#πβai-images
1 messages Β· Page 4 of 1
maybe u shouldn't be running stable diffusion on a gtx 1650
i saw some dude running it in like a 1050
just turn it down to any
ok it generated now but almost killed my pc in the process
i dont know why did it start to happen, i was generating with 500x500 earlier and nothing happened
but my issue's fixed ig
and like literally there was no slowed down performance, no errors
dam
u know what could it be?
dunno
Should've asked them for their settings 
For such an old GPU it's always advised to use ComfyUI as the engine, so something like stability matrix, which uses a SDui/A1111 UI frontend and ComfyUI as the engine is great for beginners.
i still dunno whats a good model for upscaling these types of images
maybe an interpolation would be better than just 1 model
for now the:
003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.pth
003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN-with-dict-keys-params-and-params_ema.pth
4x-AnimeSharp.pth```
work the best on average
and the RealSR are best for generally any image
did you try UltraSharp?
i use PyTorch in the end
PyTorch?
nope
oh ok
pth is pytorch tho
oh so no problem?
i use these
nah
well, u asked for upscaler for those images, and ultrasharp usually wont lose details when i use it
well, compared to other i used
i'll test it and share the results here
np
BRO WTF
the
003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.pth
and
003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN-with-dict-keys-params-and-params_ema.pth
are literally the SAME EXACT IMAGE
ofcourse thanks to this script
literally when comparing the two on a pixel level...
"C:\Program Files\ImageMagick-7.1.1-Q16-HDRI\magick.exe" compare -compose src "E:\Python\ChatGPT\Cryptization\003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.png" "E:\Python\ChatGPT\Cryptization\003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN-with-dict-keys-params-and-params_ema.png" -highlight-color red "E:/Python/ChatGPT/Cryptization/encoded_image.png"```
they are the same exact image with no red pixels 
https://imgur.com/v9q3Lab
so ig one is enough or smth
testing on a cat it looks like this
input being:
also UltraSharp is way less color accurate 
now to do Z-Stack with average color fix on it
how much denoise?
welp, i have no idea how this chainner works
eh stacking and the average color fix doesnt help it much
well try it cause its nice imho
and can get advanced easily
im too lazy now after all this time using a1111
π
it takes like 16 minutes to run
u use blender?
yah
figured
only blender users are that brave to try noodles

ChaiNNer and Blender are two different whole beasts
like both need two different sets of knowledges

u r comfortable using noodles

welp

nvm i better test it on more real life images not just a few images
input:
output with:
003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN-with-dict-keys-params-and-params_ema
i'd say UltraSharp always makes things blurry
my opinion for now is that UltraSharp sucks :3
but maybe a z-stack as always may help it
as always un ironically UltraSharp is with noise and not sharp :3
input:
output with:
003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN-with-dict-keys-params-and-params_ema
in conclusion:
UltraSharp sucks and its not Sharp and blurs eveyrthing
onto Anime
encoder : libebml v1.4.4 + libmatroska v1.7.1
Muxer : CHDBits WEB Group
Duration: 02:01:28.00, start: 0.000000, bitrate: 6849 kb/s```
so like 6GB
okie dokie
i got 9 images from it
Pov unreal engine

AnimeSharp does ruin the image only in parts with so little slight blur
like this computer screen
but in General i'd say AnimeSharp looks far better than original imo
it also acts faithful towards these small details
well, send pth? i dont think i see the things u see but maybe i should try another upscaler
https://upscale.wiki/wiki/Model_Database
lemme find it
this
also its good for Glow or Bloom or whatever:
maybe i should write a rentry for this
Fancy ComfyUI 
I also don't use Ultrasharp.
In the first stage I use fatal anime to add deeper color(contrast I guess) and then end it with Animesharp or realplusergan or what it's called.
For just sharpening (1x) I use artclarity
last person to ask tbh
@opal zephyr @quaint tiger
@strong cosmos Hey are you good at stable diffusion\
nope
im bad at prompting
Doesnt matter
then what
are you good at installing
just use Satblity Matrix :3
issue π
easy
Okay what now
just install? and launch? its simple
There are no instructions in the readme
Uhhh its been ages since i uses pip
Frick
Im working atm
No time to figure out

