🍥｜anime

[nova remnant]

[ivory pebble]

Bot? Like a discord bot?

[glacial bane]

You picked a new style i see

@ivory pebble do you save the styles u use

[ivory pebble]

It's just vibe transfer with different images

It always changes cause I use different images

[glacial bane]

Makima addict i see

[ivory pebble]

It is what it is

[glacial bane]

Find new women

[ivory pebble]

B-but makima

[glacial bane]

There are better anime women than makima

Don't stay in this cave forever

FIND TRUE HAPPINESS

NEW WOMEN NEW CREATIVITY

[ivory pebble]

Yeah but she's hot

[glacial bane]

There's mo hot women dude

[ivory pebble]

Nope

[glacial bane]

Can you just

Generate gojo being fem

[ivory pebble]

[ivory pebble]

Hmmmmmm

[glacial bane]

I like males being females

[ivory pebble]

I'll do that later, trying to generate Jolyne right now

[glacial bane]

JOLYNEEE???

[glacial bane]

that's actually a good choice

Show thou pics after generate JOLYNEEE

[ivory pebble]

Will do

[glacial bane]

Noice

https://tenor.com/view/happy-happy-happy-happy-happy-cat-happy-dancing-cat-gif-716006509349864265

[nova remnant]

[nova remnant]

Yes

Have some big eyes

[ivory pebble]

I see, why not just use like an online generator instead? If it's taking that long to generate you might as well give it a shot

[nova remnant]

Average anime character anatomy

[nova remnant]

I'm lazy

[ivory pebble]

I remember I used to generate on Seaart before I installed A1111/Forge on my PC

[ivory pebble]

Fair 😭

[nova remnant]

Most of those are very bad to use through mobile

I have to check seaart

[ivory pebble]

I haven't used it in about a year or so but It was my favorite at the time

I wanted to see if they added anything new and tbh they still don't update a lot of the loras on that site

But then again it's not that big of a deal I guess

[tepid vortex]

was this a lora or a fine tune?

[vestal crypt]

https://media.discordapp.net/attachments/1027429244478955530/1261038745390678107/00305-146506493.png?ex=66918159&is=66902fd9&hm=016829784ff0103d40216e45570741213a0844400ef5e878412e82d8d62f0704&=&format=webp&quality=lossless&width=454&height=663

[bitter latch]

@gloomy scroll

[bitter latch]

hmm its not wanting to do smally chibi likeyou did

[gloomy scroll]

Small chibi you say

[bitter latch]

haha

oops

was suppose to be inside pizza hut, but it ignored hut

doh!

[bitter latch]

ok he should have been breathing plasma fire, not havint it elsewhere. wtf

[green temple]

i liked the lighting

[steady grail]

yeah, the light are spettacular

[green temple]

indeed !

[steady grail]

i wonder how the work in other settings

[green temple]

cool

bad lighting

[steady grail]

who put jet fuel in my shampo?

[steady grail]

look at my lance, makes zero sense but is rad af XD

[steady pumice]

cute

[lyric zealot]

[nova remnant]

its the bald shiny spot lighting

[bitter latch]

hmmm thats not my prompt....

wtf

now that got crazy

😄

anyone else awake?

[lyric zealot]

https://tenor.com/view/kon-yui-yui-hirasawa-anime-anime-girl-gif-21014430

[bitter latch]

got me a new 27" inch monitor , loving it so far

[lyric zealot]

nice, been having 27 inch since forever, and i can never go back to 24

ultra wide might be my next goal tho

[bitter latch]

caught this one one sale at best buy, its a Acer KA272

[lyric zealot]

good enough for generation, but probably not the best for fps if you ever played those

[bitter latch]

its better than the 23inch acer from walmart acer SA241Y

thats my 2nd monitor

[lyric zealot]

ahhh, the uneven monitor, same as my brother

he got a 27inch and another 34 inch

for me, i just have 2 27 inch

[bitter latch]

right now i'm trying to stay up so I can try to get my sleep schedule back on track

oooh

[green temple]

noice

[bitter latch]

wow

thats better...

changes up the style

[bitter latch]

this is what I was going for

[green temple]

cool

[bitter latch]

tries to hold on for 10 more mins then bed

changes some things in the prompt and uses different style.. oooh

huh almost a magical girl transform effect.. i like it!

changes magical effect and inserts transformation between the two words

[green temple]

yay finally got something good

no , didn`t get the 😄

[bitter latch]

ok last one

[nova remnant]

[green temple]

got the closest thing from what i want

[nova remnant]

when did smol learn to fly

[green temple]

long time ago

[green temple]

nice

[green temple]

[green temple]

[junior quest]

modal?

