#🍥|anime
1 messages · Page 277 of 1
not as a function within the prompt. instead you would use separate CLIP Text Encode nodes for the parts that you want to come before and after the "BREAK" then use the Conditioning (Concat) node to achieve the same effect as BREAK. something like this:
you can chain that concat node into another if you want to emulate using more than 1 "BREAK"
like this
https://github.com/asagi4/comfyui-prompt-control I think to make it faster
seems like the 3rd one has 50% weight and the first 2 have 25/25
might be assumptions on my part
nope, it's concat, not weighted. think of it like (1+2)+3.
math 
oh
the prompt control node failed horribly for me
strangely only in the upscaling step
that bullshit is why I don't use comfy
lmao
Is there way to combine two loras like you can combine models 
do you mean just daisy chain them in comfy or do you want to actually merge them together into one?
merge them into one
https://github.com/ntc-ai/ComfyUI-DARE-LoRA-Merge
I don't know if it still works or not... it's two years old
no idea... I'm only on comfy so I don't know.
Fair enough
I'm running 4 different UIs
I'll see if it still works
Lets hope it's not comfy only because I suck with comfy 🙏
thank you 
thx man I appreciate it ❤️
something something evangelion pelican edition
it was user error 🥲
Haha I just mean the crazy node stacking and splitting you have to do
I'm gonna try out comfy
Fixed
i can definitely see the appeal. Very customizable
is yuri kissing too lewd for this server
it also kinda teaches you more about how the process works
I'm still connecting magic blocks together
but there's a rationale
She big
free gift again
(ps: it's a scam, please don't click it)

Wait... did they get rid of the warning system here?
Used to be you could report via warning symbol 
i guess it's not working properly. If you try right click and using app to report, that also doesnt work
ahhh, the bot is just straight up offline
guess whoever was hosting that bot, maybe left stability?
@dry hatch Would you mind zapping this bot for us? The report option is not working.
Many thanks to you or whomever took care of it!

im gonna take that as a "not allowed"
It's been a long time since I've actively been here. I'd say it's probably fine, though I certainly don't want to see it. :3
this server is dead lol,only bots spam here and kisses and feet get deleted + the whole drama with cw killed it so just post in touhou server 
i already posted it there 

CW?
crystalwizard
tuhou is ugly 
having a go at comfy
wonder why bots are so prevelant lately
I completely missed that. What happened? Powertrip?
i've been using uniscale, but i see a lot of people use animesharp for upscaling
i'm wondering i should switch to using animesharp instead
it really depends on what you like more
yes but it was months ago
zamn
i quit runescape before my 20 year cape
i could probably claim it if i were to hop on
i AM turning 26 soon mind you
same, i did mostly rs3 but i did have like two old school accounts
one normal and one TileMan
mr wolf, where are you?
even though i trained this lora on illustrious it works way better on noobai
it was trained on 3D images and illustrious keeps the style but loses the 3d, while noobai keeps both
Isnt noobai also illustrious?
maybe, idk
Yeah it is lol
another take on wandering aimlessly through gray void
Okay I'm done for now 😄 . That was fun
Okay NOW I'm done.... 😆 night all.
two Jinxes?
is the dpmpp_3m_sde sampler not available on a1111 and reforge?
oh wait maybe the pp stand for ++
i guess it's DPM++ 3M SDE on a1111 and reforge
love the little colored outline
i hope peaceful mode has all the mitas
i want to show mila the butterflies
Is is possible to use flux redux to do an img2img+controlnet with the style change to manga? I want to input an image and use controlnet to keep the composition, then use a manga panel to copy it's manga style onto the output.
I haven't seen flux producing anything similar to manga style. Maybe SDXL models will be a better fit considering their wide style range
I made a 3d blender render and passed it through pony img2img and got this. I just wondered if flux would be better. This one is still pretty good imo
By 'better' you mean more refined? Or you want to add more shadows. Black and white manga has variety of styles
More details like shadow would be good, yeah. The background is not clean either. Maybe upscale would help? 🤔 I'd just want to know how the flux method would compare but pony specializes in anime after all
Illustrious models may be better for anime at this moment. And upscale will help too
img2img keeps color composition even at high denoise. In other words - no extra shadows
How does illustrious/noob prompting work with img2img?
Same as pony: dabooru tags, but no score_ and artists tag are highly recommended especially for noobai
illustrious is more anime than pony
its really dumb in that sense too but it should work for anything anime better than pony like it can't do much other than the tags it was trained on
thanks friend
how does she not trip on that hair?
she's tripping all the time

