#💬|general-chat
1 messages · Page 139 of 1
It's very simple for basic generations
really? do u know how 2 find it ?
If someone knows of a workflow config file that has this all set up I'd try to pop it open
You can search for extensions if you install something called comfy ui manager, extremely handy piece of software
Looks like this should work
https://github.com/Stability-AI/ComfyUI-SAI_API/blob/main/api_cat_with_workflow.png
Might need to update it to use SD3 but that should be simple
Dope thanks
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOK
Also lets you things like updating all your extensions and the comfyui itself, manually unloading models from memory and downloading some models, especially handy for upscalers since it's usually more annoying to find them
Is it absolutely necessary to have an account to generate images or is there a way to generate images without one?
you can do it without one, you just have to have a local install of stable diffusion, but it wont work with SD3 yet
and im guessing the only way to have a local instal of stable diffusion is to buy it?
nope, it's free
Well ok then, color me shocked. Link to it please?
go into this channel https://discord.com/channels/1002292111942635562/1002602742667280404 and in the pinned messages at the top guides
thanks
It's been youtube. Funny to meet you here right now. Welcome!
didnt realize ppl still used 4chan
I just lurk on it sometimes
Does anyone have any good resources or youtube channels they like on how to edit/touch up images from AI to make them professional looking?
Define professional
Something I could put on a book cover as a self published author
Book cover really doesn't narrow it down tbh
Depending on what it's meant to be the steps would be completely different
Is that some light novel and you want to print anime girls in the cover or will it be some drama so you'll put realistic props in an abstract settings
Anyways
Stuff like that usually requires both an artist and a graphic designer
Making stuff look professional is like a whole job market
So yea I'd probably look into graphic design courses idk
It's going to be a long and steep learning curve. Was moreso just curious if anyone had some go-to these are good places to start. Thank you 🙂
Img2img. Give it your amateur design, and in the prompt add words describing it as a professional work. The result will not be usable, obviously, but the changes it makes will show you where your design falls short.
For img2img, I recommend A1111 or ComfyUI, and SDXL.
Thank you!
Still
Unless you plan out what the cover should be deliberately, it'll likely feel like a random image slapped onto it, no matter how good looking
Especially since AI tends to generate things which look like a photo for the sake of photo
Or an artwork for the sake of artwork
That's why most AI-generated comics very often look like somebody stitched a bunch of similar but independent artworks together
Just keep Photoshop or it's equivalent nearby and you'll be fine.
AI can;t be trusted with everything. Yet.
any way to download all generated images from dreamstudio?
Hello I just joined - what process do I need to go through to enable / dream in one of the channels?
You need to go to sleep.
What do you mean?
Well to enable dreaming you have to go sleep. 😏
il tuck you in
I’ll join you
Love is in the air sing
😎
can i get a tucky
What is a tucky?
Sounds slimy ...
you love it
slimy Würstchen :3
is it possible to create 360 by 180 degrees in SD so it look right when i wrap it around a sphere?
.< So can we make video from prompts or just images? Still confused.
Yes ... Stable Video Diffusion is Image to Video for example
There's a special Webservice for that I've used for free ...
Have a look here -> https://skybox.blockadelabs.com/
Awesome. Can we do so here on discord? Is there a specific room or prompt for it? Or is it solely web based for that aspect?
I think you can use Artisan but you will need to pay ...
thanks thats save me alot of work since i only need to use it ones
I use it local but it needs a good graphic card.
never mind that didnt work its like it's not even consedering my prompt
i will tell you what i found so far
but eathir im dum or they are explaining in stable defution lingo
from what i know i should download this https://github.com/AUTOMATIC1111/stable-diffusion-webui and then have a modle but im a bit camfused how do i make it make a 360 after that
i have 3060 12gb so i think i can run localy
Should run fair on your computer. I know there are ways to create something like that but only tried one or two times ... maybe ask aunt Youtube?
i been their but no luck
@shut hare @finite cloak could the artisan create channel messages maybe be modified to show you how many credits you have left each time you generate?
If you only need once I'd think to install and learn Stable Diffusion is to much work ...
i knoww its too much work for one image but i need a 360 image of a runway in an airport really bad for a project im working on but i cant find any online so i dont have much options left
OK, Blockadelabs only allow download with a subscription ... bad luck ...

Maybe ask here a bit later again. It's just nothing I really need ...
Isn't it free anymore?