[bitter latch]

back to more anima-ish

[nova remnant]

Nai

[glacial bane]

I never get tired of these adorable pics

[unkempt cape]

||english or spanish||

[glacial bane]

[junior quest]

got a subscription?

[nova remnant]

It's from the black market

My mafia friend gave me access

Somewhat

[supple raptor]

CirnoPopcorn

[zinc flax]

more angy miku

[lyric zealot]

i just img2img yours

[grizzled moat]

why she never fat

[gloomy scroll]

https://tenor.com/view/imagination-spongebob-squarepants-dreams-magic-gif-12725683

[grizzled moat]

should i unmagic the yue ?

[grizzled moat]

[grizzled moat]

@gloomy scroll help

is this ok ?

[nova remnant]

:oh: it's the neuro steamer

@lyric zealot

[rich tartan]

The weirdest generation i had

[terse ore]

I want to train a Lora which will generate landscape images from this angle (directly from the side). I've never trained a Lora before and there are so many "how to train Lora" on the internet. Any advice?

[green temple]

[lyric zealot]

https://tenor.com/view/yotsuba-nakano-gif-20160566

[junior quest]

SOMEWHAT

[nova remnant]

whatsome

[nova remnant]

https://tenor.com/view/cheesed-to-meet-you-gif-21409625

[supple raptor]

that looks like either SDXL vae decoding an SD1.5 latent or SD1.5 vae decoding an SDXL latent.

[junior quest]

[junior quest]

gimme artist tags pwease

[vestal crypt]

1girl, solo, {artist:hxxg, pigeon666}, [ciloranko, wlop, ask_askzy], wanke, [[haban35, enaa97, au_(d_elete)]],.[[nsfw]], {seductive eyes}, dynamic angle

[rich tartan]

Indeed, i realized tho, putting it into another non-sdxl sampler fixes that issue. That what you see is not what the AI sees, so the information still stays the same I just didn't use the image that got the vae applied

i was fighting to get right fingers and anatomy, using sdxl as a fixer was the greatest i could find

[unkempt cape]

): me need a1111 command line this instant diffusers suck )):

[native halo]

prompt and what model?

[grizzled moat]

[green temple]

wow

[green temple]

[terse ore]

Anyone tried the xinsir Union controlnet?

[steady grail]

what's that?

[green temple]

ugly

[nova remnant]

no

@green temple

[nova remnant]

I added this image in the vibe transfer

now she smoller

mafia girl

she'll steal your prompts

[green temple]

[terse ore]

It's a SDXL controlnet that only needs a pre processor but works as pretty much any controlnet. From what I've seen it's pretty good

[steady grail]

and what controls? the style of different images?

[terse ore]

https://huggingface.co/xinsir/controlnet-union-sdxl-1.0 here the page

[steady grail]

oh, that's good

a one in all model

so you don't need to load and unload the model every time

when doing different things

i wonder if it works well

[terse ore]

Yep it's a dream come true

Yes it's been tested let me find it

https://www.reddit.com/r/StableDiffusion/comments/1e23ong/the_new_controlnet_union_model_does_not_require/

Haven't tried it myself so I just wanted a second opinion. I can't go home yet so no PC for me

[steady grail]

i don't use xl very much

too slow for me

and when i use xl, i'm more inclined in using pony

[terse ore]

O ok yeah you might want to set it and go to sleep and see what it comes up with in the morning if you're doing SDXL

No idea if it works with pony

[stable crane]

pony ftw

[steady grail]

i doubt it, none of the usual lora and cn works properly with it

i would prefer for astra to spend more time in training cn and lora, than training new models

there is so much cool stuff that pony is missing

and it's a big draw back for pony users

one of the reasons i prefer 1.5

[terse ore]

https://www.reddit.com/r/StableDiffusion/comments/1dwlzbn/controlnet_for_pony_diffusion_v6/ I see this depth controlnet for pony. I tried 1.5 again and loved how quick it is