She's leaning down, so her hair is lower to the ground than it normally is. 
does @nova remnant ever come around here anymore? I'm still using his lightningbloom model sometimes.
Seems kinda cold ngl 
cute rat
It's tomato kind
Is this you rn
Nope. All snow melted here
Here in southeast texas, we're preparing for heavy snow tomorrow. And we're near the coast, lol. I haven't ever seen "heavy" snow, and in over thirty years I can only remember it snowing twice. It's often in the 50s or 60s during winter here, so for it to be snowing AND in the teens is exceedingly rare.
How warm is today for you
Didn't check but probaly around -5c

ultimate spiderman style
like the cartoon
I guess I was close... lol I tried with DCAU style.
that is close ngl LOL
different universe BUT both are superhero themed
you must be part psychic 
I watched... (oh who am I kidding)... still watch, a bunch of cartoons, anime, etc.
I think it's mostly that I am watching the american animated stuff as well that I think people would might pump the brakes on... but shows like "Invincible" might be changing hearts and minds.
The comic is exceptional source material.
fair, no judgement 😄 . We all like what we like. 😁
havent read the invincible comics ngl
loved the show but just havent found an actual copy of the comic yet
id rather read it as an actual comic rather than on my phone 
it messed the hands up 
j/k it just reminded me of this from the IT crowd 😆

i get that a lot
i fixed the hands the best i want to
Which AI tools you use for image generation?
forge
And model?
if you want any of the styles im using
and if you want to use any of the eye loras i use

YOOOO WHITE HAIRED SLEEPY WITCH WITH A GALAXY DRESS AND BIG FLOPPY HAT?
literal peak
it could only get better with thigh highs atp
i've never used wildcards before so i have no idea if this syntax works
it seems to be working
if too spicy, lemme know, i'll remove
Gosh darn it
@wispy canopy can you explain to me what CFG scale and Sumpling steps do again?
My memory is very fuzzy idk how things work
sumpling? sampling?
Yeah mb typi
CFG -> how hard it should adhere to your prompt
the higher it is, the more it'll listen, but also less creativity, and too high and you'll get the cfg burn effect, usually people use dynamic caps for that
sampling steps is just how many steps you want to do to make the image, more is generally better, but slower
Alright tysm for helping me
Oh i almost forgot how many steps should be used like default steps and cfg
for any non turbo model, i'd say 25-30 is a good minimum, below that and artifacting will probably happen, or cfg burn
how some ppl here are
even though its recommended on the doc and civit page? https://d0xb9r3fg5h.feishu.cn/docx/YpOQdtHTDoetcZxIO9fc33onnee
i just wanted to get clarification cus ive been using masterpiece for a while
give all 
?
are those artists ?
oh.. yeah
i want to try them 
Tsutomu Adachi
Izumi Tsubasu
Sakura Asagi
Hirohiko Araki
Yuya Asahina
George Asakura
thanks 
are you sure about this? I thought the noob page specifically called out "masterpiece" as a quality tag, but they omit "amazing quality" and "very awa"
anytime
I still struggle to get clean lines on noob.
I guess that makes me the noob.
the hugging face https://huggingface.co/Laxhar/noobai-XL-Vpred-1.0 has very awa, masterpiece
amazing quality does not exist
so its a bit inconsistent
interesting, I was looking at the civit page
ye
but it looks like very awa and masterpiece should both be trained
not sure about amazing quality
that was more from artiwaifu
seeing as my regular PAG setup that i tuned for illustrious destroys noobai -> i'm just gonna run it again. gonna match ALL fckn blocks with baseline output to see what blocks should get pag, and which should be skipped. should improve quality by a lot
that, is a pag image comparison for each possible layer
so i generate a base image without pag, and then one with
i noticed PAG destoryed the backgrounds in noob-ai so it's cause over-self attention on things that are back-ground-ish
don't think i've ever used PAG
this is on noobai -> the pag destroys the backgroundlayers totally, and causes mild burnin, so i'll have to figure out the offending layers
needs more nodes
very nice
Smol ??
yup 
pose 
White hair always looks good
bloo also 
Smol 
agreed
also bit off topic but theres been such a increase of people asking to be handheld from step 1 to expert lately
I feel like that decreased a bunch
In this server at least
yeah. like i help them with one thing and they instantly dm asking on how to even install something basic
"im bad with searching" no your just lazy and dont want to watch a tutorial
yeah this server has been going downhill for awhile 😦
"How to stable diffusion controlnet"

