#🍥|anime
1 messages · Page 243 of 1
Also I can't even find the model for it in a download-able format
:D
it's meh for now, but i'm still hopeful
Pony was wild
yup yup, true
There's just a lack of models and loras
And yeah there's more everyday
But a lot of more niche 1.5 loras that I like aren't being made yet, like the Aida iron style lora

sadly true
there's some weird lora that i have, that nobody makes for SDXL
I'm trying to make some of my own but
I only have so much money to spend on buzz on civit
And I can't find a way to make loras offline
@lyric zealot, do you know if ComfyUI does prompt switching by default? For example, using something like [cat:dog:0.5] to blend the two prompts.
or stableswarm if you like comfyui node based stuff?
it lets you put your own comfyui workflow in behind it?
Been a while since I've posted here, here is my first post in this chat in many, many, many months.
2 generic women in neon city. lol
it doesn't you'd achieve that by just passing the unfinished latent to a new ksampler node and resuming where the first left off with the new prompt attached to the the 2nd ksampler.
here's an example with the shown workflow embedded.
here's the fixed version. i forgor to apply the start_at_step change in the second ksampler (advanced) in the original
could i have the parameters and model for this? i really really like this style
ty for the example but that's very inconvenient 😄
It's easy to set up if you only want to make 1 or 2 prompt edits. But anything more complicated than that is definitely a pain
how do i make images
using the bot or a private gui?
been trying out flat soft 2.5D colors , shading and shadows
using different models
mostly comes out boring tho
but wanted to see what I could do
so i experiment like this
No lie
after so many gens, i really started to appreciate the right colors and light reflections
doesn't it get boring after "so many gens" tho?
a pianist never get bored of playing? a pictor never get bored of painting?
when you feel inspired, you just want to express it
Love the dress! Very nice!
I love how the hair lights up. Super cute clouds/penwork, too
Yeah I took a quick look at it, seems like it would be an ideal choice for me
sadly I think it's only av to a select few on experimental type branch
I choose .. you?
2.618D anime style ... ? 🤨
Grrr where make her wear gloves box
Btw you've been consistent lately i like the gens fr fr
been playing with SD a lot more these few days

Make more boxing gens pwease
THIS ISN'T A GEN IMAGE GET OUTT
@wooden coral
reported&blocked. job done
4LEX!!! MY MAN
sd3 aint so bad
More

That looks worse than NAIV1's first leak 😭
I am bread
did you try sd3 yourself?
Oh mb My BREAD
You've been gen this lately
No lol, I'm waiting for finetunes, hopefully we'll get some
ok, sounds fair
some 1,7 weeks or so before the promised update
Made a couple of those yesterday, been busy so I haven't been prompting much 
What update? Are they fixing the model or something? 
yeah, they said to improve the model
Ill give a cookie if you make 2 more

i dont expect all problems fixed though is such a short time
CHOCO cookies! yeah
Oh that's a cool concept
, gonna make some pics like that later
can't get the incoming_punch to work 
Make it guard
#🍥|anime message```yaml
best quality, masterpiece, score_9, score_8_up, rating_safe, bold line,
1girl, white hair, high ponytail, sidelocks, forehead, red eyes, dragon horns, tail, makeup, halftone, boxing gloves, sweat, short shorts, incoming punch,
(bee (deadflow):0.8), ilya kuvshinov (style), wlop, realistic, emphasis lines,
Negative prompt: bad anatomy, bad hands, error, missing fingers, extra digit, fewer digits, (worst quality, low quality:1.1), jpeg artifacts, signature, watermark, username, white background, simple background, (gpo, aca, aer, api, fla, gcx, hll, hnj, gpc, fii, fey, fbv, evg, iew, ifl, igh, iwj, iwp, ixb, ixe, ixz, jaf, jbm, jfb, jsf, jyk, kmz, ksh, kxg, kzg, lbv, xsl:0.5)
Steps: 24, Sampler: Euler a, CFG scale: 7, Seed: 866856263, Size: 768x1024, Model hash: 89e7c7518c, Model: tPonynai3_v55, VAE hash: 66f78ad136, VAE: sdxl_vae-fp16fix-blessed.safetensors, Denoising strength: 0.4, Clip skip: 2, ControlNet 0: "Module: CLIP-ViT-H (IPAdapter), Model: ip-adapter_sdxl_vit-h [68f9f08d], Weight: 0.15, Resize Mode: Crop and Resize, Processor Res: 0.5, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Hr Option: Both", ControlNet 1: "Module: CLIP-ViT-H (IPAdapter), Model: ip-adapter_sdxl_vit-h [68f9f08d], Weight: 0.15, Resize Mode: Crop and Resize, Processor Res: 0.5, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Hr Option: Both", Hires upscale: 1.5, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, Version: f0.0.17v1.8.0rc-latest-269-gef35383b```
GUARD PUCNH RAHHH
"glove close up" ??
there's the metadata if you wanna mimic it