[nova remnant]

there is controlnet for 1.5

[terse ore]

T2i adapters are supposed to be much lighter and quicker

[nova remnant]

just use those if that one doesn't work

but you're comfy idk

is it new tech

[terse ore]

T2i adapters are old

Union controlnet is new

Even with multi controlnet the 1.5 models manages to screw the hands up a lot

[nova remnant]

[steady grail]

sometimes less cn are better than more

and it's not about 1.5, but about the training

[terse ore]

True

[steady grail]

hand pose should be properly tagged

[nova remnant]

There are some furry models for 1.5 that are super good with anatomy

furry

like they do upside down people and all lmao

[steady grail]

yeah, i use furrytoonmix_v10

and it's very cool

[terse ore]

There's a color control node can't remember what it's called. But it helps keep the color consistent in img2img. I think it can help to allow greater denoise but keep the color same

[steady grail]

interesting

i attempted to do something similar but the node simply didn't work to me

[terse ore]

I have it on my PC but I can't see it yet. I can go home tomorrow though

If it works then it can really help in making consistent comics panels

I'll make a note to tag ya

[steady grail]

yup

[terse ore]

If you use comfy, do you know what's better, pixelk upscale or controlnet tile?

[steady grail]

i hate to the bone tile controlnet

[terse ore]

I use blender. My current workflow is to make a screenshot of the panel but also make a screenshot of the face up close by zooming and panning on the face. This lets me keep the same angle. Then I do img2img and upscale on the main image and img2img on the face. Then I remove the background of the face and overlay it on top of the main panel, aligned to the face. That way I can keep the expression of the 3d model the exact same while also getting the scenery denoised

What wrong with the tile controlnet?

[steady grail]

too many mutations >.<

[terse ore]

I see. Even with something like canny with it?

[steady grail]

i prefer way more to directly gen 1344x1344 on krita

i don't know how it works, but it's way better than any upscale i ever did lol

canny is applied on every tile, so it's not possible

[terse ore]

Interesting. I haven't tried krita with SD but I hear a lot of good about it

[steady grail]

it's very practical

it have less functions than comfy, but what it does, does it very well+

and being integraded in a drawing app, region and inpainting are much more easy

i usually gen the base of the image on krita and than refine it on a comfy workflow

[terse ore]

Yeah. I've tried inpainting on a drawing program with SD but I couldn't get other stuff to work. Ah well. I have tried segment anything+grounding dino in a1111 which will generate a mask on whatever you prompt and then run it through the inpaint process. Worked okay so far for me

[steady grail]

with krita you don't have open and close different gui

just load the same model on krita and comfy and you can switch without loosing time

[terse ore]

Convenient

[steady grail]

yup

[junior quest]

@supple raptor made two loras with 2 different datasets, BUT it gets "blush" as the trigger word

did enable shuffle caption tho

[zinc flax]

Maybe you accidentally set the activation tag to that

[junior quest]

but when we enable shuffle caption there isnt any tag for trigger right?

[zinc flax]

activation tag keeps the first tag so it won't be shuffled

[junior quest]

me cant generate expressionless poker faces cuz of this problem

[zinc flax]

Depends on its setting though, you need to set it to 0 so everything will be shuffled

[junior quest]

there we go, config file

[zinc flax]

"keep_tokens": 0

I think everything is shuffled there

[junior quest]

yesh its 0

[zinc flax]

Maybe you need to make a specific trigger word for the lora

[junior quest]

naur me made 5loras prev, those are so fine

[zinc flax]

Assuming you're making a concept lora lol

[junior quest]

style

[zinc flax]

oo, then it's already fine
Thought you were making a concept lora from what you said lol

[junior quest]

yesh me dun like to keep a trigger word

btw whats the new one?

[zinc flax]

Added more tags

[junior quest]

eh? wdym? I mean new one?

[zinc flax]

The v3 one, yep

[steady grail]

they are very colorful

[sly girder]

Ikr? I live for colorful stuff

[unkempt cape]

🤣🤣🤣🤣🤣🤣

so a few months ago i tried (and failed) to copy a style i found on twitter

Apparently i wasnt the only who wanted to copy their style cause i just saw another lora of segawa style LOLO

[junior quest]

https://tenor.com/view/ellen-joe-zenless-zone-zero-shark-girl-my-honest-reaction-ew-react-gif-11429165181898762904

[steady grail]

damn, i really love it

what model is that?