my recent one was on "how to download stable diffusion"
Kek
How to save picture? /s 😉
"how to smol"
i also love the insane projects people start with no knowledge and expect us to do it for them
like general chat rn
I think thats because the Anti AI haters spread how "easy" it is
you type things like smol and then magic
i already knew that
Could be worse, my requests that i get are useually worse
its why i pick and choose who to help
its unfortunate
but it keeps me sane
leave those niche ones to yall and cs1 lol
Im gonna ask what kind if model photos hes talking about and if its garbo im gonna redirect him
yup , but not good 
is very good but not for this server 
smol is good for this server 
i like the floower 
Would.
do not the... thing
you know they say "don't stick your"
and she has the yandere. don't recommend
That reminds me, i still got a backlog to upload here but i forgot. Tomorrow is early enough im not getting out of bed again
"So i got a 480, and my generations takes 3 hrs. How fix?"
First step is upgrading your pc
"Nurr, no one said i needed to upgrade. AI is stupid anyways".
Its in quotation, so obviously sarcasm xD
still probably better than my pc

I can't complain about my epeen pc
Man the artisan channels always crack me up
They use every command but the right one
Its in the frequently asked questions even
People can't read. You should know this
Oh definitely. I just keep hoping somehow
Firstly, you go onto ChatGPT obviously, and ask it to make you a prompt, so that the SD model can understand it xD
@dusky linden girl sitting in the park sunny the view should give a satisfactory relief of pure truth to the viewer, there are about 5 bushes filled with strangers sitting inside them spying the poor woman in the middle of the field
I think the best ones are like
dog
people just come in and say like that
lmao
You gave me physical recoil
"Gib Doggo"
Like i get it if its a midjourney slop prompt
lion and cat
Im a devout xl user still
Meanwhile Flux wants a thesis on the subject
When people say they're an Xl user... What does that refer to?
While not recommended i still got some ok results using tag based prompts for flux
Kinda been dabbling on my own tbh
Theres sd 1.5, SDXL and SD 3.5 (models)
The SD XL i use is a variant called illustrious
Oh right... xD Brain forgor. I don't use the base models, only stuff trained off em 😄
Ye, Illustrious is pretty dcent
Pony is alright but the prompting in there is awful
"Water"
Straw sippy
Drinking coffee gets better results imo
pony is less bad
I need to get a better pony model
for anime stuff
V7 Is being made iirc
I got two realistic models and 4 illustrious
Two flux and like a dozen 1.5 i never deleted
Just a few.
I have all the models
(I don't gen local)
lmao
what triggers me the most is seeing something cool
This be my Lora folder atm
😄
9 you said?
ugly comfy 
You wouldn't be saying that, for an eye fix like this :3
+1 
abs
ugly
Agreed. Still finding out the sweetspot for this new Lora
@green temple
angry baguette bonk
Quack
..... Only?
comfy can do math
idk im too dumb
make it the opposite of what you are doing 
I set it to -3... And the girl buggered off 😐
she went to abs heaven
I kid u not, the first set of 2 i genned after setting it to 3, was 1: An empty room, 2: Her having her back faced towards me. Like wtf
-3*
Better? 😉
WELL! She buggered off after adjust beyond -2 😦
ive peaked
On Civit?
Is that like
-3 no abs no girl, -2 abs, -1 giga abs, and then positive is more insane things
?
0 is decent
👍
why show smol belly 
It was taken from the lora's page directly 😦
Also wait... Smol?
help 
I mean, no kinkshame and all, but u do u :3
am useless, cant help

I do not have any good models for chibi making

Yes
Oh noes, someone hit both her eyes with a dart
NEURO-SAMA!
fellow swarm 
Got a little bit of a conflict, when Gif searching "neurosama"... 2nd gif
@coarse brook summoned @lyric zealot 
ugly 
smol hater gif 
smol hater is @vestal ferry
Shame Evil Neuro isn't here
not you 
Could buy a £5000 tungsten cube

Its a diglet

https://youtu.be/o-UwaetV5EM?si=SvRIUZPxUF0WpxOf&t=324 Time stamped.
The ultimate recap cartoon of the first season of the Pokemon Anime! The Indigo League! Kanto! Follow Ash as he tries to become a Pokemon master! With friends like Brock and Misty, and his trusty sidekick Pikachu he will discover the world of Pokemon!!! CHARIZARD!!!!
Should we do Johto next?!
If you want to support the channel consider signing ...
generated a bunch of images today and they all sucked
I need to go back to some basics I guess
im wondering if theyre abadoning the server this year, its like progressively going downhill and hard :/
Who?
mods, devs, etc
Oh... Ehm... Possibly, idk?
i believe it
may cosplay 
whats this
a pink frieren version ?
An absolutely unhinged, top of the spectrum Vtuber
vtuber and shorts
2 things I want to remove from the internet