@strong cosmos which UI do you recommend
@strong cosmos also does this mean I can delete the seperate source code and it's all good
imagine feeling nostalgic to GOOGLE COLAB
Since you said it was standalone
yeah
it installs in its own folder
Ye you can have nostalgia for recent things
using symlinks for A1111
Unclear
lemme show my folder
uhh
Yeah show your folder it'll be easier
It doesn't exist
Ever heard of Symlinks?
β Become a channel member for special emojis, early videos, and more! Check it out here: https://www.youtube.com/ThioJoe/join
Symbolic links aka Symlinks are an incredibly powerful feature. You might not need them every day, but they will almost definitely come in handy at some point so you'll be glad you found out about...
Also, why don't you use automatic1111
But it does
Cause I was dumb
but it's all good now I think
You didn't enter the correct path to it
Yeah use automatic111

I think the issue was I put it in the pictures folder
But the cmd directory was still C: regardless
So idk what even was going on
this is the folder
the A1111 will be in the Packages folder
the models are in that folder
then get symlinked to the packages folder
just use the UI
ignore the folder it self
I heard there was some thing hobqueer was missing
I'll need to go find what it was
but it was dynamic something
The censored files 
eh
you just need a checkpoint
Yeah checkpoint
No not where I get them, where do I use them
oh
in the stable diffusion folder
or its named Checkpoints
omagahd
the sony toilet

@strong cosmos Im guessing bigger file means better quality
whats the difference between pickletensor and safetensor
wel yeah
all my other files have been safetensor though
but do safetensors
since Pikle sucks
Okay
and its literally not safe ironically
Wdym
i assume its pickeled file
and those can have viruses
just like the .pth for RVC
Originally, when SD was released, the file format for models was .ckpt, which could theoretically include "pickles" (malicious python code, malware). To combat this, Safetensors format was developed, which is inherently safe. Both Safetensors and .ckpt files are used in the same way - placed within the models directory and selected from the model list in the interface.
Checkpoint files are sometimes referred to as "pickletensor" now, and the site does scan them for potentially dangerous inclusions; you will see a warning if suspicious code is detected by the scanner.
So, in general, always take the Safetensors version if available, and if not, check the "picklescanner" status of the checkpoint before downloading!```
Ye I just read it
well ye i was correct
:3
Damn
I dont remember if RVC was easier to get into than this or if they were about the same
But tbh rvc had kalomaze's guides
@strong cosmos okay where do I put my lora
in the folder named Lora
Makes sense
@strong cosmos and then how do I tell the checkpoint to use the lora
Do I like type in the lora name or something
just go in the tab named LoRA and choose the lora then it will add a string of text
it was smth like:
name:weight
but i forgor
so like
lora:1
Nice I found it
Bro I need to just start looking
But I swear it's like the folders pop into existence only when you mention them
False hydra shit
:3
Stability matrix really needs more exposure, such a great concept for beginners especially 
seems to be more complex than 1111
looks nice and slick though
A little yes, but you can choose whatever frontend you want, but still get the performance power from ComfyUI which is super helpful for low to midrange hardware 
i think control net also allows low vram
You still need a permission/role for that no?
you do?
Yep, else there would be much more image posts here 
Damn, How do you get it?
Hackercat βοΈ
Be in the community for a while, get known and trusted 
I mean I was apart of the previous server before the incident
What was your prompt for this (if you still have it)
Im bad at prompting right now but I think I just need to see how experienced people do their prompts
u can ask chat gpt for one or go onto civit ai and get inspiration from uploaded ai artwork
they usually have prompts under them
cant be proud that i spent 20 minutes of my time on this

