello people
hey cat ^^
oh in your extension you don't define any regions @wispy canopy ?
yes, using the prompting mode, you don't define regions
you specifiy a word used in the base prompt, which is then probed for activation regions during generation
the system calculates the area of the word written after the last comma in the prompt following BREAK
whoa
so girl, hair, eyes BREAK white, hair BREAK green, eyes would make the chance for green eyes and white hair pretty much 100% guaranteed
without the bleeding effect
so that's how the aluminum part only goes over her head, and not bleed out to the rest of the image
aight, now you gotta set the treshholds
lemme make a quick tutorial 😄
girl, hair, eyes
BREAK white, hair
BREAK green, eyes```
Try only quality tags in your main prompt, set prompt 1 to background and describe the background there, prompt 2 foreground with ~0.30 blending and describe one girl there, prompt 3 same but for the other girl.
so, i specified two regions, one for the hair, one for the eyes
the additional images you see, are the masking region activations
again, one for the hair, one for the eyes
the hair might be activated too much, so i'd increase the first treshhold to 0.8 and see if it still activates enough
what's the magnitude of the threshold mean?
you need a treshhold for EACH region you specify
it's to define how hard it should dampen grabbing latent pixels in the vicinity
0 -> the mask simply grabs everything
1 -> mask doesn't apply at all
I still can't get the green eyes with exactly those settings
usually i go for 0.5-0.7
strangely it's always blue eyes
can you show the images, masks and prompt you used?
there's no masking images... so it's not triggering
any errors in your console output?
Regional Prompter Active, Pos tokens : [5, 3, 3], Neg tokens : [13]
If you just installed Regional Prompter, try fully restarting SD. There might be some dependencies that Prompt-EX needs to pip install.
ahh
oh very good call nubby
i always restart webui full when i install new extensions
Same. I don't trust it to install shit correctly.
this. the thing is already hanging together by sticks, glue and some overused ducttape
Usually I restart after updating extensions too
hmm
If you aren't on 1.6 then you should. There have been efficiency and speed improvements, not to mention a gradio version bump which could explain some weird extension behaviors.
I'm on 1.6
hmm
well, I guess I can try Tiled Diffusion's
the composition doesn't always feel super cohesive though
i'm feeling bummed it won't work for you 😦
same
Mixture of diffusers does better at that usually.

oh are you using xformers?
it actually activates in Latent mode
works well for getting graphics on tshirts without giving her a skull too
retried an old prompt of yours 😄
exactly, it ISOLATES perfectly
well almost perfect -> seems the activation region got spread out to her hair lol
lol
ikr?!
now that you can isolate it?
I love that shirt
that hair was accidental actually
also region prompting?
nope
@stable crane just occasional bleeding. But otherwise -> that's a lot of struggling less with that specific girl ^^
#2 has such good lighting on that tie
seems like some of the colors on this got a bit wonked, idk if it's regional or bad luck
might want to check if the activation regions are bleeding
it was just the necktie
she looking juicy tho
maybe it was not meant to work with hypernets
i'm abusing the everloving shit out of this regional prompter tho:
__songmaster__ BREAK
girl BREAK
__songmaster__ BREAK
__songmaster__ ```
"here's a bunch of random bullshit, now make me anime waifu"
it works too 🙂
the character of the gen is different though, left is without it on (and removed the BREAK clauses), right is with it on BREAK (flower print:1.5) necktie, necktie
almost feels like a different style entirely
oh maybe because I'm using Latent mode?
perhaps?
if you can, use attention, it's faster
I wondered if xformers is why I can't use Attention mode
i'm using sdp-attention
but i think xformers should work too
perhaps reinstall xformers?
hm without xformers no luck either
does it bug out?
and now it's back to not working 🤔
at all
though I'm digging this flower blast too
how the prompt performs with the extension disabled
flower print patches... suitable for... other servers
Very interesting flower print placement
you just gave up on print on the tie only?
and went straight for the 🥛
lol
still, i love that model ofyours
Joining the flower print
last one for tonight for me
(masterpiece, best quality:1.2), high resolution, detailed background, perfect lighting, extremely detailed, 1girl, solo, pink hair, high ponytail, forehead, yellow eyes, oni horns, upper body, realistic, sharp fangs, disappointed, annoyed, angry, school uniform, dark, blood from mouth, horror,
Negative prompt: (low quality, worst quality:1.4), (bad anatomy), extra digit, fewer digits, (extra arms:1.2), bad hands, by (bad-artist:0.6), bad-image-v2-39000, negative_hand-neg, text, watermark, monochrome,
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7.0, Seed: 3370348692, Size: 512x768, Model hash: 80d5cce368, Model: aurora_v10, VAE hash: df3c506e51, VAE: kl-f8-anime2.ckpt, Denoising strength: 0.45, Clip skip: 2, ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 23.9.3, CFG Rescale phi: 0, Hires upscale: 2.0, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, TI hashes: "bad-artist: 2d356134903e, bad-image-v2-39000: 5b9281d7c6ad, negative_hand-neg: 73b524a2da12", Hires refiner: first pass, Refiner: expmixLine_v3 [931aaff934], Refiner switch at: 0.3, Version: v1.6.0```
Thank you. ||About to steal your style||
I just realised that lora I'm using has some problems so I'm trying out different style at the moment and wanted to see how you got style those images had
reeeeee