[zinc flax]

NAI

[ivory pebble]

NovelAI

[steady grail]

i love the style

[bitter latch]

Final boss?

[terse ore]

@steady grail the node that allows to keep the same color as the input image in img2img is called Image Color Match.

Or ImageColorMatch+

[steady grail]

oh ty

[terse ore]

I tried it at high denoise and didn't work but it seems to be somewhat accurate at levels I wasn't able to before

[brittle rampart]

It's Great Cthulhu!

[wicked ravine]

hello @white condor , is there any way i can fine tune https://huggingface.co/xinsir/controlnet-union-sdxl-1.0 on my custom data?
Being as a beginner, i need your reviews

[nova remnant]

Fuuuuuunya

Jump

Strawberry jump

[brittle rampart]

Jump Cat!

[brittle rampart]

strawberry-shaped precipitation expected!?

[wicked ravine]

??

[brittle rampart]

@nova remnant strawberry !

[steady grail]

cute

[brittle rampart]

yep! but fingers! but everything was going so well...

[bitter latch]

last one

[brittle rampart]

😉 🤣 🥴

[bitter latch]

haha

[brittle rampart]

It's scary to box with such cuteness

[brittle rampart]

heavyweight boxer! 🤣

[bitter latch]

cyber robotic Eevee schamatics

kinda

[brittle rampart]

I found an enemy

[bitter latch]

hmm looks like rough draft comic/manga

[dull jackal]

Pretty cool looking! Love the pose!

[brittle rampart]

PhaChu?! hmmmm

[zinc flax]

Bee-l

[bitter latch]

derp

[supple raptor]

afaik xinsir hasn't released a training script so, unless you can write your own based on their docs, finetuning it isn't really an option. if you want to learn more about training controlnets in general, then this would be a good place to start https://github.com/lllyasviel/ControlNet/blob/main/docs/train.md

[nova remnant]

Strawberry rain made her clothes strawberry

[nova remnant]

Chocolate bunny I'm hungry

[nova remnant]

BigGrabby

[zinc flax]

massive

[green temple]

nice

[nova remnant]

[sick cedar]

Hello, everyone. I am very new to stable diffusion. I really hope you could help me.
There is an app called AI Morph (the one that was made by Daily Joy studio). (It's not and add, cause this app sucks)It was great, until now. This app lets you transform any picture in anime style. It was the best amount the other, accurate in everything and detailed. Very beautiful. But now it doesn't support NSFW. When i put some NSFW there it says "The result may not meet our guidelines. Please try something different." Can you please tell me, if apps like this are using something like stable diffusion or are they creating their own ai? Cause their pictures was the best in everything. And now I want to understand can I make something like this in Stable Diffusion?

[gloomy scroll]

If you got gaming gpu then you can this kind of stuff on your pc
https://www.youtube.com/watch?v=ac_WQSy2ek8

[sick cedar]

Thanks, but I Tried this, it doesn't work as accurate for me as it is in this video. Plus, I have to describe my picture there which sometimes makes it even worse.

[nova remnant]

[glacial bane]

... Yeah ty

[junior quest]

this is nai too?

[ivory pebble]

Yup

[brittle rampart]

This is a magical strawberry rain!

[nova remnant]

Very cute o:

🐝

[nova remnant]

Bee strawberry

[brittle rampart]

jumping again!? these are bees, strawberries, jumping! 🤣

[nova remnant]

More beeeee

[brittle rampart]

This dress is already much better!

[nova remnant]

They all jumping from the big bad wolf (me)

[brittle rampart]

this is the wrong bee

[nova remnant]

Why wrong

[brittle rampart]

she does not work!

[nova remnant]

[brittle rampart]

but I understood! it's not a bee! It's Bumblebee!

What!?

Crap...

humble-bee

[nova remnant]

Shes doing something

Bringing honey

[brittle rampart]

I think she stole it! look how cunning he looks

[lyric zealot]

[nova remnant]

Big bee

[brittle rampart]

big booo.... bee

[nova remnant]

I approve of boobees

[nova remnant]

Are you doing something different

They look good o:

Like higher resolution

[brittle rampart]

What do you mean?