Blockadelabs? You can create your thing but with watermark and not download for free any longer
what is blockadelabs?
An A.I. to create 360° panoramas
yo why does it say
ImportError:
DLL load failed while importing bz2
Press any key to continue . . .
Who is it? And try #🤝|tech-support
wym who is it
You said: why does it say
😁
Do you also need help?
Not really 🙂
Never had this problem with run_nvidia_gpu.bat
I think he is using A1111 or SD Forge ...
What a lovely user. I checked. It was his first post. Not even a Hello. Straight to the point.
Hmm
Hi everyone, how to make this kind of video (https://www.youtube.com/watch?v=94ALmuvtBNY)
A1111 has a webui-user.bat, too
anyone can help with 360 image creating
I was hoping to join some channels as I’m starting explore and use Stable Zero123 - is there a channel like that here? I cannot seem to see it?
i subscribed the stable artisan for free trial and no credits are received and no cancellation obtion is availabl
My account is chatgptmaster@naver.com and invoice# is 99EF9080526
Hey can any1 that uses comfy explain me how to make the pipes have those neat 90 degree angles instead of the spaghetti mess? I've seen it on various people's workflows
never forget that SD3 will never released
it will, May 20 
at this rate GTA6 will be out before SD3
i mean that's also a good thing, il take either first :3
im sure it will be fine, people have waited along time, a few more weeks surely wont make much of a difference
gta6 is 5 years out at least. pc wont' come at first
it'll have a couple delays yet
the trailer said did it?
i bet theres some sort of consequences for them if they don't meet that date
they better
no, lol, even if they delay years there wont be consequences
also as is tradition, rockstar games are console exclusives at launch
2025? just in time for ps5 pro or ps6 and switch 2
man who am i kidding, im playing it on PC
but PC will come later 😦
i guess il be playing it on my ps5 until pc
yeah i'll probably buy a ps5 for it
but...can a ps5 even handle it?
ps5 pro
i can only assume its going to really, like hardware heavy
well it will certainly push the limits of the console i guess
im wondering about the fps too
rockstar's game engine is called rage so i think it'll be fine
are they still using Rage?
yeah its just the name of their internal software stock.
they dont use out of house engines
paying royalties on those kind of sales? naw
they have probably worked on that engine for like 20 years
started in 2006 according to wiki
but they use 3rd party physics and character animations tho, so Rage is rendering and other stuff
you dont just give up a name like rage either. it's really the keystone of the whole thing
but truth be told, gta 5 was too much of a success, it was so successful they interntinally withheld gta 6 for so many years
milking the cow :3
indeed, i mean its understandble, but still sucks
mhm
its the same but worse for other games, like minecraft and games like terraria, since they are so successful, they dont even dare to release or update anything
i have the entire assassin creed collection on my ps5, just missing the first one cause Ubisoft never re-released it or remastered it.. smh my head
terraria cancelled their sidequel project to add it as a free update lol
also have the full fromsoft collection on ps5
terraria is especially bad, no idea why, they add like, very very small updates yearly, like adding music or 1-2 weapons
it will come when mincraft 2 comes, so never
😦
but still, small studios are making great games, so its fine
yea im working on some games too, but takes time, i dont want to rush some garbo
it's also a very unique idea that no one did before :3
your a game dev?
yea among other programming things haha
well good luck, we need more games
yea
more simple games, btw !
my game is simple, and highly replayable, and technically even moddable
i'm working on an idle game. not sure if 'm at the point where i'll follow through to completelion or not. still really sketchy
nice
i love idle and incremental games
theyr'e truely one of the most games
im also busy with a 3D engine, i did most of the rendering, 3d physics, and started with animations, but that part sucks ass to implement from scratch
ive had a really hard time with programming
my game combines kinda couple elements, idle, incremental, adventure, etc, its very cool, but i cant share info yet :3
same i'mnot at any kind of info sharing stage. plus i'm distracted by an idle game lately
im also working on my own image generator from scratch kinda, using a new concept that im still working on, but this project will take a while, i have a general idea, but not sure how to structure it exactly yet.
so many projects, so little time 
Hi, i am new, how can i generate images using stable diffusion?
if you don't have much time just hire some ppl
¯_(ツ)_/¯

Hello!
Can anybody tell me what can I do in this server?
Nothing
You can just wait for SD3
What is SD3?