Now now. We need ads down first.
Ad banners can stay, as long as they aren't intrusive or obnoxious.
BY THE WAY! Any Comfy Users who got a flow for making images with 2 characters or more, in an easy(ish) way?
Cause most guides i find, is Auto1111 😦 And my Auto is buggy
That is a fairly good choice to remove too
I think I'm choosing vtubers and ads as the worst
Shorts and ads for me. Cause shorts ARE ads
When I download a xinsir controlnet model, it says diffusion_pytorch_model.safetensors. Can I rename it and it will still work?
Should be able to, ye.
not arcane this time... Mugi-chan in the pond
landscape for a change
So this might be the best chat for this, I'm DM'ing for a DND campaign in this new setting called Obojima. The aesthetics in the book are very unique for a DND setting. They have marketed it as Studio Ghibli meets Breath of The Wild, however i think there's a touch of Adventure time in it. Anywho, I'm trying to get stable diffusion to help me make scenes and NPC's that do Obojima justice. If anyone would like to help me on this venture let me know. I kinda suck at this lol. I'm going to drop a few Screenshots from the book so anyone who cares can have reference.
You need a ghibli and a botw lora
I've only used Illustrious and I worry that the botw models will make things more 3D.
working more on the art style improvements for this workflow as opposed to the ai side of things with griptape but this is where Im at, at the moment.
thats sick
Yeah but if you put the botw lora at a low weight and the ghibli one at a higher you might get the result you want
See that sounds like it makes sense lol however I'm not well versed in the ways of SD
i will learn though so thank you!
God I love those lines. such a cool style and workflow you've put together.
using artists names that use lots of that lineart stuff, and it comes comes together in what we both agree on is a beautiful style
all these hawt women here... making me thirsty xD
😆 This little thing is bringing me so much joy... It needs more character consistency but overall it's not terrible. Plus it's nice having a little ai assistance when I don't know what I want.
Model?
Idk. But sounds cool.
that's the actual model i'm using. It's a extreme retraining on sdxl base (basically, no sdxl left except the architecture itself) and i'm using comfy ui
- download comfyui and install comfyui
- import one of my images (open in browser and save on your pc)
- open comfyui webpage, and drag the image there, it should open the workflow.
- use comfyui manager to install the missing nodes.
- generate images
ntrmix has a lot of styles, it has them because i'm using random artist names
in that case, i'm not tho
i'm using specifically artists that cause lineart
and then it should start at some point
you will not be able to generate at the resolutions i do -> you don't have enough vram
is it low
what gpu do you have
anything below 8gb vram id consider low for sd
oh then youll be more than fine to run sdxl at 1024 without waiting long
Arcane was good though. Shame the game is shit xD
❤️ ComfyUI
Hi
o/
I mean. Could've just asked tbh. It seems the bot is having a hard time
😉
🤔 Haven't played the game
Neither, but their playerbase is more toxic than Chernobyl
don't know. Seems the Bot is still thinking 😁
Yeah
I assumed that much. What confused me was that it wasn't a new user, and there seemed to have been put efford into that user
With some 11 roles or so
Ye. Idk who to @Tag..
to cull them
@wooden dove Pardon the ping frendough. We've got an bit of a can of spam issue :3
OH BOI
Hell id love for this place to be slightly better cleaned up when bots spam, wish they'd open up cleaner applications
Oh to dream
sigh.
Shame people aren't more careful with their accounts, and what they click on
This warms my heart.
Ye. The image reported had some very
prompts in the positive.
Buzz is indeed a nice incentive to taking action and reporting
owlbearmechathing
opposite
Its ugly and cringe
the game is good (sadly its a competitive game, so its automatically shit because of people)
No, but using "Toddler" + "Bikini" in the same prompt didn't look right.
christ
Yarp.
Lmao

I'll give you some booba soon
To clean your eyes
(without abs)
Ruined
Thats it im making mikus
Hatsune?
miku kitsune 🙂
where boba

am doing things for a secret server
booba later
also I don't really have an idea of what gens to do


Promise Booba.Fails to deliver. smh.
You're just like UPS

where smol
i couldnt gen today 
Mmmm do love local webui
my local is smol i cant gen 
Time to invest in a $20k Gpu so u can gen
🥲
The world is burning anyways, so why not have fun? xD
burning ?
World is fucked. is the TL:DR
why ?

look away
no smol
im just
burning
too hot
I have a bottle of water full of ice cubes
not enough 
Do a warcrime ^_^

ugli
"handing out gun to chibi" lora moment
actually ugly 
evil elf @green temple @coarse brook 
She doesn't look taller than 1.5m. Just need to stand on a table, and she can't reach
booba 


He doesn't have the ability to make booba. He lies to us
who ?
Chaos obviously.