[nova remnant]

Hmmmm

looks good

Very clean images o:

[nova remnant]

Honey 👍

[bitter latch]

useing wildcards for actress, clothing and background

[brittle rampart]

That's Honey!

[bitter latch]

you mean this?

[brittle rampart]

it's a dark forest. where the evil and terrible wolf lives (@Cat with 4gb vram send help)

[bitter latch]

goes back to bed

[brittle rampart]

have a nice rest

[brittle rampart]

how to make this emotion >:<

[zinc flax]

> <

[brittle rampart]

alas... it doesn't work like this

[brittle rampart]

Dickens!!!! 🥴

[brittle rampart]

I forgot the token for "cat lips" again

[vital raptor]

cat mouth

[zinc flax]

:3

[brittle rampart]

💯

an umbrella in your hand is an impossible task!

[vital raptor]

it's not that bad..

[brittle rampart]

I am not sure.

[vital raptor]

i get a good result every 5-10 images, it's not exactly consistent

[brittle rampart]

essentially two parts of an umbrella. what's on your result and what's on mine.

[vital raptor]

SD isn't very good at continuing the line, but as long as it''s not far off, inpainting could help

[brittle rampart]

oh yeah...

[vital raptor]

so close..

[brittle rampart]

exactly! so close but not quite right!

[brittle rampart]

Well, we can say it turned out almost perfect!

[green temple]

i really liked the style

[green temple]

and this style too

[brittle rampart]

both results came out completely by chance... 🥴

[brittle rampart]

I need to try again...

@green temple this is from the morning one. it really didn't work out very well.

[green temple]

animagine can do it yay

[brittle rampart]

oh.... oh well! 😝

[steady grail]

guys, someone that uses blender? i'm looking for the best version

for what i found online, thee are very discordant opinions

[green temple]

wrong emoji 🌚

[steady grail]

wrong emoji?

[green temple]

nothing

[green temple]

why big

[nova remnant]

but pillows

[green temple]

what`s the haircut called ?

[maiden narwhal]

TWINTAILS,double bun

[bitter latch]

not the leia twintails?

or however it's spelled

brain fried from training class.. .I hate lectures

[terse ore]

I loaded this image for img2img and used the multi area conditioning node to try making an image that is part jungle on one half and a fruit tree garden in the other, with a flat view. However, the image doesn't blend together but instead makes a disjointed image. Anyone know how to do it correctly?

[brittle rampart]

I think the problem is in the node that is responsible for area masks. above the sampler node

[terse ore]

Yes, I correctly divided the area into regions beside each other but I don't know why the image is not cohesive

[brittle rampart]

because, in fact, the division occurred according to a rough mask. what if you try to change the "strength" parameter?

[terse ore]

I changed strength to 0.6 but still the same problem. Completely different images from the line

[brittle rampart]

I think then you need to refer to the manual for this node

@green temple

[nova remnant]

I want the manual for the cute fox girls

[brittle rampart]

I think @Sooma will provide a complete manual!

[green temple]

I can’t get the grabby ears still

[brittle rampart]

ears? which? and for what?

[brittle rampart]

There was honey in the morning. and now the cakes?

[nova remnant]

honey cake

[brittle rampart]

hmm

[brittle rampart]

she looks so sad

[nova remnant]

she'll get fat

cookies

@brittle rampart without the style (the one that makes them have black hair and blue eyes)

@green temple big ears

[brittle rampart]

where are the cookies!?

[nova remnant]

[brittle rampart]

I don’t know where the cream comes from!?!?!

[green temple]

from the cookie

[brittle rampart]

oookay!

[green temple]

[nova remnant]

I want to be the cookie

[green temple]

🤣🤣🤣

[nova remnant]

ear cookie

[brittle rampart]

@green temple

[nova remnant]

eep

[dull jackal]

Those cookies look beyond delicious

[nova remnant]

[nova remnant]

i got xl to work for me

it still takes like 15 minutes per image

but i actually am very patient

i just throw on House and generate on my pc

@lyric zealot i got that ponynai model working, tbh it takes up s much ram it makes typing slow down visibly LOL but it generates images in like 12 minutes, think theres any arguments that might make it a bit faster though

im just not going to switch models because i think thats what takes the most time

[nova remnant]

misewy

on which UI

[nova remnant]

wdym which ui

oh wait

like if its forge or not??