You're just a hater

You know what keep hating
About this image
Prompt: (perfect quality, masterpiece, super detailed and intricate), (epic composition, beautiful cute hacker cat-girl, short hair, sitting at computer desk, cyberpunk), (anime:1.2)
Negative Prompt: Low Resolution, Low quality, Bad CGI, Bad artwork, Poorly drawn, Smudged, Distorted, (Conjoined, Twins, Duplicate, Merged, Overlapped, Fused, Mutated, Dismorphic:1.0),
Created with: Stable Diffusion XL (SDXL)
Seed: 354961
Guidance Scale: 7.5
Dimensions: 1024x1024
Steps: 50
Sampler: Euler Ancestral
Servers: Horde
Created: July 26 at 3:45 PM
π π π
Thank you
#1160633319537520700 channel dedicated to lolipop art fr 

damn this looks amazing
it looks like a fractal
Finished (86.51s) - @shadow bane
I've been here all night (Ariana)
I've been here all day (Nicki Minaj)
And boy, got me walkin' side to side
(Let them hoes know)
I'm talkin' to ya
See you standing over there with your body
Feeling like I wanna rock with your body
Compose a brief, lively scene inspired by Ariana Grande's 'Side to Side.' Infuse it with flirtatious
@shadow bane liked a clip
Finished (75.97s) - @shadow bane
And we don't gotta think 'bout nothin' ('Bout nothin')
I'm comin' at ya
'Cause I know you got a bad reputation
Doesn't matter, 'cause you give me temptation
And we don't gotta think 'bout nothin' ('Bout nothin')
These friends keep talkin' way too much
Say I should give you up
Can't hear them no, 'cause I
I've been here all night
I've been here all day
And boy, got me walkin' side to side
I've been here all night
I've been here all day
And boy, got me walkin' side to side
#πβai-images message (Clip 2)
Compose a brief, lively scene inspired by Ariana Grande's 'Side to Side.' Infuse it with flirtatious
Bruce Lee Boxing??
Cause its supposed to prevent people from publishing copyrighted music audio samples.
Only a certain few that the staff know and certified can get the role of "Trusted Member"
:3
I'm a Mod from AI Hub Brazil lol, I would never do that because I know the importance of this server
but this is AI Hub OG (Rebooted)
There would be no AI Brazil without AI Hub 
This guy 
yes
Remembering that this server has been restored by a Owner from ai hub brazil π€ βοΈ
Then remembering that whenever this server gets inevitably popular again, they go after them lmao
@opal zephyr help 
how can I remove the watermarks/text
Uhh wait show
managed to get rid of em mostly
omg
think im gonna have nightmares now
bean
scared of getting boykissed in the dark hallway? 
no 
that's not canyon east 
yup π
got a question, does the order of loras effect the results?
oof
maybe if you have an object and a style lora, then try putting the object first and then the lora 
otherwise it might just depend on the strength of the loras
Michel Jason
ahh okok gotcha, just wanted to ask lol
never really used sd before
I am not an expert in loras
I only used stable diffusion since 2022 august
when it was only 1.4 with absolutely no fine-tuning
Michel Jason
well the programs all do their own thing, so it might depend on that, also there are trigger words for some that definitely are influenced in where you put them in the prompt
I'm gonna turn my 2 loras around in ComfyUI, since they don't use trigger words the change should be obvious, so give me sec
tiny bit, mostly at the last facedetailer stage it was visible π So yes it does have an influence, but unless they interfere with each other, you probably won't notice it too much
ahh ok, appreciate the help 
I forgot that "old" edited messages get auto deleted 
Yeah if you edit a message after itβs 2 minutes old it gets deleted
Does anyone know if there's a way to make SD work with AMD graphics as of now?
I think there is a way of doing it but it requires you having to run it on Linux
Do you know if it there is a way to do it on windows?
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs
Follow the instructions for windows at the top and you should be good to go
install it normally and add it to webui-user set COMMANDLINE_ARGS=--medvram --backend directml --no-half --precision full --opt-sub-quad-attention --opt-split-attention-v1 --disable-nan-check --theme dark --autolaunch