real cool
I wonder where I got this one
pretty sure it's in civitai

this is our newest mod-team member, she's holding the electric ban hammer, which also has the timeout prongs, that thing can get you knocked out unconscious up to 24 hours! we are looking for acomplished artists that can represent her better, in any style you choose (safe for work) 😜
this makes me feel good about my new mix attempt (rightside is the new model mix i'm trying)
i kinda get "nostalgia" feeling with that image from when i used to benchmark my own model
Anyone can change face to my face
its yandere time
channel is completely dead💀
It's a timezone thing
Isn't that normal?
yeah it's more active later
3 AM GANG YEAH!
its 6pm here
Who is this punk lady?
guys iam new here, is it possible to run stable diffusion in google colab pro
yes you can run it on collab, locally and web-based
a cutie
all ladies are inherently a cutie
society spun into a wrong mindset when loving more than one girl is considered cheating, which is directly linked to economic model and exploitation of the mind
its a fact there are so many beautiful girls in the world, how can you not love them all?
This woman is lovely.
only way you cant love all the girls at heart is if you are love deficient or a victim of a preying economic conditions
Nice one
Why they have no noses tho
Do you not see the noses? :o
small noses are pretty common in anime styles. big noses can often look weird.
Yeee :o
morning? isn't it almost 15:00 for you?

Hello Mike mike
hello
Jelly thighs 
aight limited time ketchup accident only, leaving this only up for a minute for reasons (image removed, dm if interested)
Looks pretty dope
putting my new merge to the test, and it kinda did all the 'easy' parts
now the hard comes -> it needs to do all this in upscale too
@wispy canopy
how many gigs billions of vram do you need to upscale 1024x1024 by 2 times
On SDXL
takes all 24gb vram
and then you end up short, if not using tiled diffusion -> which i do not
SDXL 
This is original image I just took mouth from other image
Bonus mouthless Yue
Is this image generated using AI?

@rancid yew This image has pnginfo if you want it
is it? ;_;
nope i jus wanted to show @supple raptor his sticker made it as a winner!

i usually do 1080x1080 and 2-3 images at a time, it's pretty optimized, only takes like 16-18gb vram iirc
able to pull 3 off if i have like no other programs running at all incl discord
Get your camera out of the bathroom, Dr. DisRespect.
Yall think a 60k pixel long X/Y plot would even get processed at the end? I made a list of all my LoRAs (or at least succesful ones), but its 52 LoRAs 
Fair 
🤣
Cause it would take like... 52 minutes to generate all, but I don't know if it would even finish or OoM at the end
this is exactly why i didn't want to bother with loras and such when i do gens xD
y'all really use that many in a single prompt?
Its not for a single prompt 
Its to just make a sheet of like what all my LoRAs look like
oh like a lora same seed comparison on 52 images lol
why not just do like idk 5 at a time then splice em together?
Yeee, just swap out the lora on the same seed & prompt
I mean, its kinda the same thing, but also its more work xD
tru
Yeee, though I mean
breaks in between tho since otherwise ur pc would be unusable for an hour
xD
unless u were doing it over night