im still using a1111 tbh

ive been sick/in a lot of pain recently

hadnt really been up to things that require thought LOL

[lyric zealot]

let me check my cmd arguments

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=

@REM Uncomment following code to reference an existing A1111 checkout.
set A1111_HOME="D:\Downloads\A1111 Web UI NEW\stable-diffusion-webui"
@REM
@REM set VENV_DIR=%A1111_HOME%/venv
set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^
  --ckpt-dir C:/Users/User/OneDrive/Desktop/Stable_Diffusion_Models ^
  --hypernetwork-dir %A1111_HOME%/models/hypernetworks ^
  --embeddings-dir %A1111_HOME%/embeddings ^
  --lora-dir %A1111_HOME%/models/Lora^
  --vae-dir %A1111_HOME%/models/VAE^
  --realesrgan-models-path %A1111_HOME%/models/RealESRGAN^
  --esrgan-models-path %A1111_HOME%/models/ESRGAN^
  --no-gradio-queue^
  --cuda-stream^
  --pin-shared-memory^
  --cuda-malloc^
  --api^
  --listen

call webui.bat

cuda-stream might be worth it

[nova remnant]

bet

after this image ill try it, see what happens

ill let you know in a bit

also

[nova remnant]

throw my art into your vibe transfer like you do everyone elses LOL

i wanna seeeeee

[lyric zealot]

based on forge github https://github.com/lllyasviel/stable-diffusion-webui-forge

[lyric zealot]

oh wait, you are not talking to me right?

[nova remnant]

I think its to me

lmao

[lyric zealot]

lmao yeah

[nova remnant]

do it

[lyric zealot]

@nova remnant i updated command to main bot

[nova remnant]

i wanna see vibe transfer

[lyric zealot]

you can now view your vibe transfer

[nova remnant]

helllaaaaaa

[nova remnant]

not sure if thats a good thing

[lyric zealot]

[nova remnant]

I mean

[lyric zealot]

just make sure not to use it here, definitely

[nova remnant]

https://tenor.com/view/twho-hours-later-sp-funny-gif-18933136

[lyric zealot]

since i made it an user installed command

[nova remnant]

testing this

[nova remnant]

[lyric zealot]

good luck

[nova remnant]

im waiting on it to actually start up rn

it says theyre unrecognized

hmmm

[lyric zealot]

[nova remnant]

yeah ik

whyyyy meeeee

[lyric zealot]

wait

omg

im stupid

that's for forge

[nova remnant]

LMAO

that might explain it 🤣

[lyric zealot]

i don't think there's much to do on auto1111

except maybe xformers?

[nova remnant]

already have those

[lyric zealot]

yup, then there's nothing much to do

[nova remnant]

maybe its time to switch to forge

[lyric zealot]

maybe multidiffusion?

[nova remnant]

how to i do that easily again

multidiffusion??

[lyric zealot]

"easily"

[nova remnant]

"in the least stressful way possible"

[lyric zealot]

https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111

not sure if this is the corrent extension

[nova remnant]

i need to make a server with just you and cat so i can ask yall questions and be able to find the answers without asking yall again 😭

[lyric zealot]

basically instead of generating one full image in one go, they split it up in chunks, and that makes your vram usage go lower, and maybe in turns, get faster generation speed

[lyric zealot]

[nova remnant]

im trrying to find when yall found me the working version of forge

[lyric zealot]

https://github.com/lllyasviel/stable-diffusion-webui-forge/tree/29be1da7cf2b5dccfc70fbdd33eb35c56a31ffb7

this probably

[nova remnant]

bet

downloading

lets fuck this up

i mean, lets do this

[lyric zealot]

[nova remnant]

this is insane

already having issues and it aint even downloaded yet

okay its downloaded

[lyric zealot]

[nova remnant]

idk something weird happened but imma try running it anyways

what now

i redownloaded it but when im extracting the file from the zip this happens

[nova remnant]

and i think thats whats causing this

[lyric zealot]

that looks like the download didn't complete properly

[nova remnant]

do you have forge??

[lyric zealot]

check if you have the folder called modules

[nova remnant]

can you somehow just send me a zip of it 😭

[lyric zealot]

yeap

i don't think that's the best way tho

[nova remnant]

im not seeing one

[lyric zealot]

how did you download forge btw

did your git clone?

or use the one click package?