@simple badger Wrong channel
Blud thinks hes making ai cat girl images
@dire knot has reached level 4.
@valid gale dont use commands here
@brittle ocean has reached level 12.
Mr Bean's lookin worse for wear these days.
well i dunno
Bean Sensei 
What the fuck happened here
I saw smthing about Loli girls lmao
huh what? xD you mean the thread that has lollipops?
besides command spam I didn't notice anything else much 
Ok I'll check logs tgen
indeed, the timing of it is pretty hilarious though with @young spear post 
created like 5 min ago
True
lmao
im tired.
what's up yume? busy day? need some sleep? or just some rest? 
I still can't post in this chat boowomp
i dont feel welcome, i dont feel happy, i dont feel energized. everything just sucks today it seems.
thats all
ima just hop off for today
aye depression days suck, take some time off of online, do some stuff you enjoy and just do your thing, not anybody elses.
you beat that shit off soon 
@shadow bane has reached level 14.
I so want to post some of my stuff here but its so unsettling lol
@uncut slate what do you think? can we get the tT some rank up?
find anything in the logs?
really that bad? if it's non-adult you can just put a triggerwarning and spoilers around it π
ill just run it by a mod just in case haha
good idea 
The tT?
oh I just realised I cant post pics here anymore 
It's all good lol
so many roles... no image perms 
got so used to the helper role having perms anywhere 
@ionic hollow has reached level 18.
wtf happend here
omg lol
well while getting bonked somebody INSIDE her ate her eye... or something like that I guess

When the inpaint fails

Mouth eye ew wth
nah
What's a tt
tridenT π«£
Ok u lost me
Here u go babe
Ahh cheer up. It's gonna be alr. Ur very welcomed by me
Thanks now I can post my boykisser horror arc stuff
@uncut slate

the line on clothes looks sooo good
you didn't need to do so. Thank you! β€οΈ
Get creative ππππππ
Hii does anyone know a new site or app where i can create new AIs image of celebrities?
so you didn't use it yet, just install?
im trying
@shadow bane has reached level 8.
you could use https://github.com/LykosAI/StabilityMatrix
it's an easy installer, also I would advise to use A1111 since it is more popular and therefore more easy to get support for and find online videos/tutorials for it
will this work as root?
that's the whole issue im having
Linux?
if im root, it won't let me run it, if im not root, i can't access files
yeah
is this a ui tool or?
it's an installer for all popular AI tools, you choose which one you want
does it have a ui
yes you choose which one you want
that's an issue
i don't have a ui
im running this on a headless server
geeezus 
well you can try to send me the error in DMs, but I won't promise I can help with terminal Linux problems π
You can try installing base stable-diffusion-webui and run it on the server then launch the link it gives you in the terminal into your browser
@tawny peak has reached level 3.
you just need python3 and git
How do i make the images that look like real places but resemble memes
He means Illusion Diffusion
Ion got the brainpower to teach him rn
So if someone else could that would be a W
Cheese
Discuss in a channel that isnβt here please
Wait thought you were talking about music
For the movie posters use dalle3 or stable diffusion
Thabks
Stable Diffusion is the better one of the two
A lot of times yea
Since openai and bing neutered it
That "censoring" is literally just a coverup for them having shitty bandwidth on their servers
It can and will draw penises just fine 

