#🍥|anime
1 messages · Page 141 of 1
filters on things that make men lose control
openai policy is since girls are generally beautiful certain aspect of girls are censored
I'm saving this hilarious quote for future use.
Right in the "Stuff People Say" folder it goes.
well usually men dont lose control easily but that's how dalle interprets feminine beauty
definitely not stolen pngs
Thick 
haha thick walls
oh yes
Can you do older
or a more ruined/battleground
background
like battle wrecked zone?
Yes or just older stuff in ruins
ohh
also inside an arena
That would look cool I need to try that later
1girl, roman warrior costume, battlegrounds, ruins, wreckage and weapons,
let me see what that renders
not too epic yet
Is that a gun in the corner lmao
And jet trails above. She's just a time traveler, no worries.
Fire the tattoo artist.
Unless it's meant to show a constipated cat. It is over her abdomen, after all.
When the eyes are going completely the opposite direction of each other, the angle doesn't matter lol.
Its the muscles pushing them around
I'm liking this Roman arc we're on.
hello purple man 
meow
Well uh... that didn't turn out exactly like I thought it would. I wanted an assassin but got
looks neat
Nice asse

@untold glacier
better cat I think
also no idea why they have such weird poses, im trying "standing"
That cat has the full 4gb of ram.
nice lol
dalle give you this? lol
sᴏ ᴄʟᴏsᴇ ᴛᴏ ᴘᴇʀғᴇᴄᴛ. ʏᴇᴛ sᴏ ғᴀʀ...
Yes

Hello lotta!
I organized my mouse folder

Many of them, of all classes
They are twins

Don't play with lightning, kids.
Fabulous results here.
Good to see ya Smol
yoyo
That would be hot if that happens. 😛
Mmmmmilk tea??
Disgraceful
Magical mouse deal with him


Beanie bunny ears
Bank robber gloves
And student sweater
normal cute girl

How is she getting more adorable right now?

I'm leaving 'those things' to Sukuna and I will not bother to help, in fact I'd rather run like hell even harder.
Now that is groovy.
I recommend just reacting with ⚠️ on each of these and not interacting, peeps.
It's certainly nsfw. I'd say nsfl, but that's not featured in the rules.
is it nsfw? more nsfl
nsfl is nsfw.
I mean I wouldn't be caught watching this whole channel at work
so it's all nsfw for me ;p
horror?
oh boy
nice teeth 
They look lovely.
Terraria blood moon bunnies!
ok thats freaking me out I may mute you so I dont' see it... shivers
its ok dalle said it was safe 
no thats nitemare fuel... not safe. no no
@native halo time to get sacrificed for the cursed images you brought
decent hands 
cracks open one of these bad boys
lemme see that cola
That's it
Mouse girl give me something so I can unsee those 
i see things clearly now
Beautiful stuff.
As for the cursed images...
my arm hurts
You're going to jail Steven
ok ill see myself out 
We miss this adorable lady. 😭
Are they twins?
Triplets there?
yes 
my bedroom is in dire need of cleaning
I'll need both
@native halo After having to warn multiple times in this channel, with you included-- you have been timed out for 24hr after consistently posting nsfw/gore/horror and being told by multiple members to pls spoiler tag and or not post-- Please keep the rules in mind and keep it safe in the channel.
24 hours
miku tree
i thought you were supposed to be sleeping 
I know I know. going to bed. but first
/dream眼镜.脸红
bot over #1100170312106127410

还要加别的什么才能生成嘛
sorry i yoinked your prompt!
typing ./dream and press this
/dreamprompt?
wait what?!
press that, then type in your prompt
i got that as well, but after putting . infront of the command, it works
look
maybe translate it to english first?
OK
he's done
missed something? someone generated something creepy? XD
but guys please, we've talked about the nsfw stuff on multiple occasions and keeps happening, as much as anyone wants to see miku in bed wearing lingerie and see-thru gawns, please share those via DMs, we won't tolerate that anymore on the channel. And I am going to ping @native halo @nova remnant @nova remnant and the rest, horny jail is a reality
anyway, better that the channel is maintained clean of obscenity of any form! we have enough of it anywhere else!
I’ll try to remind them if I see any
What did I do 
@lyric zealot 
Gen one of those on top of a moving train
or anything that is moving
that's probably super hard for SD1.5 model
are you prompting a fisheye lens in that one? looks cool the distortion
Someone rannnnnng?
yup, fisheye is in prompt 
taken care of Sunny, had to ban him
You're free to ping me--no problem
can confirm it's not the first time 