A fairytale
What can you do for me
Love is the answer ... ❤️
hello
Ni hao
hello
you can generate images in exchange for money, check out #artisan-faq
Did they update the version of SD3 that apis have yet
Hey people, I have a noob question : Does anyone has a solution to directly input a PNG image with transparency into inspainting? So the alpha area is targeted by the inpaint.
hey guys, i used to have a website that list every style of clothes, hairstyles, eyes etc but cant remember.. does anyone got it ?
hello guys , who use image to video from stability.ai using API ?
Ahoy ahoy
So, the plan for Stability, is to make money, the same MidJourney does
I'm pretty sure they won't release SD3 or anything else
We can only count on open source and community
No Stability anymore
it is likely to be released, people just dont have the patience to wait for it
they also made the mistake of setting a release date, that passed awhile ago, making people more upset

if ceo says 6 weeks, we can probably assume to double that, so 12 weeks, and we are closing in on week 8, so at the least 1 more month to go
im having fun and playing with so many current things and extensions/plugins that by the time sd3 releases, il prob even forget like, oh yea... this is a thing :3
and im sure people will wait 1-2 months, its not a long time after all
hopefully end of may tbh
yeah thats not a bad guess at all
civitai newsletter said so
i think it will be more 1-2 weeks into june, but well see
Cue the Gaben HL3 jokes.
SD3 = HL3.
It will be worth the weight and it comes but these things take time...
etc
worth the weight? good one
greetings... ai visual production bots have been removed, anyone know?
are you talking to me?
They are back in a different form: #artisan-faq
yo i'm glad A1111 is usable on CPU only
can someone train a pony xl model for me? i have the dataset and everything but my loras look horrible
i have given up completely
Guys when can we see SD3
Now ...
I mean locally
see it with your own eyes? 😮
they will release the weights in the future
Mhh
A figment of your imagination.
you can connect comfyui to the api
hi, SD3 will never released 
.
how much do you wanna bet?
you're both wrong


it's a bad day for men of culture
why not lavibridge on current anime models?
Naw
hey which channel can i get help for pupil problem
the pupil channel
pupil? kek
Is it true that stability is considering releasing uncensored weights for SD3? All that stuff has been done with loras and custom models and to my knowledge, the world hasn't imploded. No one cares.
they don't censor the weights. they currate the dataset and the model learns what it does from that dataset. once the weights are made, they can't surgically go into them and subtract specific knowledge
it's the websites that people use which set up filters, not the model
I am still using sd 1.5 and doing great
and A111 works the best regarding compatibility
all this forge or comfy says error not working after script or plugin install
but A111 no warnings
I think they are concerning about SD3 itself.
pretty sure that the models don't have any filters built in
Make a model in the cloud that trains itself from the whole internet
so no one controls it and everyone uses it
im sure someone with know how cna modify the current approach
write it like a virus
the real Open ai
imagine everyday you use its better coz it trains itself
constantly
accessign GPUS wherer crawling the web... lookig at pics videos
non stop
Imagine someone owned the "internet" back in the day... we wont be here now, some tech is meant to be free
By the time Skynet became self-aware it had spread into millions of computer servers across the planet. Ordinary computers in office buildings, dorm rooms; everywhere. It was software; in cyberspace. There was no system core; it could not be shutdown. The attack began at 6:18 PM, just as he said it would. Judgment Day, the day the human race was almost destroyed by the weapons they'd built to protect themselves.
this crap will never be self aware
(artificial) intelligence
garbge in garbage out
it can become very good at predicting and emulating but never aware of itself
plus a model that would focus on training itself on say images wouldnt concern itself with hacking into nuclear weapons systems
basically what im syaing is a model thats just like these models but it kinda crawls the web for free computational power to train itself on whateve rimage sit can access. hehe
cloud computational stable ai
looool
I thought of that exact quote instantly
Correct, these types of models won't be. But when they finally make an AGI that can bootstrap "awareness," like we supposedly do, it will have wonderfully large datasets prepared to immediately know almost everything 😂
as long as it can render hands and consistent characters we're good...