Butt
cant booba
🌚

we just do smol here
Ugly my little pony 
you say WHAT? 🙂
@green temple help 
nooo 
this character is a win
oh wow you got her again
is this an actual character from something?
I want to say yes
But like
It's not really
It only resembles her, not much data
In the end it just looks like an average dark elf
Sylvannas windrunner
hmm ok
the better windrunner
alleria windrunner
(although now lyralei windranger cuz copyright?)
I never played wow
Alleria Windrunner was a ranger-captain of Silvermoon who led her rangers under the Alliance of Lordaeron during the Second War. She was also the head scout of the Alliance Expedition to Draenor. Alleria and her lover Turalyon disappeared after the planet was torn apart and were not heard from for many years. They resurfaced during the final war...
sylvanas is a traitor
alleria is pure.
alleria is a whole hero in hearthstone.
hero has 30 hp
a whole deck of effects.
sylvanas does 1 thing, alleria can do anything (within the hunter archetype)
that looks kinda like my character from yesterday.... #🍥|anime message
yeh looks a little bit like alleria's dota form
A hero card or hero skin?
a hero skin.
I don't really play hearthstone
whizbang is #1 hero in game tho
Plus she looks...
cuz whizbang can use all 11 classes
generic
and even break the rules
and add cardsl ike "wish" or duel treasures to ur deck
so whizbang is the best hero in all of hs
he also turns ur whole deck golden
whizbang is the only true hero with infinite potential
but sylvanas has infinite hunter potential.
I pulled it from some random spot .. wasn't even sure what was in it 😄
Ugly
alleria is totally not generic.

actually looks like I did a few denoise levels... not sure which is best...
let's get some more of that dark elf character
i'm trying, but she's hard to get right
you should do a prompt like mine
"1girl, sylvannas windrunner"
:clueless:
a bit on the spicy side, must be the tag cian yo at work
its so hot
@modest fog complicated? perhaps. but that's half the fun!
Idk where else to gen it 
Misty looking different 
I did it. Finally, after endless hours of painstaking experimentation and research, I found an excellent workflow for tuning 3d renders into decent manga styled images. Lo and behold. This is the input image:
This is the output image:
Of course it will not produce a finished professional image in a single pass, but considering this just is a simple img2img which can consistently output something that does 80% of the work, it can easily be made to do much better. I will test how it can do if I add shadows and lighting to the render but I do like this type of simple style. I also see that sometimes it will add color even if prompting not to. And it will make the wall, pants, etc blank at times unless I prompt for it, in which case it will follow the same coherency as the input image. I would like to see what would happen if I change the original render into black, white, and grey colors instead of a colored base render. That should make it even better!
Another trick I like to use is to zoom in on the face in the blender 3d program (at the same angle) and make a render, and then do img2img, then use that img2img overlay on top of the img2img picture of the scene. That will make it follow the same expression and face detail as the render. But I may inpaint the zoom-in because it gives more dramatic emotion than the 3d character maker.
never really messed with this button before
before
after
i like you tiny button
it just runs hires fix on the last image generated
ik
i like it
so easy
so nice
lovely little button 🙏 wish i noticed it sooner
yeah, it was the one thing i missed after moving to swarmui since it didnt have a similar button
so i made one 😄
thats real ngl
i wish i had known about it sooner, i just never bothered with the little buttons at the bottom before
it just never occured to me to click on them
@desert island youll find it under where your images generate in forge
bottom right
I'm in comfy so I don't have that button 😢 ... I think theres a node for it though... lol 😆
thats so tragic but at least theres a node for it-
possibly LOL
its separate
They are all just interfaces... it depends on what you want... if you want to play with virtual building blocks and do crazy things comfyui is for you... if you are new and don't want to spend the time learning stuff start with Auto. Either are good options it's more about what you are looking for and how you want to interact with your models. What kind of experience you want to have.
or if you want both, SwarmUI
i prefer forge personally because it looks like a1111 but it doesnt killl my laptop as hard
a1111/forge is probably the easiest for beginners, comfyui is the hardest for most people
You can make it with either but you'll have a little more control outside of what can be done if you go with comfyUI/swarmUI over forge or Auto...
ah, i made that using a lora version i havent released yet but the current version thats up should still be capable of doing something like this, i can give you the prompt as well if you want but im honestly shit at prompting i feel-
the lora
just a word of warning it's more work but I swear by it. 😄
the prompt
comfy is def more work
i tried to get into it, i use it sometimes but forge is just waaayyy more user friendly imo
comfy is for like
super autistic people 💀
theyre the only ones who can use it right i swear
HEY! I resemble that remark... 😆
i use it too broski 😭
just not nearly as often as forge
i could not tell you honestly
i have 4 different UIs installed though so
it cant be much space
(teasingly under his breath) 😮💨 ... casual.
my autism lies elsewhere 🙏
comfy could not grip me like it did everyone else for i was too dumb to understand