I'm still not banned for using the jailbreak lmao
And to think I found it entirely by accident
that's actually how many of these kinds of things are found and pandemonium ensues 
I hate this so much omg
true lol
and it looks cool! Nothing bad about some creepy stuff
Rizzar
sad, I cant even get vhs loras to work correctly lol
Proof
This is Bing image creator still btw lol
oooh, dall-e 3 is really good
Hot
indeed 
the sample pic on that article 
iirc google is also busy making a generative image thingy
i can box it up and one pump it, it's all good
no i will do naughty stuff with it
Funny that this is posted on msn

hi
yes assult the thing so that it doesn assult u
:3

what type of jailbreaking?
with dall e 3
making it generate uh, weird 18+
ohh yeah let me in on ur secrets π€π€
I W I L L M O D A B U S E
what?
I'm the one who discovered it 
yeah
lol I love these
:0
I generated the image then made a meme from it lmfao

YES
saend it to verified chat fr
it is super funny
think my paperspace gpu is dying 
naah only 90C
yes lil bro on paperspace
have a ralsei
been having issues with paperspace the past few days tbh lol
ohhh rlly why
notebooks have slow download speeds and sometimes they dont even start
its random tho, guessing its a demand issue
ohh well my download speeds on my rigs are fast but i do get that glitch with the a4000 machines
u just gotta wait a lil
buildings are poopy 
just regenerate a bunch of times and u'll be fine
I would but the machine turned off lol
ok launch a new vm
ill do some more stuff later today, need to go to bed 
yes king self care π
@unreal scarab has reached level 9.
Hot
The universe will end by boykisser maximizer
@pale briar has reached level 1.
more boykisser
I felt daring lmao
satisfying your values through friendship and boykisser
ONG
What is optimal illusion strength
@abstract glacier has reached level 1.
Once you hit level 5 you get access to #β¦βchat
Also that's for you to FAAFO
Furry.
He can't
@wet bramble has reached level 23.
Found myself a style that I like :3
Those make some damn good looking shit when merged
Hot
Pokemon Pixar movie
Rate my ComfyUI workflow
Fuck you lmao
BITCH I GOT A FUMKIN 3060 TI AND A WYZEN 5600X
What awe youw temps undew woad
whart
Cause mine haven't gone past 70
YOUW PC TEMPS
Mine doesn't go past 70 wid Aiw Coowing lmao
I bewieve in de 40's
i think
Don't wemembew dough
oki
I certainly love it.