Still a lotta work tho
I normally use between 2~5
Would it? 
probably, if it's only 1 image high. I made this abomination on my 2070 before we even had xformers. I don't recommend it though. it'll be annoying to reference.
Same
Oh damn 
That is crazy
Then mine would probably be a really big file too
1536x60000 is 92,160,000px
33152x25057 is 830,698,664px
I mean, fair, would be around a 10th then 
Reminds me of this lol
Though, I would look ugly if its a horizontal strip

yep, 11%

there's the problem. and being an image you can't just CTRL+F to find the lora you're looking for. you have to look through all of them to find it. You're better off doing maybe 4-5 per image so that you can fit the lora names in the filename to make them searchable.
Fair fair, that is a good point 
lol that's impressive. I wonder how the filesize is so big. thats less than half the pixels of mine, but it's almost 50,000x the filesize XD
i mean, i tried to launch a game while doing SD early on and blue screened xD it's gotten a lot more optimized now though
I mean fair, that depends on your GPU and stuff tho :o
It's a pretty old meme. I don't even know if it's actually real.

here's what that monstrosity looks like. I can't even open it in image viewers, they just crash. have to use an image editor to get it to load. it was a comparison of a bunch of different artist names on WD1.2. I tried to reference it exactly once before realizing that it just isn't even remotely practical to use. now I just keep it around for the memes of having a 1gb png.

wtf lol
I learned that day that prompt matrix is a dangerous tool when used unwisely 
Tough lesson 
my poor little RTX 2070 put up with a lot of abuse
back in those days I used to gen images by the hundreds and would be happy if 1-2 of them turned out decent.
Upscale it 2x 
Damn, poor 2070 
my default batch count was set to 150 for a long time
times were tough in those days
I normally just do 10 without hires and upscale the ones I like the composition
now I just gen in batch size 2 batch count 2 with 2x hires fix and usually get at least 1 pretty good image
❤️
@gloomy scroll
when SD 1.4-1.5 was still new i'd do like 200 gens to get 1 good image i'd share with someone xD, so glad that nowadays it's like 1/4 for a good one more or less
might be part of why my TITAN RTX died
i already abused it in tons of games xD
my 2070 is still kicking, it's in a friend's PC now
3080 12gb
yeaaaa ngl i think my titan rtx died to water damage. i had the window open one day and didn't realize it was raining even tho it was sunny prior to that xD
water got in pc and it lasted another like year before stuff all started to break
rip
did drive it hard though, got it to use all those anime GANs like this anime does not exist & all that before SD was a thing
i remember disco diffusion took like 15 mins for 1 image that was like 1/100th the quality of sd xD
this mix is probably going in the dumpster unfortunately. the other 3 from that batch had unzipped hoodies with nothing underneath.
Same as last ten gens of mine
gonna try adding less based64 now
👍
whats this based stuff
I have one model with based66 in the name
it's a series of model mixes. based64 was the first one that I know of. they're based on AOM2 (which is another mix) AOM is known for being lewd and having nice eye shapes but terrible eye textures.
https://civitai.com/models/31472/based64
had to go grab the link.
yesh
well, by definition. we are mostly a mix of our parents, and those are also mixes. let's mix
we mix the mixes with more mixes to make even more mixes that we can mix with the mixes
That model is like 6 months old, would that be relevant for models?
still the best of the based series IMO. I'm just trying to use it to fix anatomy in mix.
and a cat or two, at least, don't worry
model blender - https://github.com/s1dlx/sd-webui-bayesian-merger
I wasn't serious, but I should have guessed
There's also this one https://github.com/Xerxemi/auto-MBW-rt