i think it's more rng 
I usually like to add "motion lines" and "wind" to these types of images that involve movement, makes it feel more alive
@ivory pebble wanna try prompt yuji in the boxing concept 
Stuff like floating clothes or speed lines might also work nice
With or without the boxing gloves
yup, 'emphasis lines' seems to be tdoing the same thing
With bread men
very sus outfit but...
I have no idea what that means, I'm just gonna turn him into a loaf of bread
Make her guard
You said you're BREAD
I said With, Bread
Dummy
That's a good one as well, I keep forgetting about that tag 😭
Too late he is now bread
NUH UH ALAKAZAM ALIKABOO
4LEX SHALL BECOME BREAD TOO HUU
✨
i don't think there's a tag for boxing guard

🪄
might have to use controlnet open pose
Just make mur pls
No but seriously what did you mean by "with bread men"? 😭
I answered your question
With" and i was referring to you
By saying "Bread men"
Because you said you are Bread
So i didn't want to he homophobic so i added BREAD
Because you identified as a Bread
So hread
Bread
Bread
More"
Ah ok I see (I have no clue what you just said) 😭
Ill put this in computer terms
I meant Bread men
Referring to the database code of 4lex
And the assisted object With"
... That sounded dumb im out
I will smooch your floor
BREAD just go prompt yuji in boxing gloves
Idk why but It just feels like it would look goofy on Yuji for some reason 😭
If she was abit more looking old i wouldn't be raising an eyebrow
Maybe cause he's always punching without them, idk
Maybe he punches
Because he does it
Because hes the puncher
Like bro he's the best puncher 🙏🏻💀
probably cos of the reference images that im using
Yeah but that's why it feels kinda weird, since he always does it raw (pause)
Him having gloves feels weird 😭
Understandable
Im calling my friend
9
1
1
(took some trials)
Not bad
It's best when he does it raw
OH NO SHE'S GONNA DO THE PIECE PUNCH 
Oh how oh how i would like to just hug her

niice
yo what model is that? looks good
switched sd3 for sd15
NovelAIv3
oh, its paid right?
Yup
Too sus, send her to me and I'll remove it

haha




can you give me pnginfo 
I made this a while back when I was experimenting with some artist tags so I'm not sure which ones are used here 😭 (cause I deleted the original image)
Unless you're asking about the other prompts
well tbh, any prompts are fine, since i just got NAI sub, and was testing out different ways of prompting
It's a shame cause this looked so nice but now I don't know how to get it 

I recommend going to AIBooru and typing in "NovelAI has:metadata" and looking at some posts there 
There's a lot of interesting artist mixes on there
ahhh interesting, thank you

Just found out that site exists earlier today, so I've been trying to mix and match some artist tags from there 
But a lot of the images there have.. very odd prompts
I'm still confused how the artist tags work honestly, sometimes putting "artist:" before typing the artist's name gives a more accurate style but it also overpowers the other artist tags, so honestly it's best to just play with these prompts until you find something that works
i think it's a problem with SDXL, kinda
so hopefully will be fixed with SD3 finetune and stuff