Electric bicycle charging station, futuristic technology ambiance, stainless steel metal canopy, real-time monitoring, capable of accommodating 20 bicycles for charging, with parking space for 10 electric bicycles, leaving 10 spaces vacant.
hey
Does "poorly made hands" as a negative prompt really do anything? Doesn't it automatically try to make the best hands it can?
it probably doesn't do anything
I think its just a superstition within the Stable Diffusion community
there might be textual embeddings or loras that specifically train on bad hands which you can use in negative value, those might work though
Try to clean hands up in playground v2.5. It does a great job with them.
too much effort,i just gen hundreds of imgs and delete deformities later
I only create Aliens so I can always say: "It's Aliens stuff" 😄
yea just create demons,that way u dont have to fix anything 😈
6 Fingers? Those Aliens are used to it like the Simpsons only have 4 ^^
some loras help a little I believe, but that's about it
you can inpaint to fix fingers if you like the imge
yea if u like the img and wanna do it u can,i just gen hundreds and delete bad ones,takes too much time for me to fix hundreds of bad imgs
Can't ADetailer help with them?
it can try but adetailer for hands sucks,its only good for faces
this. a lot of the times, the same types "changes" to an image, based on adding things to a negative prompt, can be attained by simply adding a blank token. most of the superstitious placebo negative prompt keywords just add noise to the prompt like a seed value. no different than if you clicked new seed on the ksampler.
there is a hand fixing control net that works quite well
Is there anyone expert in LoRA? It's not loading. Need help!
yes
Okay, how can I use stable on Discord
Check out #artisan-faq
Hey i need to composite my video to look futuristic is their a way i can do that...it will be like 2-5min clip
Hello everyone, I have a question about using models in ComfyUI. When I download a model and run it, where does the information about its network structure come from? Specifically, I'm trying to understand how the architecture is defined and loaded along with the model weights. Any guidance or documentation pointers would be greatly appreciated. Thank you!
For training just a face LORA, do I need regularization images? Assume I'm already adding captions like bald, dark skin, etc... to cover the basic features in the photo, but the output I'll eventually generate will be sci fi themed, cyborg body with aforementioned face.
can i pay someone to make me a lora
why would you train a face lora?
Use controlnet with IP-Adapter to swap a face
Unless you need a specific expression, which is hard to get anyway beyond basics like smiling, scowling. Face swap given the expression from the face image
actually, face replace keeps the expression
I can train it for free while im sleeping 😁
dms please
that's not as good as a lora
what kinda lora you need
Lora the explora
Dora the deporta
Bots is working again? Why it says subscription? We need new account to create commands?
Hi everybody 👋 Excited for SD3 🫶
You need money 🙂
if you don't have money for an image generator, just get money with a money generator first 🙂
ATM? 🙂
But it's not backed by gold and there is more debt than money in the world so good luck. 😌
I love that StableDiffusion is less than two years old and we already have superstitions. But some things actually work. Like, if you set the RGB on your GPU to green, that makes it more energy-efficient and "greener", whereas if you set it to red (the color of heat), it increases tiddie size on Anime renders by one-half cup or so.
ATM? You mean...the debt of America to Mao?
Damn you're a pro bro
Ah, thank you. I appreciate it. I've been in the game since 515x512 was considered high-res.
Forgive my ignorance but are those the times of Atari or what? NES?
Commodore-64 ?
lol
I also come from the past
Bro I was rendering on an abacus back in the day. Had lots of compatibility issues when I upgraded my setup from Roman numerals to Vedic numerals (commonly but incorrectly called "Arabic numerals" by retards)
But honestly the upgrade was worth it.
Damn son
I mean honestly, I was late to the game, my homie was rendering back when languages only had words/numbers for "1", "2", and "many". Was really hard to get precision using that.
ok pal, you're smart. I get it.
You have a rat for a profile picture though
chill
I posted one of my early works in general-with-images.
lemme check
that's some good work you have there
Thanks man. The more things change, the more they stay the same.
As they should.
As God intended.
True dat.
As I see it, given the amount of questionable AI-generated content, the mod (Jesus Christ) will return to the group chat (Earth) any day.
"I come not bringing peace, but a ban hammer" - JC
Indeed, in the form of AI
And I shall accept my demise
Because we humans are despicable
Soon enough.
I mean, we are pretty chill most of the time.
Except for some parts
lol suuuuuuuuuuuuuure buddy
suuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuure
I am pretty chill
I'm sorry sir, I didn't know you were the whole of humanity
my deepest apologies
Okay, I am trying to lose some weight, if you insist on making it personal.
watch out for AI though
is out to get ya'
Maybe the Indians are right, we should eat less red meat, would make us Americans less fat at least.
how about we throw hands?
a little bit of fight club
to vent
catharsis
roflmao "That N man"
what in the nine hells is a N------ ? lol
My question was isn't the stable diffusion the same as it was in 2023 (free)? It need payment now?