noice
I stopped doing the horror stuff with this lora lol
i still think they could make the "comfyui" a little more comfy and simple
make it crank 90s on fortnite
wheres my cranking 90s lora 
fuck thats a good question
how the hell do we manage that
it's obv harder to train a lora to follow a certain style
still need to have a look into training sd stuff tbh
i think yume gave me a tut
let me share rq
but it prob might be a lil hard to set up on paperspace
rate my unreal engine workflow
im sure there is a way, just gotta try it out lol
ohh yeah it's possible however the videos show code tailored to windows machines
i think paperspace machines run on linus
linux
Ubuntu 20.04.5 LTS to be exact π€
mhm and there arent any ubuntu type tuts
however it relies on python
and the terminals are based on bin bash
bash for short basically
ill have a look into it laters lol, cba to do it now
ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models
I mean it won't beat HighResFix (cause smaller details still have weird duplications or distortions sometimes), it's still really good for not needing any fine-tuning and is just plug-and-play
On another note: ive noticed people generating shit based on spesific artstyles or generating spesific characters, how exactly is that done??
@zenith spindle has reached level 1.
complete newbie here fyi
either the model understands when you ask for that in the text prompt, sometimes an artists name, or the artstyle name are directly recognized when mentioned,
or you use an image as a base in img2img that gives that whole picture it's style
or you use a Lora/LyCORIS/etc. that can give any model a style you want, for example https://civitai.com/models/122567/papercut-sdxl or https://civitai.com/models/118845/jasmine-art-style-sdxl-10
Same for characters
Prompts to start with : papercut --subject/scene-- Trained using https://github.com/TheLastBen/fast-stable-diffusion SDXL trainer.
Lora
thats something ive seen mentioned a LOT yet can find fuck all on
when looking into it
And from what i understand about it perfect for what im looking for, a model which lets me train it on a spesific artstyle or character
the max i get is "you can use x" or "made using x" but thats about it, nothing about how to actually use it, where to find it or anything
x = Lora/LyCORIS etc.
well yes exactly, @quaint tiger made multiple character Loras for people here and their friend circle, you can see one example here https://discord.com/channels/1159260121998827560/1160735385975009310
and there should be loooots of videos on how to train loras yourself, so that shouldn't be a problem:
https://www.youtube.com/results?search_query=how+to+rain+lora
there should be loooots of videos on how to train loras yourself
Actively searching for it ive found absolutely zero before your specific link xd
ive seen noone covering what loras even are let alone how to train one until now xd
thank youuu~~
since lots of people want to make images with their OCs this has been booming in demand and of course many video makers are trying to fill that demand π
for me its kind of a mix of that and wanting to see whats new in generative image tech
considering the last time ive seen it it was in like
the DallE days
and recently came across some that actually look ok
still not believable as human
@zenith spindle has reached level 2.
but nice regardless
i think this one is pretty sick though
https://twitter.com/uozmmy/media
This account
idk, the rendering looks weird as fuck
I can def see it useful as a art reference
Well if you spent looots of effort you can get it to basically not being identifiable, but yeah the technology is basically in it's baby shoes, and I mean looking at what the recently released DallE-3 can do with recognizing what you want(thx to chatgpt helping with the text prompts), it shows what more is possible
cute
my pfp right now
ah the classic, it's the first one of yours that I saw 
I'll still share it here, but this was so close to perfection
yes...
not bad, some small fixing touches and this would be perfect body wise 
also still messing with my profile, not used discord for a while π
can be edited away!!
i still think this one looks sick
i often find that using ai and then patching it up is better then just raw
then again, only used it for reference so far
well it's a modern artists tool, so many have already adopted it in there workflow, some just use it since they hate backgrounds, others use it as a full base, others use their base image to be filled with random details by AI...
This is why yiff exists
wha

Yeye! i can see generative images used as tools
unfortunutely, most people ive seen just want the death of art
and will die on the hill of "ai is good artists get a job"
there's one i found recently which lets you select a part of a image and move it as if it were 3D!
the ai just autofills what should be there

you would not believe how useful that is for art references
like the recent gaussian split revival? I think that did something similar
I haven't really noticed that much, but yeah there is a mud slinging party as always with these kinds of things, this will all dissipate as the years go by. Nothing will stop AI from entering this space and nothing will kill artists as a whole, just like it always has been
then youve not been on social media as of late
Your sona is honestly very nice.
deliberately so 
I heard enough of both sides having 0 idea what they're talking about, so it's as always just extremists screaming their heart out, they will move nothing with that... so meh
most of the arguments are just the same types of people who used to do nft's and stuff arguing that "artists are useless now" and shit and the other side just not wanting their stuff fed into ai without consent
atleast im using irl images i took !
and like
old ass paintings
like propaganda posters
(i like the style dont judge π )
All my images are for personal usage 
yeah same
I'm not gonna post them on social media and claim that I made them
anyway its like
3am
i should sleep haha
thank you everyone for the tips!!
i will give updates or ask if i encounter problems
eeee
@zenith spindle has reached level 3.
Is it strange that I run my images through a bit of sharpening before uspcaling?
I feel it makes the results of the upscaling better IMO
there is no consent with machine learning, Same as there is no consent of you looking at other peoples images and styles and getting inspired, you might have copyright over your character, but that is only applicable to people using then generated images to use them in copyright protected use cases
if someone in private wants to make your OC in lewd, then they can do so, even if it's mickey mouse.
and cryptobros are as obnoxious as ever, that is nothing new, but some artist bros are not that better and try to argue with facts that don't exist
there is no consent with machine learning, Same as there is no consent of you looking at other peoples images and styles and getting inspired
i completely disagree with that
because to me
My friend is a furry artist, and he hates the usage of AI for generating art 
He doesn't even give any proof as to why he hates it he just says shit like "That's fucked" and nothing else
it only learns from what it sees, taking parts of it and pasting it onto another
there's no transformation
That's when model merging comes into play
well they aren't, factual analysis shows that objectively you cannot argue against this with current law
if you feel different about that, that is your right, and you can try to make it law, but it is not fact and not law
if anything, all ai generated images should be public domain imo
well they aren't, factual analysis shows that objectively you cannot argue against this with current la
because current law has not caught up yet
it can barely keep up with older stuff like the LGBT
so many places still hate them
yeah I think most people have noticed that the last few decades 
at the same time a good and bad example haha
if someone in private wants to make your OC in lewd, then they can do so, even if it's mickey mouse
if someone in private makes lewd content of someone's character or even them irl using ai, i see it no different then something like rape
youre a horrible human
regardless
well lawmakers have 0 incentive to stop ML from looking at what is available, so it's 90% likely nothing will happen, the rest will be lobbying and it's a done deal, AI is too big money already