Hmm, probably
how can i discover what is the model just by its hash?
Can you send me the image with the info
oh thats cute
Does autism pony need a vae or is it baked in?
pretty sure it has it baked in, but it just used the basic sdxl vae
I see. And if I'm using comfyUI, do you know if there's a tool to help find the danbooru tags I need?
i just search for things on the danbooru site, and i don't use comfy so i dunno
probably is
Just realized I had downloaded it a while back. Going to update it real quick
It works yay 😊
I heard that pony models aren't good for inpainting so should I have a different SD model for the auto hand refiner/detailer?
They are good
Probably not for background stuff
Like a landscape
Oh I see
i tried making a lora and i guess civitai's lora trainer is pretty intuitive considering i didn't have a damn clue what i was doing, and it still turned out great
https://civitai.com/models/562796/tibetan-sand-fox-kemono-friends
i made it for a friend since pony couldn't do the character on its own and there weren't any existing loras for her
furry
sea
Imagine a sleek, modern laptop displaying a vibrant and futuristic website interface. The screen showcases innovative design elements like smooth animations, interactive features, and bold typography. In the background, there are creative tools scattered around, symbolizing the process of crafting cutting-edge digital experiences. The scene conveys a sense of forward-thinking design, blending creativity with technology to shape the future of web development.
#🍥|anime Imagine a sleek, modern laptop displaying a vibrant and futuristic website interface. The screen showcases innovative design elements like smooth animations, interactive features, and bold typography. In the background, there are creative tools scattered around, symbolizing the process of crafting cutting-edge digital experiences. The scene conveys a sense of forward-thinking design, blending creativity with technology to shape the future of web development.
nice ! ]


wow !
👍
I mix matched a couple of words on my prompt to see what would the model follow more
2020 anime style magazine cover, animals, British, kids-book, nature, watercolor, whimsy, female-artist, masterpiece, detailed eyes, best quality, very aesthetic, absurdres
that's why the weird words kek
name is funnya 

nice
thank you, went back to some of the older sd15 models
is this SD 1.5 ?
yes
custom merge of some cute anime model and lunar lighting model
oh ok
I think the merge was cuteyukimixAdorable_specialchapter + lunarRadianceMix_luasNight
crepe
Yup
In the not so distant past it was considered an issue that for example green eyes would overflow into other areas of the images. Looking back, colorbleeding almost seems ancient attractive feature. It certainly balances out the color scheme of this series of images. One example to illustrate.
controlnet reference style fidelity : 1
early preview of WD-V looks pretty promising so far.
where do i see that? 
neggles gave it to a few people (including me) it isn't public yet. those 3 i just posted come from it.
yesn't
idk how much i can say about what it actually is. i will say hw requirements are the same as normal sdxl models.
can duplicate images or variations of an image in a training set negatively impact the quality of a style lora?
potentially, duplicate images mean it will train on those specific images more which will make it more likely to overfit on the concepts in those specific images.
not a particularly good example. but it tried.
That is pretty cool tho
i trained a lora based on sivil from macross 7, it turned out pretty good considering my dataset was a whole 35 images, most of them being screenshots from the anime
My man 

upload it to civit 
I will eventually
I need to make preview images and I'm not gonna be home until Monday
did u trained it on pony?what settings did you use
Yes, and I just used the default settings on civitai's lora trainer other than changing clip skip from 1 to 2
oh you used civit lora trainer i thought u were using onetrainer 😔
I trained a Lora for a friend using free buzz and he ended up tipping me 10k buzz for it
The lora trainer on civitai is fast and idiot-proof
And it only costs the equivalent of like 50 cents to train a lora, which I think is worth it considering my GPU would probably take days to produce a mediocre result and would end up costing more in power
yea i guess it works better since u dont have to worry about settings
it's a little funny considering i've seen much more popular models that haven't been tipped this much

What about this trainer ? 
she can train me anytime 
i'd lift those weights all day