[nova remnant]

i just used the link you gave me

hit download

it downloaded

[nova remnant]

looks like this tho

or it just immediately shows it downloaded

[lyric zealot]

okay

try going into your folder of forge

in the path above, type cmd

and enter

[nova remnant]

alright

i did it

[lyric zealot]

wait a minute

that doesnt make sense as well, you should still have modules folder

try this instead

https://github.com/lllyasviel/stable-diffusion-webui-forge/tree/29be1da7cf2b5dccfc70fbdd33eb35c56a31ffb7?tab=readme-ov-file#installing-forge

then there's that one click package

probably easier

[nova remnant]

okay i think i got it

it only took 7 tries for it to download properly

welp

think my internet might just be to bad to load it??

[lyric zealot]

weird that it couldn't install torch

[nova remnant]

is there a way to go around torch install??

like an arguement or smth maybe

[lyric zealot]

you ran update.bat right?

[nova remnant]

no

[nova remnant]

[green temple]

noice !

[brittle rampart]

@green temple did you use the animagineXL V3.1 model?

[green temple]

yes

[brittle rampart]

I'm getting some pretty strange results.

[green temple]

how

[brittle rampart]

if I add watercolor style. I get this

or here

[green temple]

maybe too much prompt weight ?

[brittle rampart]

how much?

[green temple]

i use less weight for every prompt

idk

[brittle rampart]

oki-doki!.. then in a few minutes I’ll try to use this model again.

[green temple]

or maybe not

[nova remnant]

(cute, kawai, fluffy, very cute:3.1)

[brittle rampart]

hmm... it's not her!

[bitter latch]

she's suppose to behind the tree.. tho technically she is.. sorta . sigh

[brittle rampart]

everything in the world is relative! 😁

[bitter latch]

true

hmmm...

my what big ears you have!

[brittle rampart]

she went all out! one. at night. with wine and cookies?

[bitter latch]

huh.. 2d and 3d at the same time?

[brittle rampart]

2.5d 😁

[bitter latch]

Little red riding hood Goldie locks

goes back to bed. thud

[brittle rampart]

have a nice rest

[nova remnant]

dinner with me

cookies dinner

[brittle rampart]

dinner and went to bed... 😉

[brittle rampart]

Why did you get the girl drunk, you scoundrel!?

[nova remnant]

she's not drunk, just a bad photo cx

I want a cookie outfit

[brittle rampart]

draw!

[nova remnant]

I got one

Lmao

[brittle rampart]

This is a visual illusion!

[nova remnant]

🐺

Grabby

[brittle rampart]

here's a little cookie for you

[knotty wasp]

can you guys help me create this type of semi realistic art

[bitter latch]

The cookie chef

[bitter latch]

Miku from the land of cookies

[brittle rampart]

she looks like a rag toy

[bitter latch]

no she's a cookie doll

[brittle rampart]

good good...

[brittle rampart]

her ears will be bitten off in the crowd!

[bitter latch]

strawberry cookie girl

[unkempt cape]

just gonna drop my 2 cents in

so like i remembered playing with NAIv3 and basically the Control Tools were unable to be used

It reminds me of like Pony where you can't use any sort of SDXL related thing with Pony models you gotta use specifically stuff made for Pony, it's so far from SDXL because of its training

So my theory is that NAI trained on the same dataset repeatedly like 10 times or someth

also my phone screen got purpled 😭😭😭

[bitter latch]

hmm wonder whats she's looking at ?

[bitter latch]

thats better

ok enough cyber miku

[nova remnant]

throws your image in my vibe transfer

👍

[lime patio]

Lesbians are obsessed with her, huh?

[nova remnant]

[glacial bane]

I want more Ellen joe

[lime patio]

I'm just telling you, women loved Wriothesley.