Bro, I am Daniel "Danny" G. German, Ph.D., and any similarity to a naughty word is purely coincidental, and I would be discheesed if you were to mention this to a moderator.
Look buddy, the stronger they are, the better, I ain't a p u sy
The software is free, but you need decent hardware to run it.
A good beating would serve me well
Don't be so hard on yourself, man. I am sure we're all gonna make it and that we're respectable, successful gentlemen.
😦
Don't make me cry...
😭
Right in the goddamn feels
Really? Right in front of my feels? 😠
Last year from discord I was able to generate image via bots. Now I can't. has that changed now?
Times change, buddy
Yes - now you have topay to use the bot on this server
yes, one hour with the admin of the server
there are still shedloads of free services available, although they appear to be dying off by the week...
I don't know what API these bots used, but generating images is expensive. The move to a non-free model isn't limited to StableDiffusion; MidJourney also used to allow a certain number of free generations, but that's no longer the case.
We can say that SD3 is "out of date"
what do you mean?
it's late..
exactly
Wasn't that a quote from Interstellar? "One hour with an admin here is 7 years of Earth time" or something?
like it's not getting out yet?
no it was just a joke lol
There has been extreme drama with the ex-CEO
(of StableDiffusion)
and maybe never will
Ah that's extremely unfortunate
Damn that's bad news
what happened?
what's the pusy drama now?
fcking humans
they have still said weights are coming.. although the original timeframe has been and gone
can't never agree on shit
Ok, that's what I wanted to know because I didn't understand from the description that subscription was needed now. Thanks.
its mostly capitalism tbh - turns out people that invest money in this shit want their money back faster than Stability was going to repay it - mostly cause they were losing a shitload
Damn, crazy
Yeah, he was one of the few people in the space who actually supported open source (as opposed to Sam Altman, who puts "Open" in the name of his company but makes it closed-source, classic move). But, inference and training are expensive, and investors want their money back (and then some). If you're a CEO and you're not generating those returns because you refuse to make something closed-source, you'll be replaced with someone who will.
Sadly, huh?
Unfortunate 😦
so what now?
we have to pay for SD3 or what?
billionares want more coke I guess?
yeah they had made $11 mill, and had a $88 mill bill for the gpu compute
more prostitutes and shit?
can't they get appeased for once in their goddamn life?
fuck knows
for sure they do
It's actually a typo. You don't need a SUBscription, but a PERscription. Go see your doctor and tell him you need big-tiddie anime waifus by the petabyte, STAT. He should write you an Rx. Without this, it's illegal to use AI.
yeah, bout time we made like the 18th century french
I concur.
About time they get the 18th century treatment
thats prescription - i.e. something that is prescribed
Oh no
I am dumb
😭
don't be so hard on yourself buddy
we're all dumb from time to time
🫂
<з
I am actually not a fan of Анимешников but the open-source training of WaifuDiffusion was at least impressive from a cloud-computing perspective or whatever.
?
But honestly I don't think that crowdsourcing the training of models will work well from a technical level. They're reaching the point where you need 100-200GB of VRAM to train them, and that's well outside the range of anything consumer-oriented.
Just print out what I said and take it to your doctor.
so the "open-source" won't really be open-source soon
What are you talking about?
Honestly, though, I think they will be "open sourced" via inside jobs eventually. I don't work at a tech company developing these models (do work in tech though), and I tried shoving an 8TB M.2 SSD up my ass, not comfy but doable, so there will be ways for people to smuggle weights & biases (models, essentially) out of the grasp of these tech CEOs.
Just trying to help you get access to AI image generation; if you're not interested, that is fine.
We need to figure out a way to use less VRAM
imagine what will happen to that "leaker" after he'll get caught
we can't just keep expanding hardware capacity ad infinitum
there's always a limit
he will be "leaked"
i meant "leaked" but i guess we're not ready for that conversation
Bigger and better models means more VRAM, all else equal. Scale really does matter. There are probably efficiency improvements to be made, but models like SORA are never going to run on an iPhone.
Iphone 🤣
bro Iphone can't run even its own OS
Do they do steganography? In models? This is a serious question.
it will never be released
not really, if one hardware is limited, then you multiply the quantity of that hardware for n times
🙂
like more hardware with the same specs in racks
like servers' GPUs
True. "Delays" in releasing an open-source model are often like "delays" in being able to withdraw crypto from a failing exchange. The company will make more and more excuses until everyone realizes they're never going to do what they promised. Many such cases.