#доналдспистолетами
definitely 
"Ahum. Class, pay attention! plz"
cooked fingers 
trips @nova remnant
as stoopid as you 
I call it cutelunar, a merge between some cute sd15 model + lunarRadianceMix_luasNight

oh lunar radiance is very cool model

grabs it
whoa
same seed but with my NOMD anime merge...
grinz and....
huh it will not let me merge it with mine. boo hoo
ah ha, I can't use it in the B merge slot for config but I can use it in the A slot
cackles
and NOMD_CUTELUNAR merge result is.....
huh...
wow.. .I like it!
I like experimenting
😄
same prompt but WTF did this come from? cause I changed the resolution? wow
then again it could have been the sampler euler
yep it was
huh.. that doesn't look right..
then again maybe it does
shrugs
no no the back and front are.. wrong.. wtf
oooh
I LIKE IT!
just cutelunar
huh...?
idk what is this lol
Hi
noice !
looks like a record cover
I was going more with an old game vibe but record cover does track
both is good
accidental heterochromia is the best
noice
i like the art style what is it ?
the promt is kinda 
#🍥|anime message```yaml
best quality, masterpiece, score_9, score_8_up, rating_safe, bold line,
1girl, black hair, ponytail, sidelocks, blunt bangs, red eyes, smile, v, bandaged arm, clothes around waist, halftone background,
(bee (deadflow):0.8), ilya kuvshinov (style), wlop, realistic, (ligne claire:1.1), aogisa, (shion (mirudakemann):1.1), (cogecha:0.9), sho, (solipsist:1.1), (impasto:0.7),
Negative prompt: bad anatomy, bad hands, error, missing fingers, extra digit, fewer digits, (worst quality, low quality:1.1), jpeg artifacts, signature, watermark, username, white background, simple background, (gpo, aca, aer, api, fla, gcx, hll, hnj, gpc, fii, fey, fbv, evg, iew, ifl, igh, iwj, iwp, ixb, ixe, ixz, jaf, jbm, jfb, jsf, jyk, kmz, ksh, kxg, kzg, lbv, xsl:0.5)
Steps: 24, Sampler: Euler a, CFG scale: 7, Seed: 679386028, Size: 768x1024, Model hash: 89e7c7518c, Model: tPonynai3_v55, VAE hash: 66f78ad136, VAE: sdxl_vae-fp16fix-blessed.safetensors, Denoising strength: 0.4, Clip skip: 2, ControlNet 0: "Module: CLIP-ViT-H (IPAdapter), Model: ip-adapter_sdxl_vit-h [68f9f08d], Weight: 0.15, Resize Mode: Crop and Resize, Processor Res: 0.5, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Hr Option: Both", Hires upscale: 1.5, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, Version: f0.0.17v1.8.0rc-latest-269-gef35383b```
this is the artist style right ?
(bee (deadflow):0.8), ilya kuvshinov (style), wlop, realistic, (ligne claire:1.1), aogisa, (shion (mirudakemann):1.1), (cogecha:0.9), sho, (solipsist:1.1), (impasto:0.7),
yup, but tbh, i don't even know if all of them have effect on it or not
np i will try it
ok
ooohhh so you get the same style right ?
similar style, yes
nice
it's kinda similar to nai vibe transfer, but you can't really control the information extracted
np i will just copy this i think ..🌚
good luck 
do you want to copy the style of an auto1111 gui? good luck lol
Wondering what SDXL controlnet to get? I see xinsir controlnets for openpose, canny, etc. but there's also the new all in one Union controlnet.
grabby 
no grabby
-# stooopid
it worked i think 
mememmememem ME
youre alive
always
I'm testing nai stuff o:
Will test goths
whats nai
Paid model/service
oof
i think i know what youre talking about
novel ai??
Basically same thing as pony
Yes
i used to use it when it it first came out
it kinda sucked tbh
@nova remnant
I like everything but the quality
Nai could never get the hires look I want
That dress is so cute though what the fuuucckkk
I have no idea what I'm doing dog meme
@nova remnant 

I'm having trouble with pony/sdxl models
They come out worse in quality than 1.5 models