[glacial bane]

https://tenor.com/view/happy-happy-happy-happy-happy-cat-happy-dancing-cat-gif-716006509349864265

[formal shoal]

[nova remnant]

no shark girl for you

[glacial bane]

NO I WANT MORE

https://tenor.com/view/ellen-joe-gif-11214938808008893309

[nova remnant]

@glacial bane

@lyric zealot what is thiiissss

im using better wifi now and now its just giving me this stupid ass yellow text no matter how many times i redownload forge

wtf

@nova remnant have some goth shark tail girl to avoid some of the pain, I have no idea what black magic is that

[nova remnant]

shes cute 😭

idk what black magic is happening on my laptop either

this shouldnt be happening

hellllllllllpppp

[lyric zealot]

Jesus my lord

[nova remnant]

helppp meeeee

[lyric zealot]

Might be due to those new update

Maybe running update.bat is a bad thing

[nova remnant]

i havent run the update.bat yet

[lyric zealot]

What did you run then?

[nova remnant]

the normal web-user bat

[lyric zealot]

You are using one click version right?

Then you probably should run the environment.bat then run.bat

Skipping update.bat since you don’t really wanna update to newest version

[nova remnant]

hmmm lets try it

[lyric zealot]

https://tenor.com/view/kazuma-konosuba-lmao-anime-praying-gif-14008536

[glacial bane]

But but

It's not Ellen joeee

But i do appreciate the efforts

[nova remnant]

@lyric zealot @nova remnant 💀

[nova remnant]

I want cookies maid

[nova remnant]

why do all my images come out looking like shit 😭

i use the scores and source_anime

but it comes out looking more cartoonish even with anime style loras

[green temple]

[lyric zealot]

jesus it takes 3 hours?

[nova remnant]

Yeah 😭

I couldn't get forge to work

[bitter latch]

whew today was a long dayyy

[bitter latch]

cyber red wolf power ranger

interesting

tries for a group

hmm...

ah ha!

almost

at least it was consistant that time

hmm cgi-anime?

huh...?

ooh

ummm wtf!?

stares into her eyes

yawns

ok i'm done. to tired

good nite..

thuds

sleeps

[brittle rampart]

the contents must be replaced with ice cream. or crushed ice

[spare relic]

A girl

[brittle rampart]

I am!

[acoustic locust]

Why when I'm on linux PyTorch loads everything to RAM

Including the model

My ram just fills up with usage

[vital raptor]

wrong config settings maybe? or no gpu detected?

it's not because it's linux

[acoustic locust]

it seems to be getting everything just fine

[vital raptor]

what webui are you using?

[bold ether]

https://tenor.com/view/hatsune-miku-gif-26297771

[acoustic locust]

comfyui

[brittle rampart]

Perfect Miku!

[bold ether]

what makes you think that ?

[vital raptor]

not sure how comfyui handles vram settings, but i'd guess it's running out of vram so it's moving things to the ram

[acoustic locust]

There's a process that runs called something like PTorch or PT or something and when I load the model it starts using a shit ton of ram

and then when I gen something it uses more

so it just keeps using more and more

[bold ether]

probably because it loads the model into ram and then pass it to gpu's vram

[acoustic locust]

well it never clears the damn ram and it keeps filling until my pc crashes

[bold ether]

uh oh

[acoustic locust]

It just never clears the ram

[vital raptor]

that's vram, not ram, it's supposed to leave the model there

[acoustic locust]

the other one is ram usage

this here reaches about 16 when in usage

[bold ether]

try asking in #🧣｜comfy-ui

[vital raptor]

I don't have a pt_main_thread with swarmui using a comfy backend, so it must be something in comfy

[acoustic locust]

pt_main_thread should just be the pytorch process

[vital raptor]

best bet is asking in the comfy channel like @bold ether suggested

[lyric zealot]

even if you are genning like low res? 512x512?

[steady grail]

maybe he is genning on ram?

[lyric zealot]

that's super likely

[steady grail]

not to be arrogant, but i'm an expert in genning using ovens lol

[lyric zealot]

[steady grail]

honestly, i would never gen using only ram

it's very painful for so many reasons

[vital raptor]

that reminds me, i need to play with cpu (no gpu) genning on my server, it has 28 cores (56 threads) and 128GB ram

probably still slower than my 4090, but it might be useful for automated upscaling or something

[steady grail]

would be cool to have a parallel machine to do secondary light jobs

[vital raptor]

yeah, I got the server for running a NAS and a few VMs, but I got a pretty good deal on a used server so it's way overkill for what it's actually doing

although for a NAS, the 24 hotswap bays in the front are nice 😄

[steady grail]

cool

[unkempt cape]

may have just found free niji hopefully it works