thats the annoying part
USA is closest to doing anything about it
i know for a fact Korea doesnt care
i just dont want a future where even something as human as art becomes souless and automated
yeah I think Capitalist uberstate US will not stop the new money making machine, even Hollywood and co already announced their interest in the technology
from what ive seen public wise
the outrage is biggest in the west
get used to it and make your own bubble of human content, it will always exist, just as the little egg market around the corner
i dont understand how that doesnt sound dystopian to you
it should be the other way around
when it comes to art
music, writing, drawing etc. should remain purely human
because in a fully autoamted future
what else do we have
nothing
ai could be used, but as a tool
not a replacement
souless corporations do what souless corporations do, even when they had humans doing it it felt souless
because its a serious topic
not suprised about that
it doesn't matter what it sounds like, it was even as a kid obvious to me that it will come.
we can't stop AI entering the industry and ML being the standard, what we can do is fight what it gets used in, but that is a fight actually worth fighting and winnable to some degree.
Hollywood using AI to write scripts and ML being allowed on a big scale, is not winnable. Simple as that. And I sure don't care if Hollywood poops out scripts, that is their thing, I watch movies that I want and if they don't offer that, then they won't get my movie money and others will
im not arguing to keep ML from entering the market, its already been used for a while
im arguing against ML becoming the only factor in it
Hollywood using AI to write scripts and ML being allowed on a big scale
and those scripts end up garbage
then again, this is hollywood
theyve not made good shit for years
because in my eyes machines are physically incapable of coming up with anything creative
Chat is this real?
"im arguing against ML becoming the only factor in it"
the human "slaves" that are tagging million things for ML still to this day, are so far keeping us away from that reality 
only time i will change my mind on that is if we reach a point where machines achive true sentience
very 
@ionic hollow has reached level 19.
That's probably coming soon, just not in an efficent or widely available way
I love the little flesh jesus head on the statue 
with our current speed of advancement im giving it atleast another 1000 years before that happens
π
because we need to figure out how to stop ai hallucinations, how to make computers actually give random valiues without patterns etc.
the Ai Revolution is coming 
oof I don't want to know how many people living in the year 1923 said that π«£
once again i am emphesizing, im talking about TRUE sentience, not something pretending to have it
because we already have that
if this was 2007 i would have given humanity 8000
but again, stuff changed
who knows, maybe in 200 or so years we get a breakthrough
that speeds it up
but at the moment im giving it 1000 for true sentience
or like
old painting
@zenith spindle has reached level 4.
vibe
yeah that too
I'm not making any predictions for even 10 years 
Gn8 
I think jesus is trying to hand me USB sticks... though nowadays I guess it would be more likely vape sticks 
real
litten from pokemon?
I unno
I mean, I can technically use it locally right now on my phone
Used to use it on Colab but Google said f^^^ you no horny
Go for paperspace if ud likr
What is that
It's cloud service
Ok
U can buy a subscription for 8 d a month
Nevermind then
Yeah nowadays since colab just decided to give us the middle finger everything is paid for
Mhm
But if u doing commissions it's worth it
True
Try happy accidents