These are super adorable
yeah the chibi mikus rock
my chibi isn't as cute as Vil's tho
the weird thing is that those were the ones who triggered sasanek to insult and call pedo everyone here
Mods wanting wine surely mouse girl version 
imma be honest, the mouse ears look like a speaker
🐱 ?
Those are cute mouse ears, tbh
They are biig
i'm more worried about the legs and knees
probably SD1.5 models fault 
What do you mean?
those shadows with a backlight make a weird thing
👍
those leggings/socks are like two sizes too small 😂 those are going to leave marks for a while
that's seriously the reason why i disliked using thighhighs
Speaker girl on the job
I always found those cute
But you saying they look weird makes it look funny to me
xd
i like it, but i kinda dislike how it's like a few size too small
like if it's just nice, then 
look at her face!!! she's like looking directly at Smol saying... dude I can't even walk straight on this
imma refrain from making ||nsfw|| jokes 
i bet sdxl model can do this fishnet stuff better tho
Serious mouse x3


Most of these have a super serious look lmao
Angy model
You could try something like scales as a negative of some sort.
Love this one, too
Lmao
Getting the thigh highs treatment
At least they look better than the war one

that's true
Kinda looks like a couch rip But great pose
Easier pose to do without caring about hands 
Easiest *
RIP 🖐️
These are some beautiful hands
True 
Awwwh! That's so beautiful! Love the whole autumn atmosphere.
That was the prompt!
Hahaha, right on!
what hands lora are you guys using? today i made my best one yet 😂
i think only this
but it's because im too lazy to look for better ones
That looks like a Rinner from Tohsaka residence. 😮
A very beautiful woman is here.
A little wonky around the legs, im still working on things 
How is she so fabulous with or without masks?
was trying to go for this vibe
Can i ask ? Is this app can colouring manga or comic ?
it can add colour, but probably not the way you wanted
if you want to be able to turn any manga to coloured version, then this is still lacking in that aspect
super cursed
Cursed horror
Now the face sucks, but the moves 
I made a new Sailor Moon Video, what do you guys think?
My sole halloween post for the season. I love low CFG.
Just in case anyone was curious about upgrading to torch 2.1 or torch 2.2 nightly, here are my benchmarks.
TLDR: speeds are within margin of error so not worth going out of your way to upgrade.
what is that CPU? threadripper?
nah, just a lowly r5 2600
thanks for that data !
I'm also on amd cpu and some python stuff just doesn't want to work, numpy and numexpr give me lots of headaches
as did the first torch 2.0 version
I had some issues with early torch 2.0 stuff too. Wasn't sure if that was just me not really knowing what I was doing with pip at the time but I kept ending up with a broken venv
what model are you using by chance?
I really dig these images.
Mommy Miku
Miku is Magical, another image that proves it.
Mommy Miku 2
Sponsored by Aáiddas ™️
thx for the nice hand....
Classic skirt hand
the pickpocket hand xD
True XD
smile dont make me bonk you
The lvl 20 rogue perk, grow a specialized pickpocket hand
Me when fruit wants to bonk me 
how dare you fruit
Cyan 
True 
Cyan in my cutesy style
vampire with mouse ears? Mouspire?

How do I read this?
Just the third and fourth rows? What do the numbers represent?
Mob psycho vibes
Very good
This one is nice
let me google what that is
hmm
They are the same species :p
Not that much similar
Thanks
my thoughts exactly lmao
Column 2 is the effective speed at 1/2/4/8/16 batch size and 512x512 resolution (I think it uses euler a 20 steps, but since it's just a comparison that part doesn't matter much). The values are not the actual it/s like you would see when genning but rather the effective it/s (it/sbatch_size*) since that is more useful for comparing the speed across different batch sizes.
Column 5 lists the installed libraries.
Column 7 lists the active optimization.
so like
on the first one you're basically getting 19,48 it/s ? >.>
with 16 batch size
also 16 that is pretty crazy
Is there a world where upgrading torch can help my poor gpu 
yep. It would show at 1.21 it/s, but since it's doing 16 images at a time it's effectively 19.48 it/s
most of what I do is very GPU heavy with low CPU demand so it's fine for me. There's only a few games that I play that the CPU bottlenecks.
The only outlier seems to be batch size 1 torch 2.2 ?
a tiny upgrade
yeah torch 2.2 at batch size 1 with sdp optimization was consistently a little faster. I didn't get xformers working on torch 2.2 though. The difference is meaningless in the real world with that gen completing in ~2 seconds on any of them.
wait
What does xformers do exactly
I thought it increased speed >.>
increased speed and better memory efficiency. sdp does the same thing, but is a bit less memory efficient in my experience.
ohh
I mean, I sometimes get a 2~ second diference from gen to gen
I reran all of those benchmarks 4 times each to make sure that I could watch for fluctuations.
wow really nice, how long this gen take?
My time or gpu
you do the inpaints at the very end?
@supple raptor
should I change this stuff
TAESD will significantly reduce vram usage, but also reduces quality in general. with 4gb I would definitely run TAESD.
Any other settings you can think of that would help vram ?
This is on 1.6 im trying to see if I can get it to work as well as it does on 1.5.2
token merging would help with speed, I don't think it really efects vram usage though.
civitai dead?
Idk whats up with civitai
looks like it runs on a potato
its up for me
you're buzzing
I'm moving up in life
hopefully one day they at least fix the search in that site
thoughs on the hands adetailer? crap?
its so bad
haha
hands adetailer looks like it doesn't know what to do
it can't see the whole picture
seems to strictly make things worse for me
my face when out of vram 
RuntimeError: Could not allocate tensor with 301989888 bytes. There is not enough GPU video memory available!
angy fox girl
1.5.2 I can do 768x768 but 1.6 I can't