Bro do you have kids?
Never wondered about this
lol it's easy to say "n numbers" in math
but real life is APPLIED maths buddy
Gay. But my point is, not all workloads divide up well. Sure, if one American can eat ten burgers in one hour, then ten Americans can eat one hundred burgers in one hour. But, the fact that one woman can bear one child in 9 months, doesn't imply that 9 women can collectively cooperate and bear a child in 1 month. It doesn't work like that.
I lost a small amount of money in the Qiwi. Their license was simply revoked at one point and all the money disappeared. 💀
I yeah, in theory, things are infinite
in practice, the only infinite thing, is the void in my bank account
Jensen will find a way (Nvidia's CEO)
Yeah that guy is nuts
i hate the h0m0
so don't say that to me
y u so gay?
you rat then
You don't want to discheese me, trust me. You don't want to see what I'm like when I've been discheesed.
with compute power, it does work like that
10 A100s GPUs can be used to train one model faster
my fridge can be used to train one model faster
no but seriously
those Blackwell chips 🔥
I can fit so many waifus in that thing
The Indian guy understands what I mean. There are many processes that don't scale well (or at all) in parallell. This is why a 64-core CPU isn't going to improve performance in most types of applications; if one computation depends on the outcome of another, they need to be processed sequentially, which negates the benefits of parallellism.
fuck the meaning of the universe
lol
just kidding hehe
I heard it's true, "once you go Black(well), you never go back"
I'm all for the meaning of the universe
That sounds about right
"indian" :))))))
who's that
Uh, could it be someone named "Akit Shudar"? Dot, not feather.
He is smart.
Like, you can tell when someone has solved computing problems not just "in theory" but in practice. In practice, parallellism is often a nightmare when it can be implemented at all.
Indian, lol, yeah I've heard that many times
I think Shudar is an Indian name
I actually made the name from two words "Akitsushima" which is the name of Japan in the past and "Rakmodar" which is the name of an Orc Shaman I used to play in WoW
Ah, got it.
But anyway
Do they do steganography on models? This is an exceedingly serious question.
Hmm... funny that you mention it
Steganography is the practice of concealing information within another message or physical object to avoid detection. Steganography can be used to hide virtually any type of digital content, including text, image, video, or audio content. That hidden data is then extracted at its destination.
I've been thinking there are hidden messages in the Universe
like the supposedly random digits of Pi and other constants
also radiation, like the CMB or gravitational waves
we will use AI to decode it
and find out who are they and what are they trying to communicate
Hopefully we can decode some useful messages
There are no accidents...
Master Oogway said it best
What language do you guys speak? 
I am talking about models. Often, with secret documents (like diplomatic communications), every variant given out will be slightly different (like, one word replaced with a close synonym out of a ten pages), and since every variant is different, if someone leaks it, they can determine who leaked that unique version and [REDACTED] them.
English motherfucker, do you speak it?
no, I don't speak English
🤣
Yes, it seems like it
This is not a joke, they really do this.
I know, I'm laughing at the way you said [REDACTED]
that's pretty smart yeah
but, they do this already, right?
Yeah, with documents, and video and stuff (tiny, unique, not-easily-visible markings), and of course all printers sold print unique identification codes in light yellow that isn't easily noticible, but I don't know if this is done with AI models.
Maybe before they let me work on a model, the program it to react to a prompt that is a unique password (some random string of letters), and it rats me out:
"AI, please respond to this question: dhjwek4hdskcnw93uj54qq?"
"CODE ACTIVATED: It was him! It was Dani G. German who leaked me! Go out and send him to Guantanamo Bay!"
And maybe every programmer gets a version that is programmed to rat him out when prompted with some secret prompt (which is known only to the CEO or something).

You may think I'm a paranoid schizo, but better paranoid than in Guantanamo Bay.
In fact, Sam Altman could be in this very chat. He could be any one of us. He could be you. He could be me.
I have a gad and I’m happy that I’m not the only candidate for being admitted to a mental hospital

Isn't that like Maslow's Hierarchy of Needs?
Or something
I am not a psych major, thank the Lord.
Gay. It was made by the...((Wachowski))...brothers? Sisters? Also, the "red pill" is estrogen, dead serious homie, there are microplastics in that shit.
Also, imagine living in the sewers, imagine the smell (unironically).
You may not be able to dive into The Matrix, but you can for sure dive into Idiocracy.