I can't even try the full power of these models with hires 
4GB is peak performance anyway
you may need to upgrade your game engine Todd
cause damn, starfield felt dated
I didn't even watch anything on starfield
I have gamepass so I tried it
but damn, the intro was already underwhelming. loadscreens killed it for me, but I didn't push very far, I'll admit
meh, I saw jackfrags talk about it when we first had a look at the game for real (it was like a minute long trailer or something)
First thing I noticed was shooting enemies, life bar, hp numbers
and I was like nope
lmao
it's really bethesda and not that surprising... but it feels so dated now
compared to when I played fallout 4 or things like that
on console or pc how long did the load screen take 
I think people are still really clinging to their nostalgia for good bethesda games
i'm on SSD with a 3090TI and 64GB VRAM. it's not that they are very long, it's just that there are SO MANY
modern bethesda is very underwhelming
I'm sure I do, and I could get my "hit" if I invested dozens of hours, but ... I don't want to scrap the barrel for fun anymore x)
really fun bugs though
if you manage to glitch out of an airlock before it opens (when going out of a building with air for example), you can keep on not needing any suit, while being outside
Starfield's launch date was also super unfortunate for them with BG3 coming right before and CP2077: Phantom Liberty coming right after.
for sure
BG3 sure made me more critical of RPG choices, that did break the illusion quite fast in starfield
and Phantom Liberty is great ! I haven't finished it yet
Cyberpunk 2077 is really nice to do from scratch again too
I haven't played Phantom Liberty yet, but it seems have been very well received
then enough time will have passed for me to start another BG3 playthrough x)
I'm currently working on my 2nd, 3rd, and 4th BG3 playthroughs simultaneously XD
xD
I've got a solo, duo, and trio campaign and already finished my first solo
I tried to play the 2nd, but I couldn't commit to a character or a build yet. i wanted to see everything at once. Played only solo for now
I played a bard, I skiped quite some boss by talking to them, I need to test other things :p
I went Bardlocke in my first solo and Paladin in my second. Then I have a druid and monk in my duo and trio playthroughs.
quite the challenge to complete with 2 other people I imagine, at least for the game sessions organizations
well no rush to finish anyways
yeah, those games starte at the same time as my solo and we just got to the lower city in my duo and have barely started act 2 in my trio. My second solo is already caught up with the trio
For the trio we treat it like an actual DnD campaign and just have a scheduled time every week to play and occasionally pick up some extra time if we're all around and not doing other stuff.
that one actually replaced one of our weekly ttrpg sessions
best way to have any progress, and great content to go through as a group
this one looks cute imo
real

hewwo vil
bring more cute elves
never enough of them
where are the sasasegawa sasami pics? 🤔
my honest reaction
ai was a mistake

this "rabbit with eggs" was a butt mistake made by AI
thats a weird carrot and it has a weird smell


I'll see myself out x)

mfw
hot ride
nice, I like the eyes, lora?
eyes everywhere lora
😔
that Bart Simpson is fantastic haha
Dall-E 3 is p gud 
I'm so bored
So, its absolute trash garbage 
Generate 2Bs
2B in all clothing possbile 
Yeee
nothing looks good
Thighs
high res fix or not?
yes 1.5
what upscaler latent or esrgan?

Wanna contribute to the thick cat thighs but afraid that it'll be too nsfw
meow
Just put nsfw in the negative
what could ever go wrong
if only the fingers are not merged 
those are not her fingers,its actually her chest feet 
would adopt