Send something with PNG info
I check 
bet
Do at least 768*1024 resolution

this is one of the better ones, the rest come out even worse quality, like these
hmmm alright
I think 512 will be buggy
At least before hires
It should look good even without hires at 1k resolutions
try using this as negative prompt bad anatomy, bad hands, error, missing fingers, extra digit, fewer digits, (worst quality, low quality:1.1), jpeg artifacts, signature, watermark, username, white background, simple background, (gpo, aca, aer, api, fla, gcx, hll, hnj, gpc, fii, fey, fbv, evg, iew, ifl, igh, iwj, iwp, ixb, ixe, ixz, jaf, jbm, jfb, jsf, jyk, kmz, ksh, kxg, kzg, lbv, xsl:0.5)

i will use these two tips
and come back with the resukts
results*
And yeah embeds don't work on SDXL from 1.5
and also, this format for prompting: ```best quality, masterpiece, score_9, score_8_up, rating_safe, bold line,
1girl, solo, ...
{artist}```

thought i removed those
oops
itll take a bit but its making a new image
40 minutes oof


hmmm how do i check that
task manager > performance > gpu top right
and also check your Dedicated GPU memory
hi forget, im dad

are you using forge ui?
a1111 yeah
noooooooooooo
forge is said to be better in vram usage when compared to auto1111
yup it's kinda dead
somebody made a fork of forge tho
and im still on the old version of forge, that's working great for my usage
Same UI but it uses comfy memory managing I think
should i try comfyui
is it like a1111 r is it like an addonn for it
comfyui is node based, totally different than auto1111
is it better
what are nodes

It makes everything complicated
You should try to get forge
Didn't you have it before
well, if you wanna learn better, like how everything works in the background, then it's better. But no, you can't just type a prompt, get image
yeah i used to have it
well damn
Did you delete it
you need to connect wires, from this node to that node, then this node to another, then another
it stopped working
You updated it

and when i got around to redownloading it
it wasnt working anymore
so i got a1111
this is kinda like how comfyui works
Forge became some experimental stuff
You'll need an older version to use it
😔
Go pester @lyric zealot for it
Smuggered
and forge will let me use sdxl models??
I think yes 

I am not sure how much you'll be able to hires
man ill have to delete a1111
then redownload forge
Smol has 8gb and can do big resolution
then set it all up again
just install Forge in a separate folder
you probably need this version https://github.com/lllyasviel/stable-diffusion-webui-forge/tree/29be1da7cf2b5dccfc70fbdd33eb35c56a31ffb7
dawg thats for people with space
you just need to put Forge in a separate folder.
and then indicate the paths to the models and Lora that you have in the installed A1111
no

only if you don't have much space on your hard drive. which I doubt
i dnt have jack shit space wise
which is why i said id have to delete a1111 if i get forge
plus resetting up my commandlines is annoying
sure its just editing stuff in notepad but

ive done this so many times
I get you tho
it's kinda a chore
times
yeah 😭
my ai stuff breaks down often

so doing the same shit over and over
is making me wanna

Download more space
the cat* is crazzzyy

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
💀
if you ever need my webui-user.bat
i have my models and checkpoints everywhere LOL
its all strewn across three different drives
well, the best is to put it in one place, preferred SSD for fast loading/unloading of model
not enough space for all my models and loras unfortunately
C:/Users/User/OneDrive/Desktop/Stable_Diffusion_Models
store all models on the system partition?!
i have like 2k loras and 30ish models
plus a bunch of random stuff thats non ai related on my normal drive
for me yes
cos that's the only drive that's fast af
i could partition out my ssd but that's a chore
Do it, delete unused loras and models

but is speed really that important!? except when loading the model.
well, when i was doing a lot of testing
i like to do plot xyz with multiple models
and i was also hosting a discord bot that could gen on my machine
going to name her...... Bea
i
Got it.
now that bot is more dead than alive

okay but does it make stuff faster
could i also do random shit to my shit laptop to make it faster
that would acyually help tbh
i think half the reason my sdxl and pny models take so long is the loading of the models
so if you are okay with just taking a long ass time loading that model once, and then don't switch model
then you are fine
theyre all on an extended drive which doesnt help
i switch my models all the time
and i hate how long it takes to load 😭
yup, then it's definitely best to put those models and lora you usually use on your ssd
for me
it took about 1 minute before changing to ssd, and now it's like 2 seconds
Smol, hype me up for this chore
i have to make new folders now, move stuff, change cmmnd arguements....