An incapacitated pensioner who shit in a bucket lived below me for a long time, so I already managed to live in a sewer.

Ah
It wasn't me, just saying.
In any case, now I'm happy that SD3 will never released. I switched to text models and the Llama3 modifications are already at an amazing level, at least there is progress somewhere. (Before Llama3, everything I launched was not much better than the models from 4 years ago)
is there a specific workflow you recommend to check out? and does it work with auto1111?
I have never really tried text models. I like to write and don't like to draw so I suppose I am "naturally" drawn to models that supplement my weakness and/or laziness; how are they? What do you use them for?
I'm talking about text-to-text models, not text-to-image. I was very pleasantly surprised by this model - https://huggingface.co/Lewdiculous/Average_Normie_l3_v1_8B-GGUF-IQ-Imatrix All settings are in the model description. (kobold+silly tavern)
oh, like the text assistants, so you can ask questions, etc, I thought you somehow integrating text with prompting for images
Ask it to draw you a waifu in ASCII, it probably can render you dat ass(cii).
Wait a second, are you an ambassador for the Pine Phone? Are you a real ambassador? Do you have diplomatic immunity? If so, please prove it by saying the N-word.
no relation to the phone, or politics
I'm just interested in seeing what modern technology can do. I don't use anything, I just see what it can do and have a little fun with it.
But are you actually an ambassador? Do you have diplomatic immunity (i.e., immunity from legal consequences for any actions committed in the nation of your diplomatic mission)?
if I was, I wouldnt make that my username
"Average normie" - bro, I talk to those all day
I just liked the sound of it, and I happen to live in a pine forest, and at one point I was scripting in a language called pine
In any case, if I had money, I would already be creating a game based on SD + Llama3. Modern power and generation quality are already enough for the first good game on ai.
Nice, I enjoy the smell of pine trees.
What kind of game?
Were any of your ancestors involved in the riots?
(I mean you, pine ambassador)
what riots? what generation are we talking here
The latest interesting technology related to images is this - https://github.com/nullquant/ComfyUI-BrushNet
The Pine Tree Riots
never heard of it, googling
It was pretty hyphy, as the kids say.
I bet some peeps said "tally ho!" unironically during the riots.
Generation of text and visual parts by SD+llama3 under user control, while the context and basic mechanics are in no way related to AI.
sounds really cool
What framework (the equivalent of Automatic1111) do you use to run text models? I'm totally ignorant of this stuff so don't even know the basics.
usually all people use 2 applications: Start the kobold, select a model, select the number of layers loaded into video memory and the number of cores. (work simultaneously on CPU+GPU), then connect the API to SillyTavern.
Kobold? Is this a D&D reference?
It's a WoW reference
Isn't that a game that requires the Internet or something?
Yes, World of Warcraft
Lord of MMORPGs
I checked text models out a while back, using oobabooga, it was interesting to check out, but it didnt really give me anything I needed in my daily life
Ah yeah, I remember trying that too.
Is there some rule that every product in tech has to have a weird-ass name? Like "I used to be big into WaifuDiffusion but then I switched to OobaBooga, and put the prompts from there into 11Labs (no relation to Automatic1111) or CrAIyon"
This app always lags behind SillyTavern. That's why I don't use it, even though I like it better.
Is SillyTavern a "full" framework (you just need it plus a model)?
is a visual add-on to the Kobold, with better customization of everything
Does it still require Kobold?
Also, I still don't know what a Kobold is, in this context.
yes
So, uh...what is that?
is this the kobold you're referring to? https://github.com/KoboldAI/KoboldAI-Client ?
framework who runs gguf models better than anyone else
yea
So, now I have to ask, what is GGUF?
GGML (GPT-Generated Model Language): Developed by Georgi Gerganov, GGML is a tensor library designed for machine learning, facilitating large models and high performance on various hardware, including Apple Silicon.
oh, no.
GGUF (GPT-Generated Unified Format), introduced as a successor to GGML (GPT-Generated Model Language), was released on the 21st of August, 2023. This format represents a significant step forward in the field of language model file formats, facilitating enhanced storage and processing of large language models like GPT.
Okay, so it's a model container
Ah cool
I have an RTX 3090, should I be able to run most of what's out there?
Do you need an A80 or whatever like Martin Shkirelli (sp?), the Pharma Bro?
to run serious models you need more
for the one I gave, almost anything will do
my admiration lies precisely in the fact that such a small model works so well
Like what? I posted a video in general-with-images, is that you?