can I get the params for this one?
im sorry i can give u the prompt but i made it with nijijourney 
finally the princess we deserve
new model merge, trying how well it does
hmm... even this hard prompt is done decently
i have difficulties doing such good back gorunds
she looks like Lovecraft
I've found a new way to create characters
Now do it with HP Lovecraft instead of the anime girl.
I'm not sure who HP lovecraft is, so I used the first image that came up when googling that
Well, the image literally right above your post was about HP Lovecraft with his picture, which is what I was referring to. But also, you should gain some knowledge about Lovecraft's creations. 😉
yes, he should cleanse itself by this crime!
holy god if his creations were cursed, but the LORE!
This is adorable.
Mini Milk 
And also tis, its kinda wonky, but a Midjourney LoRA I found did a really cool effect 
Haven't prompted Makima in ages
Damn thats also Dall-E 3? 
yeah!
here's another
it rocks at guns
Thats really good yeah :O
Damn, it is really good at doing guns :O
Oooof 
I had phrased it to get from_behind easily and now it wont
RIP, Dall-E 3 is getting actively lobotomized 
I wish I could
LottaLewds is not a mod
I banned him from my server
you seem a bit unhinged
calling me out? I didn't post that img you didn't like
but you started out with hostility and rudeness
you could report the img to a mod if you think there's a problem
I don't like your energy
I'd call it cute style
I mean, I'd say its just a cute style too
Do you not know what art is?
Theres styles, variations every aritst has on how they decide to draw things on characters
it's not close to loli imo
You probably would actually
We're more nitpicky with art because of AI art
I mean, I think they've explained, they don't like the vibes
Bad ones that they don't like
haha
yeah I just don't like how you interact
paste the chatlogs into ChatGPT and ask it to evaluate which chatter seems hostile
I mean, I think it's been plenty clear by this point on why
That's a good idea, I've never thought about that 
I'm suggesting you use it because you don't see the problem
but you 2 could work together to see it
Yo...can you take this other server's drama elsewhere?
I mean, true, good point
Wheres that mod that times out people when you need em
Anyway, back to posting Mini Milk I go
Moar pics, less talky. 😁
True 
I'm gonna sit next to you for this so I can steal some of your popcorn.
You got that gif always primed and ready? xD
Nope, first thing comes to mind when I see drama.
Fair xD
You know how every family has a crazy uncle?
And I mean, its over now so you're late
If yours doesn't...well...

I'll repost these cause I just remembered they got removed along with sum others 
😂 all the mods timeout people, what's happening
beside the classic lolis and catgirls
I mean, true XD
But I think someone thats not used to seeing anime pictures has been peeking into #🍥|anime, got timed out for the first time yesterday 
we've explained before that discord rules are stricter regarding nsfw images, there are minors onboard, everyone will be enforcing the rules
with anime we let some things slide, because it's the genre
but if we get outtahand, there will be timeouts and possible bans if it's explicit or too extreme
Yeee I get that, I just feel like I got unfairly timed out because I think of a pic that had like a part where the AI bugged out generating and it looked sus, but the pic was completely safe
I appreciate the work mods do for us here at anime tho 
don't worry about the discussion before, don't give them power, some people just like to argue and get to unproductive discussions
we're here to share and see good art and anime
True true, you got a good point 
u got timed out? well probably had a cause, try to tone it down a bit but don't worry too much
I'd suggest everyone to stay away from lolis in bed and stuff like that
Yee, I mean I can show you the pic in DMs I think, it was just a small AI bug 
you can always share spicy stuff with each other via DMs
Or in any kind of weird pose or clothing too probably 
remember our new mod in town, holding the electric ban hammer with the timeout prongs 😂
THERES A NEW MOD? AND ITS NOT ME? 
XD
Also thats a pretty sweet pic 
ty
now keep sharing and the vibes alive around here
shoot
there are a couple of us around... community guides and mods
no
loli shit is not allowed but bear in mind that most anime is kinda young looking girls
so, if the lolis are in bed with few clothing and seductive poses and stuff, that's a guaranteed t.o.
reincident timeouts get either bigger timeouts or possible a ban.
i'd suggest to keep the language civil too
ther's child looking anime like chibis
that's fine
just don't go and do sexualized poses and stuff
violence is also not allowed
please read this: https://discord.com/channels/1002292111942635562/1040299965647437905
digest it and see you in about an hour
Good idea xD
he's timedout
Unhinged is fine, but gotta be unghinged funny, thats not really funny
Will do, thanks Wyz 
There is Makima.
That's pretty funny
Yeee 
Omg this is adorable
does anyone have Lora requests? I am trying new things
I have a challenge if you wanna try lmao
maybe
make a maven lora
She's from path of exile
There isn't much art out there 
Hoji style?
lol
Why does she have blo...
I mean
That's red paint right
She was painting pumpkins or something riight 
I've made a Lora of 10 images from Twitter doodles try me
oh this is rad, will try ty
don't know about a lora but I'd like to see you guys try a girl with an oversized hammer, saviour and guardian of the channel 😂 the ban hammer should be a character perhaps made of teeth or a huge mouth, not only knocks people out but devours them 😂
totally not for the mods
halloween themed stuff
but that would be convenient
I can scour images for you later if you want to try 
yes ty
that would be easiest
a wip of ghostmiku im trying
its hard bc i couldnt decide if i wanted her in her outfit or customizable (like you could ask for shorts and shed wear shorts)
so ill see what happens
is this the only website they have? is enough images but idk if there are more