U can choose whichever model u want and stuff but it only goes up to 1024x1024
Says 200 per month in free tier
We need a Stable Diff bot 
Canβt download all models anyway, most of them (even my civitai models) are locked behind the paid tier
Also holy crap that heated my PC up quickly
Even then the thing Iβm using right now (βDraw Thingsβ on the App Store) kinda works
But
I canβt find where the negative prompt is on it
Actually nvm itβs constantly crashing when trying to load the preview
Just kidding I think I fixed it
Cause apparently the project was broken
Where is negative prom
pt

This
Iβm looking for this in the settings but I donβt see it
It feels like it doesnβt exist
I'm not even gonna bother upscaling this image. Nick wilde with the "My Honest Reaction" lora
normally it should be right next to the normal prompt
i love this lora lmfao
o3o
Oh I had to swipe Ty
where the FUCK is the ai hub chan lora
SD frogs, what's your preferred sampling method?
WHERE IS IT
private
:(
I've recently started using dpm++ 3m SDE GPU with exponential scheduler... gets some pretty good results and supposedly works well with lighting and skin
before that for speed I used dpm++ 2m with karras scheduler
also making images with SD on an Intel Arc GPU takes about 5-12 seconds at 512x512
what model is that?
sdxlNuclearGeneralPurposeV2Semi_v2
downloaded it like 10min ago 
ah nice
I need that Lora
Totally not gonna combine it with Boykisser LoRA
You should
can't wait for that...
@ionic hollow I would really love to see your horror skills combined with this Lora 
I appear to be the BoyKisser master 
Yume has created a monster out of me 
A big gay fluffy white one
Anyways the MHR didn't appear to affect it at all

Wait
It defo did affect it
HOWEVER
It appears to have just made him smile a really big goofy grin
i think im still the master here buds
I've created too many boykissers to count 

We're both Boykisser Masters
How about that
It's funny how I went from this
To this
Hot
I need to do this lol
Gonna combine the Boykisser Lora with and Eldritch Horror model
love the face and pose 
comfyui for paperspace
u may run it but
i am not sure if its okay
cuz they do not allow local tunneling and ngrok tunneling
i havent tried it myself
so proceed with some caution
ill have a look into it, might need to change some stuff and get passed what you mentioned
no it has 3 options for tunneling, local, cloudflare and another one i do not remember
i will reach out to the team to confirm running cloudflare tunnels does not result to a acc termination
this is creepy π
ahhh okok
Or if you want to do it psuedo-local you can buy a Saturn subscription or something similar 
no why pay over-priced companies wtf?
what
paperspace or runpod is the way to go
@ebon dome has reached level 18.
oh this is cool
I said "or something similar"
yeah i see but i just dont recommend saturn cloud
there are some much better options out there
yeah hi litsa
migth as well pay for colab lmao
why were u banned again?
threatened vecken
yeah why did ya do that
got mad
ohh well consider this ur second chance
i threaten ppl when i get mad (all empty threats)
migth needa chill a bit ok?
u post wayy too much so here ya go
we'll never know
usually i just leave it on 7 or 8
I'm also using CFG Tonemapping so I have to increase it
wow this one is quite sharp
like seriously
fingers π
they Blocked Peter Griffin
Microsoft had a good thing until they ended up blowing it up
With their pride and their ego
did you combine it yet?
Didn't do anything good sadly
I'm doing Protogen Boykissers now
F
What
hi
how can i make an image??