doesnt work on my amd
dalle?
Yup
Yea; honestly a lot less restricted than ppl think it is, bikini armor/blood etc is all fine
got a bunch in the characters channel and some in the general and sdxl spots
Just censors the weirdest stuff like cherry blossoms
yeah I think they eased up some
They might’ve cuz there was a tweet from a dev about the overly strict soft blocks
a couple days ago the generation waits were basically 0, so after my 'tokens' were used, I was able to just keep going. It eventually blocked me from making any more but I probably did 100.
That’s how it always is for me, zero waits but maybe after like 300 gens it gives a “you can’t gen anymore” for several hours or til the next day
yeah
the error message was full on wrong though.. something about waiting for existing prompts to complete.
It's pretty close to being able to do animation sprite sheets. Once we get there, I'm gonna be hooked. 😛
the game's we'll be able to make with this...
ooof, and tonight, i can't make anything without tokens at all.
oof 10min wait
omg
my 2x upscales were taking 20mins instead of 1min33sec due to batch size on tiled diffusion
just noticed default is now 4
wompwomp
yeah I was wondering what was up
Wait, so even with the latest update there's still no support for SDXL??
I'm using 1.5
quiet tonight
ㅤ
ô¿ô
Yeah, I've just been waiting for months for the creator to make good on their statement to update to support SDXL "once the SDXL custom checkpoints come".
Here's today's Tifa.
the adetailer models are only responsible for identifying where the face, hands, or full people are. in that regard the hands adetailer models work very well. the problem is that inpainting hands very rarely improves them, and usually makes them worse unless they were already good to begin with.
What Nubby said. The reason it's usually worse than before is because with inpainting it has even less context than it did for proper hand positioning and placement. It doesn't know which way the arm or body is pointing as part of the gen process so (depending on the denoise factor) you'll end up with backwards hands, strange positioned hands, or even feet. (So many feet for hands.
)
at least the face one is always solid
has even less context
not if you use "whole image" rather?
uncheck "Inpaint only masked"
but it still sucks
Adetailer doesn't take that into consideration by default if I remember correctly. And the "whole image" thing will create its own problems.
seems easier to reroll the img with slight variation amt
For hands, I'd say yep. Slightly changing CFG or altering the seed to a decimal point.
you don't use the variation amt slider?
Hires fix and upscaling can help them though, especially when using the targeted region and prompting things in for characters when using Tiled Diffusion.
I do not. Never looked into it.
IDK how it works but it's in "Extras"
it can give you slight variations or larger ones, depending on the slider, and has its own seed
Oktoberfest for sure.
The person_yolo8n adetailer model can help with hands sometimes. If the primary reason that the hands are bad is that the character is too small. That one will mask the full character and inpaint just the area the character is in at a higher resolution so it gets the full context of the person and can increase the resolution that the character is being generated at in some cases.
Nice. Yeah, that's essentially automating the process I described with using the tiled diffusion targeted regions and manually doing it yourself. I should try that sometime.
Have you had much luck with it?
yeah, I like to use it for some of my super wide images. where the character is far away.
How often does it think there's a person there and there isn't? I've had to up the min recognition of my face adetailer model to 0.5 recently. I was getting too many faces in stuffed animals and slightly open doors. 😅
I don't know that the person one has ever done that for me. The face one does it pretty often though. I've just embraced the fun faces that it adds to rocks occasionally.
((what is that ball from the movie called again with a face painted on it??)) that thing
Heh. Makes me think of my dad, a painter, in whose early works he put a lot of faces in trees and background elements. Funny that we get that as an issue in AI generations.
lmao
WILSOOOOON
I honestly had tears from that moment in the movie.
LOL. Of course it's a lora.
Here's one that I coulda used that adetailer model on. (the background person)
Though, it probably wouldn't have helped the off height comparison.
it'd probably help with both of them to be honest, the closer one doesn't really need it, but it would still benefit.
Fixed it for ya
You need to appreciate what's beautiful on things
Well... I guess if you like yandere.