@nova remnant gen some waifu on cheerleader outfit for @nova remnant
doing stuff
slowly
@nova remnant clearly I'm focusing on makeup here 
my loras im fine with being on an external drive
they arent 6 gigs so any extra waiting isnt actually that long
my first model folder (xl and pony models) 57 gigs
not too bad
my 1.5 models can stay in hiding for a while
i refuse to delete them though cuz 1.5 will always be supreme in terms of sheer lora amount
cleaning out the trashcan be like
moving stuff takes foreevvverrrr
its only 50 gigs smh laptop
and how do you specify the paths to the models in this case? or do you just copy the models you need onto the SSD for use?
#🍥|anime message```yaml
best quality, masterpiece, score_9, score_8_up, rating_safe, bold line,
1girl, solo, hatsune miku, cheerleader, pom pom (cheerleading), jumping,
(bee (deadflow):0.8), ilya kuvshinov (style), wlop, realistic, (ligne claire:1.1), aogisa, (shion (mirudakemann):1.1), (cogecha:0.9), sho, (solipsist:1.1), (impasto:0.7),
Negative prompt: bad anatomy, bad hands, error, missing fingers, extra digit, fewer digits, (worst quality, low quality:1.1), jpeg artifacts, signature, watermark, username, white background, simple background, (gpo, aca, aer, api, fla, gcx, hll, hnj, gpc, fii, fey, fbv, evg, iew, ifl, igh, iwj, iwp, ixb, ixe, ixz, jaf, jbm, jfb, jsf, jyk, kmz, ksh, kxg, kzg, lbv, xsl:0.5)
Steps: 24, Sampler: Euler a, CFG scale: 7, Seed: 3623574722, Size: 768x1024, Model hash: 89e7c7518c, Model: tPonynai3_v55, VAE hash: 66f78ad136, VAE: sdxl_vae-fp16fix-blessed.safetensors, Denoising strength: 0.4, Clip skip: 2, ControlNet 0: "Module: CLIP-ViT-H (IPAdapter), Model: ip-adapter_sdxl_vit-h [68f9f08d], Weight: 0.15, Resize Mode: Crop and Resize, Processor Res: 0.5, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Hr Option: Both", Hires upscale: 1.5, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, Version: f0.0.17v1.8.0rc-latest-269-gef35383b```
ip-adapter??
for forge, i only load my commonly used models, so yeah
but you can probably do it with some commandline
yup, ip-adapter, with this image
ps that image is not mine ^
Do you think Forge will be able to understand how to work with two paths to the models specified in the commandline?
whats ip adapter
now that i think about it, it might not be possible to look at two different place
unless you symlink it maybe
I don't think this will work.
yup the more i think about it, it's probably not gonna work, unless you change the code slightly
which is gonna be irritating
it's bulti-in for forge under controlnet
Me and the program code are two absolutely incompatible things! 🤣
bet
its so cute 😭
I understood! this is your summoning ritual Vil
if i was the teacher id be the reason she failed

i keep ranting about my divorce and making my students watch the polar express instead of actual classwork
Strawberry ritual
flashback to my school days when that actually happened
in one word - Men!
its happened to everyone
that's it... now wait for donuts and bitten fingers...
Donuts and bit fingers... That's a job for @gloomy scroll

that's what I'm talking about.
okay my fucking XL models are finally moved
checkpoint directory changed
now to test an image
fingers crossed