thats so real
check my discord pronouns 💀
theyve been like that FOREVER bahahahahahaha

okay, i think i gave you the wrong link before
idk
but here
this should be the right one
when you type up prompt and hit generate 🙏
youll need a model first if you dont have one already
the model you use would depend on how good your computer is though-
if you have at least 6gb of gpu you can manage?
Just become mentally able next time
👍
i do 
never 🙏
how good is your computer-
whats the gpu broski
Ugly 
you think everything is ugly, your opinion on ugly cannot be trusted for you are biased 🙏
its what i use at least
you can ask around for more models or browse that site for some
id look into better models like sdxl,pony, illustrious and if you can, idk if you can tbh, anything higher
but those three types of checkpoints are a good start
ouch... I think it only has 4gb of Vram? I don't think he can run that model.
does it?
hmm
LOL
double checking
15 minute gens are fine...
i deal with 40 minute gens ngl
sometimes i set up a batch of 4 for like 2 hours of waiting so i just play games while waiting
lol, i get impatient when it's over ~15 seconds
pretty sure it's 4
im too medicated to not be patient
thats so tragic
okay so
you cant run anything good ngl
i dont know if youll be able to run it at all with 4 tbh-
sdxl can run on 4, it's just really slow
sweet
you can use sdxl models it seems
I think you're using system ram to make up the difference though if you run in low vram... how much sys ram does it have?
||abs are still ugly||
runs away
yes, you would be using system ram at that point, which is why it's so slow
here
this model would suit your computer better maybe?
its sdxl
and a popular model
@rancid remnant you know pony and illustrious/noobai are sdxl also
i didnt 🙏
i learn new things everyday
but i was told the model i gave him earlier wouldnt work
so i found a new one
I love it and use it.
they're just so heavily trained that the loras won't work across them (usually)
its a good model, i used it when i used sdxl 1.0 stuff
wdym
my only issue with the newer models is that theyre over trained on women so even prompting men results in women often
like, a lora trained for pony specifically, won't work well on a base sdxl model, or illustrious
male focus does not exist in these newer models istg
oh well ye, i knew that
you have to match the lora and checkpoint, they have to have matching uhh
models ig?
but the underlying architecture is still sdxl, so some things can work
pony loras for pony checkpoints
1.5 loras for 1.5 checkpoints
i dont mix them at all
skulls for the skull throne.
i've seen alot of loras lately that say they work on illustrious and pony, but i haven't tried them so who knows
exactly, bro gets it
i havent either tbh
not any that say they work on both at least
no
wait
i have
this guy jujohotaru has a collection of eye loras
it works on pony and illustrious
i use them all the time
reply Flux is right out for you my friend
big no, flux won't even load
tbh, with a 4gb card, you can try some 1.5 models too, there's some good ones
and they'll be tons faster
I was just typing that
1.5 isnt bad tbh, if you can run better do so, but i loved 1.5
you just have to do more upscaling once you get something you want
it has tons of loras for different stuff, its super fast
Alright too many good chefs in the kitchen... You all are giving great advice so I'm gonna do the homer meme and go back to work. 😄
i remember people trying to sway me to try sdxl and i wouldnt because id have to download different loras 
only reason i ever switched to sdxl was because a1111 broke something in an update and my entire workflow and custom models and loras broke, so since i was re-training things anyways, i switched
otherwise, it was performing just fine
they never did fix that bug as far as i know...
i switched because i tried it out and just decided i was willing to wait in exchange for the quality
im not sure what would be good for making a webnovel cover
Smuug



twins shipper