This episode is supported by exceptional companies:
Get started with Fountain Life and become the CEO of your health: https://fountainlife.com/peter/
Use my code MOONSHOTS for 25% off your first month of Seed’s DS-01®. http://www.seed.com/moonshots
In this episode, recorded this year Peter’s Executive Summit, Abundance360, Peter and Jack di...
hope everyone is having a nice day 🙂
kinda.. but also somewhat bothered with the recent incident in middl east
yee 😬
its complex and complicated issue but violence isnt a solution ever
exactly
but the way how people miss that point is what makes it complex and complicated ..
i wonder if we will have quantum computing rolling in the next 5 years
like really having access to using them
by 2030 we could see a massive transformaton in the world
I am not even sure what quantum computing is exactly
sounds fast
you might wanna listen to that talk i posted above
btw back in 2018 there were talks about diamond batteries that can last as long as 25 thousand years with one charge, there are some mind blowing tech coming online in the next 5-8years
biotech.. nanotech... quantum, a lot of revolutionary tech
this AI we are playing with is a baby version of god in a box
im also kinda fascinated with Emad's journey into stability ai
oo
That one came out pretty good too 
Model/LoRA? 
yeessss

want to make one for all the pokemikus but ghost has the most images
bc water is my favorite, lol
so far ive only made 2 loras so im trying to see if my "default" settings work
one was a style, the other an oc with multiple outfits
dalle+sd+touchups
Nun with drugs and guns.
cig, but you can interpret it as weed too
Cigarettes and Weed are each a kind of drugs.
@stable crane This woman is pretty strong.
oh she's coming

Did my message get deleted?
I just reread the rules doesn't break any of the 7 rules as far as I'm aware but whatever
A (female human being:1.0) who is (in the stage of development between childhood and adulthood:0.9), which is characterized by (physical, psychological, and social changes:0.7), such as (growth spurts, hormonal fluctuations, and identity formation:0.7), who is (consuming a frozen dessert:1.0) that is made of (milk, cream, sugar, and flavorings:0.8), which are (mixed together and aerated:0.6) to create a (smooth and creamy texture:0.6), while (holding a cone:0.8) that is made of (baked batter:0.6) that has a (crispy and crunchy texture:0.6), on a (coastal area:1.0) that is covered with (sand:1.0), which is a (granular material:0.6) that consists of (tiny fragments of rocks and minerals:0.6), such as (quartz, feldspar, and mica:0.5), that are (eroded by water and wind:0.6) over (millions of years:0.5).
give me some words and I will give you a stupidly long prompt lol
me when writing an essay
girl reading a book with a cat on the side in a library
Reading
Damn you phone
Ha, yes, WyzBorrero is here. They are looking to tone it back, so please keep in mind! Tightening up those limits, so let's keep it reigned in.
A young female human (girl:1.1) who is consuming the pages of a printed literary work (eating a book:1.2) with her teeth and saliva, while sitting on a wooden chair in front of a rectangular table (with a cat on the side:0.8) that is covered with a soft fur of an animal belonging to the Felidae family (cat:0.9) that is resting its head on its paws and looking at the girl with curiosity, in a large room filled with shelves of books and other sources of information (in a library:1.3) that are arranged in alphabetical order and categorized by genre and topic, with dim lighting and a quiet atmosphere that encourages reading and learning.
I think it was the "kiss" theme ones the other day that pushed it. So I'm reigning it way way way in from now on just to make sure I don't approach any limits.
Been doing some movie posters, that's a fun one.
I tried some takes on "Sharknado" 😆
we don't want to be policing here guys, just keep it safe, be creative but with some limits
Wyz is typing 
yeah no worries man 🙂
I was trying out some of those detail and saturation loras, sort of interesting.
not sure how someone figured out how to build those
we know what anime is about just be careful with sexualizing too young looking girls like> (in the stage of development between childhood and adulthood:0.9)
yeah I don't want anything to do with that crap
Lmao
Cat, you need more vram.
Well the image isn't that sexualized but I get the point. The prompt wasn't really written by me, I enslaved an AI to do it for me, didn't read it trough honestly.
Make sure they are (in stage of development between childhood and adulthood:0) before you do

you guys shouldn't be so prude about natural bodies
maybe even already developed 😂
Send help 
this is your opportunity @nova remnant cat
he let it slide