yandere huh...
How it feels to use ultra wide monitors
You're not wrong.
if it wasn't so bright, I would make that one my wallpaper for the winter. I run darkmode everything though and really don't want to be flashbanged every time I go to my desktop.
Just a second
Quick, invert it and throw it into i2i. 
oo mouse gladiator
maybe you can do her in the "this is sparta!" kick pose 😛
I could
I kinda want one, but I worry about window management
ok that's enough
win11 is pretty good with the zones now with powertoys
The ᴛʀᴜᴇ experience.
this is all I get for the topic XD
I run discord on one half of my 2nd monitor which is only 1080p.
I have a lot of shit open all the time.
Window management is as easy as the hotkeys and zones you use with Windows. (If you have Windows.)
Snap is your friend.
my youtube browser window lives on the left of the 2nd screen, discord on the right then on my main monitor I have webui on the right, cmd in top left and my explorer window in the bottom left.
gaming is where I would hate it. I like to have discord visible on my secondary monitor, but don't want to have to play in windowed mode.
90% of the time I prefer games on windowed
Borderless windowed is far better these days then it ever used to be. I can't think of a game that doesn't run better or just as well as fullscreen mode.
It used to be super terrible though, with a major FPS drop in most games compared to regular fullscreen.
They are not the best on optimizations lmao
Heh
Sorry, I shouldn't laugh at other people's vram. I'm super lucky to own my card.
not skinny enough
I do like this style.
I need to (thin:1.5) next time
(very skinny legs:1.5),
for negative prompts: fat, obese, overweight, thick thighs, curvy

The secrete!
Last one 
Nice
Anyone know of any good LLM cards for an Overlord-type experience? There's so much crap to wade through on the pop sites that I can't find any good ones.
can you translate what's that
Eh... LLM = Large Language Model. I have a local LLM where you can run AI models like chatgpt that are made for specific purposes. Things like roleplaying, being an AI assistant, prompt generating, etc. People make "cards" which are sort of preset configurations of pre-prompts that the AI uses to give you the experience you want.
For instance, there's one that starts with "{{char}} is a writer and will help {{user}} by creating new content for a fantasy universe." where you can talk to the AI to have it help you create a world, character, or other RPG elements.
That's the short definition anyway. 😅
Hmm
@nova remnant chopsticks.
Kinda like models for SD
Pat pat pat
The model I like using is very bad at chopsticks

It's also RNG
There a tree coming out of the little fountain thing

smoooooooool
probably don't have to reply to you since you already left the server, but just in case you come back metadata
sus

beeg legit
What's that
trying to sign up for free oracle account
it was doing maintenance till 17 Oct 2359
then after that, it changes to the above
rip. good luck. oracle was an absolute pain in the ass to sign up for.

how did I manage to reply to that message
Huh
sorry for wrong reply cat


well, since i just realised AWS free tier EC2 instance is only for a year

hopefully Oracle isn't 
oracle is free forever afaik
if only i can sign up 
I think I saw somewhere else that gives a 1 or 2 free EC2 instances indefinitely. Maybe one of google's things?
I'd say your problem may have to do with this, lol. #🍥|anime message
i found a github that compares services like these, you might be interested @supple raptor https://github.com/cloudcommunity/Cloud-Free-Tier-Comparison
Imagine how it looks on my odyssey g9 
Typing so slowly makes me frustrated when I have to reply or edit sitting on the can. I need my keeeeys.
basically what i'm seeing here is that Oracle is a chad in this space (surprisingly) and I will definitely be keeping that account going. I've been running one of my free ec2 instances for like 4 months now and they just reclaimed it for the first time in the last few weeks. turning it back on was as easy as logging in and clicking start on the management page for the instance.
is there a reason why it was reclaimed? Was it due to inactivity or was it because it's a spot instance?
they'll reclaim free tier shapes from time to time if load necessitates it and usage is low. since all that one is doing is running Prompt-Dog it looks basically idle most of the time.