i hope i didnt do all of that for nothing
that was like 10 minutes just waiting for stuff to move
but I’m not sure that it will be reduced significantly
ITS NOT WORKING
DAMMIT
it says some bs about unrecognized arguments
Big coffee
The tip of her tail is funny
error on the path?
ig so
its not recognizing my commandline that i used to change where the checkpoints are pulled from
i see nothing wrong here
not quite clear
@lyric zealot
model name MODEL.SXL?
its my folder for sdxl models
because ive downloaded a1111 three times by accident dw about it
Was it working with a (2) in it
yes

moved it to a folder outside of the sd folders but still on my main drive
It seems to me that the error is due to the fact that you have now transferred the model that was specified at the last start
lets seeeeeeee
??
this seems to have worked

lets wait for it to actually start up though
jic i fucked something up
i wouldnt be suprised
draw a cat
my sd refuses to allow me to use the original sd model folder looool
@nova remnant - draw a cat for the person!
i had to make a new folder
probably need to wrap the path in either single or double quotes like 'path' or "path" since it has a space in it. i never remember which one is right

ill try both
ill try "
example
it turned out gorgeous!
but the thing is that he doesn't use Forge. and launches A1111
What did you do 
Cthulhu's tentacle popped out of my computer but you deleted it so its gone now
img2img your masterpiece 
Only I am able to generate the most quality images
i need three more but in black and purple
They don't want the color black

:D
LMAO
fucking TELL EM cat
no wait
dont be a furry
💀
why does yours turn out so good
and mine just
looks ass
i did better on 1.5
if you got forge working, then just copy everything the same as mine, and should get the same result
i put in the negs that you told me
im still using the same image for ip-adapter tho
and the score thing up front
eh, still using a1111
i think if you mimic exactly on auto1111, should be same result
alright

#🍥|anime message```yaml
best quality, masterpiece, score_9, score_8_up, rating_safe, bold line, medium breasts, mature female,
1girl, solo, hatsune miku, cheerleader, pom pom (cheerleading), half-closed eyes, stadium, smile, hand up,
(bee (deadflow):0.8), ilya kuvshinov (style), wlop, realistic, (ligne claire:1.1), aogisa, (shion (mirudakemann):1.1), (cogecha:0.9), sho, (solipsist:1.1), (impasto:0.7),
Negative prompt: bad anatomy, bad hands, error, missing fingers, extra digit, fewer digits, (worst quality, low quality:1.1), jpeg artifacts, signature, watermark, username, white background, simple background, (gpo, aca, aer, api, fla, gcx, hll, hnj, gpc, fii, fey, fbv, evg, iew, ifl, igh, iwj, iwp, ixb, ixe, ixz, jaf, jbm, jfb, jsf, jyk, kmz, ksh, kxg, kzg, lbv, xsl:0.5)
Steps: 24, Sampler: Euler a, CFG scale: 7, Seed: 2277079525, Size: 768x1024, Model hash: 89e7c7518c, Model: tPonynai3_v55, VAE hash: 66f78ad136, VAE: sdxl_vae-fp16fix-blessed.safetensors, Denoising strength: 0.4, Clip skip: 2, ControlNet 0: "Module: CLIP-ViT-H (IPAdapter), Model: ip-adapter_sdxl_vit-h [68f9f08d], Weight: 0.15, Resize Mode: Crop and Resize, Processor Res: 0.5, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Hr Option: Both", Hires upscale: 1.5, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, Version: f0.0.17v1.8.0rc-latest-269-gef35383b```

ill try this again tomorrow LOL
thanks 😭
What model you used ?
NovelAI
anime
anime
I like the mage
Im denji ofc I'd become a dog for makima
Lady D.E. or Dark Energy, more powerful than gravity. She expands it all on the big scale, almost invisible universal creation, like an inflow of pure existence.
anime
hi 
Source anime > anime
MORE MUAHAHAHAHA
Grrrr you tooo
It took so long
thanks
DN
DARN
Pc sucks heheheaw
Jk

It's from a bot
It's free but slow because of queue

My PC just blue screens at the sight of AI nowadays
Nice