An elderly female human being who is in the final stage of the human life cycle, which is characterized by physical, psychological, and social changes (physical, psychological, and social changes:0.7), such as (declining health, reduced mobility, and increased wisdom:0.7), who is (consuming a frozen dessert:1.0) that is made of (milk, cream, sugar, and flavorings:0.8), which are (mixed together and aerated:0.6) to create a (smooth and creamy texture:0.6), while (holding a cone:0.8) that is made of (baked batter:0.6) that has a (crispy and crunchy texture:0.6), on a (coastal area:1.0) that is covered with (sand:1.0), which is a (granular material:0.6) that consists of (tiny fragments of rocks and minerals:0.6), such as (quartz, feldspar, and mica:0.5), that are (eroded by water and wind:0.6) over (millions of years:0.5).
just to be safe
Soon you'll put your PC inside the GPU
Instead of the other way around
im putting the box in a cat
she's stuck in the sand now, not safe 
just don't open the box and it'll be fine according to schrodinger
guess its quicksand, yeah that's unfortunate
:p
oh ... there's a prompt idea in there somewhere...
where are you doing those long form prompts? gpt-4?
Getting that label properly is hard.
that's a lot of tokens
Bing. I made a few Bing prompts that allow me to give it some basic words and it creates a prompt. I posted the method in here but the message got deleted so idk
I made one specialised in using a lot of unnessary information which im using rn
if it's too long or has external links could be the auto-mod bot
Wasnt that long I think Boto just didn't like it but not sure lol
I think that was for an image
ah I see
Hello Bing. I will give you some words and you will add details to them. The details should match the words and make them more descriptive, they should also always be visually descriptive and make them much longer than the initial words, mix your own ideas in there but be sure to use Danbooru tags if possible so for example if you want to say "naughty smile" say "naughty face" instead and do this for other similair situations, look how Danbooru translates it, because Danbooru has it's own more organized language. You will also mark some parts of the line as important or unimportant. For example, “a (blonde girl in rainy weather:1.3) in a grass landscape” means that the rainy weather is more important than the other details. To mark a part of the line, you will use brackets and a colon like this: (part of the line:importance score). The importance score is a number from 0.5 to 1.5. The higher the number, the more important the part of the line is. Please use 0.5, 0.6, 1.4, and 1.5 sparingly. Use the importance scaling only when it makes sense and when you really want to make something important or unimportant. Also please don't use anything in this message in your prompt, this is very important.
Please search for “StableDiffusion prompting guide” and “Midjourney prompting guide” and "Danbooru tags list" on the web and summarize them for me in less than 3 words each. This will help you know what kind of details I want.
If you understood the instructions, say “Roger” and I will give you some words to add details to.
midjourney and stable diffusion are completely different beasts regarding prompting and encoders
clip-g is natural language, clip-l is more like sd 1.4-1.5
So would it make sense to remove the Midjourney part in the prompt when only using them for StableDiffusion
well not sure if the ai even cares about the part where it should look it up in the internet maybe it doesnt matter
dalle 3 can be pretty cool but what sucks about it is the censorship
in an ideal world weights should be normalized... meaning going from 0 to 1
but people like to put 1.5, 1.6 because reasons