They send you an email when it happens and you just have to start it back up and it's like nothing happened.
maybe in your promptdog, add a function to add something to database every hour or something

Honestly, if I just have to turn it back on 3-4 times per year when they reclaim it, I can live with that. If it becomes more frequent then I'll figure something out.
true, since it's free stuff anyway
and since your promptdog have been running non-stop for months, you can probably claim the active developer badge
I'd have to put it into a community server. As of now it's only in AnimeAI server and Luna's testing server.
the easier way to do this, is to just create a simple community server for that purpose 
True
Type faster 
noooo not Rem!
pink trees are kinda cool
It looks cool
Like lying down on grass
The hair mixed with it


Hello @nova remnant !
saw you were asking for girl laying on grass lol
actually these are my first attempt in rendering a girl laying on grass
Do "from above" generally makes them less awkward
Lmao
yes these have from above prompt
Hmm
Maybe it's the resolution
hmm they are 768x768
with hires fix
More portrait style might make it less prone to have a distant view
Like focus more on the girl
these images do look a bit like the girl is not really blending in with the surface
especially her head
Ye
That's what I mean by awkward
Like they wanna get up
right i see it
any ideas?
idk how to fix that
Make width 512
And see
ok
And "Lying down, grass" in prompt usually works for me
Hmm
Maybe cowboy shot?
(full body view from above), 1girl, laying on grassy field, flowers, outdoors, blonde hair, blue eyes, crop top, skirt, bare foot,
512
Try full body, view from above, lying down, grass, flowers
And then the girl description
im gonna remove the character description
(full body view from above), (highres, detailed:1.2), (masterpiece), (cinematic, hyper realistic), 1girl, laying on grassy field, flowers, outdoors, <lora:epi_noiseoffset2:0.6>, <lora:more_details:0.4>
bit better i think
These are not 512x768
from haivng removed her attributes that kinda seemed to come out too strongly

its 768x768
What I was saying I think the wider view makes the model want to see the horizon
And the girl tried to get up a bit
To align with the camera angle
odd
how about 512x768

i dont think width is the problem tho
but let me try one
hi smol sus

my pfp or ? 
The emote

oh man
lol
The convenience pillow
whats the explanation @nova remnant
whats up with reduced width?
Hmm
they do look better right?
What do you expect from a wider image
more view
lol
It's usually the horizon
Yeah
Not to see something flat
So the model is prone to that
ahh yes true
interesting
Resolution changes a lot of things
gotta remmeber that concept
For ai
hell yes they do
but yeah this is my first ever to render images of a girl laying on grass
I was making them lay down on a busy street
Was quite funny
lol
i dont want a cute girl get ran over
im very visual
i dont cope with sensitive images too well lol
her head looks much better now
I'm the absolute complete opposite of you

But I'm normal 
you want a cute girl get ran over?
No
you hate girls?
I mean I'm not visual sensitive
Lmao
You chose the worse options xd
lol
well some guys dont like girls which i dont understand really
@nova remnant 512x512 she wants to get up x3
which one?
This one I sent
her head is bit lifted yeah
whats that top she is wearing called ?
thats not a crop top
Headband
Or the shirt
Hmmmmm
thats called shirt? no
Sweater?
nope
has to be some kinda blouse
Shirt-dress (I don't know)

not sure whats it called
Oversized clothes
Long sleeves
I'm pretty sure that's what smol prompts
@lyric zealot any idea whats her top called in this image?
Look at torq messages
Ohh
I thought you meant smols image
I dumb
no mine lol
i would say crop top 
i would proabbly use multiple prompts to describe it, like crop top, collared shirt, tight shirt, tied shirt
interesting thanks 🙂
short sleeves look cute too
might not work that well tho 
damn, thats brilliant
i cant read that last word but thanks
brilliant = amazing
Ah yeah just late at night and i have trubble with letters many thanks
Why?
D:
Also hello 
im dyslexic as shit auto correct is my second half and test to speach and hia cat
is that SDXL btw?
na just the normal one
yes but i still can't figure out what her blouse is exactly
trying to find the name for some clothing?
@gloomy lodge this shirt
Send me this 
crop top button up - something along those lines
gonna try those suggestion
these look very similar
she looks so sweet
with (crop top, button up)
so close
but her crop top is hanging down a bit more than the original image
She drew herself