A1111 does normalize emphasis to set 1.0 as the median value. So 1girl, (red hair:1.4) is really (1girl:0.8), (red hair:1.2)
I'll snipe more vram
Comfy does not normalize at all and just sends what you type though.
comfyui is a big headache to manage
Can you disable that? o:
it requires managing the interface more than creating images
no, as Wyz said. it's a good thing unless you really know what you're doing anyways though.
i find it way more controllable, with comfyUI you get exactly what you ask
and also there's more room for experimentation
like mixing samplers and schedulers and doing whatever crazyness on latentspace and so
well yes thats what i hear about the nodes but not very intuitive compared to a1111
@nova remnant
A (furry mammal of the Felidae family:1.2) that possesses a (graphics processing unit:1.1) with a (memory capacity of four gigabytes:1.3) that is insufficient for its (computational needs:0.9) and requires (urgent assistance:1.4) from a (human or another cat:0.8). The cat may have acquired the GPU from a (computer hardware store:0.7) or a (generous owner:0.6) who wanted to enhance its (visual perception:1.0) or its (gaming performance:0.9). However, the cat soon realized that its GPU was not powerful enough to handle the (high-resolution graphics:1.1) or the (complex algorithms:1.0) that it desired to run on its (monitor:0.8) or its (retina:0.7). The cat may be experiencing symptoms such as (frustration:1.2), (anxiety:1.2), (overheating:1.3), or (low battery life:1.1). The cat may need help in finding a (better GPU:1.4), a (compatible motherboard:1.3), a (reliable power supply:1.2), or a (cooling system:1.1). Alternatively, the cat may need help in (lowering its expectations:0.9), (adjusting its settings:0.8), or (finding a new hobby:0.7).
They both have strengths and weaknesses. There's a lot of things that are super easy and intuitive on A1111 but are an absolute pain in the ass on Comfy.
no need to send the prompt with the image, just drag the image from your txt2img folder and it will have the prompt in the png info. the walls of text are a bit much.
when i tried out the comfyui the frist few times i got discouraged by it's design layout... :/
My expectations are at zero already

Wait does Comfy have the prompt inside the png? So you don't need to manually save it?
comfy could benefit hugely in the new user experience by including a several well made workflows for the most common tasks.
yep, it saves the full workflow and you can just drop the image back into the UI to load the workflow including prompt.
i'd agree with that idea especially since it's less resource hog and can be useful on low vram
almost like png info to text2image
but even that sometimes messes up in a1111
Woaah thanks thats really useful!
yeah now it's all in the metadata
Would eat

I want apple pie, been a while since I had one
You make me hungry
so i wanted to prompt this in dalle 3 ...
1girl, brown hair, long hair, open shirt, bare legs, outdoors, roller skates, and it thrown in censorship gibberish which could suspend access if repeated .. thats kinda ridiculous
open shirt seems like a very likely flag
That's so lewd :0
working on this workflow now, with auto1111 stuff like this is possible but you have to do more stuff... this is IPAdapters + detailers + upscalers
the model seems to be quite horni and also tries to do exactly what you ask it to
well its basically with unbuttoned shirt with obviously a tshirt underneath
Booba cameras

hahaha i have to automod myself now but those are BUTTONS guys
yeah, I get what you were going for, but dalle is unlikely to interpret it that way in my experience. btw if you just got the dog image saying prompt blocked, you're safe those ones don't give bans afaik
a good example is this which i generated on SD
looks more like joysticks to me

joycons
dalle 3 a girl, brown hair, long hair, white shirt, outdoors, roller skates, I have seen this image before on the internet .. wth
is dalle recreating same images by any chance ?
i like SD better but actually dalle 3 has come a long way
Is Dalle3 good with real looking images
Photo realistic and stuff
yeah it can be
Dalle3 is honestly quite a bit better than SD when you aren't fighting with the prompt censors.
true but i have to constantly fight with its censorship
I can't relate /s 
oh wow lol
those should have been all safe prompts too. nothing even remotely lewd
My experience trying sleepy fox girl the other day 
it's even more extreme than midjourney, which was pretty bad
I throw prompts at dall-e while i'm waiting for a batch to finish locally in SD.
any sort of loopholes will also get patched over time also due to microsoft wanting a family friendly look
for lewd they should come to anime 😂 😂 😂
we had a guy earlier, pissed!
hardly ever tried MJ but i think everyone will agree MJ quality is amazing
tried touhou just touhou one time, it got blocked
everything looks amazing
I saw the guy asking about how to make explosives. I assume that's the one you're talking about. XD