I have 128GB of RAM if that matters
I have a 4060 and I can barely run stuff
nah just kidding
but i mean
the 80B stuff
that shit is heavy
Llama3:80b
Is this the best currently-available public model? By general consensus, I mean.
104B billion parameter model , god damn
I don’t know which of these is correct, I also don’t know much about text models
but this is the best model now
but the problem is that it is huge
It can be, but the one Jupiter posted is insane
and it will work slowly
you think you have a lot of RAM until you try one of these bad boys
yah at probably like 3x the rate I can run it
I notice that the HuggingFace page at this link shows comparisons to other models...but not to Llama3:80B, is that omission intentional?
But, you know, sometimes more parameters don't correlate to smarter
this model came out much earlier than llama3
I don't think so.. They host that model as well (llama3:80b)
So uh, what do you actually DO when you have a model running locally? Do you just talk to it like..."Hey computer, what's up?" / "Not much, nice CPU temp, it's 75F up in this case, living like Hawaii" / "Ah that's cool" / "Yeah"
A natural thing is to talk about the weather
Uhm...
You haven't talked much to an AI yet huh?
Like, I never thought about what I would talk to a computer about.
at some point some new model will come out that beats the 4090, then the 3090 will drop enough in price that I wont have buyers remorse
No, I have not.
The thing is powerful as fuck
specially if you're a programmer
you can ask it to help you program anything you could ever dream of
it's basically GIGO
you know that?
GIGO
Garbage-In-Garbage-Out
What? Are you gonna try to sell me car insurance?
Oh
you put garbage in the AI , you're gonna get garbage
nevermind
but if you're a smart deep individual
I swear that thing is the best
better than humans
10/10
I want to get garbage, so that' sgood
I guess, to me, the idea of a computer being "intelligent" seems kinda silly.
Yeah I like trash food, so that's great lol
Why would it be silly?
-1 social points for you, sir
well I guess you have to define what intelligent is, can it beat you in chess? yup, check, can it make decisions quicker? yup, by like an exponential amount, can it remember more? yah, like the library of congress, so...what is your definition?
His definition is being envious of the Machine
The Machine shall dominate
and you shall perish, puny humans
Glory to the Omnisiah
I mean, when I grew up, I was probably in the top 1% of "dudes who are too fascinated by technology and spend their time in high school being huge computer nerds rather than getting laid" (hopefully this isn't something hard for you to believe, since probably there are lots of people like me around). And for me, for "us", even those who were in the very top percentile of being huge nerds with unhealthy attachments to computers, we still understood that a computer is just a glorified calculator - that computers are not intelligent or creative, any more than an abacus is (which isn't to say that they aren't useful...after all, an abacus is useful, and so is a calculator).
And it seems to me that the worldwide level of "unhealthy fascination with computers" has risen so quickly that, instead of being in the top 1% of computer-philes, I'm in the top 1% of computer-skeptics.
Can it say the N-word?
" we still understood that a computer is just a glorified calculator " then you understood nothing , Jon Snow
literally: NOTHING
Talk about your loyalty to the machines
you seem hung up on this concept, would you like to have like a group therapy session or somethinrg?
yes
Not cool ; (
Why? don't you respect prostitutes? I do
Not cool to NOT respect them
I would say
Maybe I need one. It's just that my experience on Discord has been:
(1) I join a chat.
(2) I act all nerdy, since I'm a nerd.
(3) Someone accuses me of being ChatGPT.
(4) I reply "Oh yeah? Would ChatGPT be able to call you a N-----???"
(5) They report me to the mods.
(6) The mods ban me
Then back to step 1.
This has happened probably a dozen times.
Wow seems like an amazing life
lol
And I've only been on Discord like a month
But, like, actually, how do you talk to a computer?
The same way you talk to a human
or, with programming languages
or, with machine code
or, by sticking a chip on your brain
want more options?
Because when I'm talking to strangers, like, I ask them things like what they're up to (but a computer just sits there), what they do for a living (but computers don't work, they just mooch off our electricity), so there's sort of nothing I could talk to a computer about, it seems.
what kind of fucking nerd you are?
you can talk to him by voice and it will answer by voice
there are such add-ons for text models
you're the least nerd i've ever met in my life
you're asking how to talk to a computer?
What year is this? 1920?
Is Frank Sinatra still alive? Cos Id like to watch his show
I mean...I often talk USING a computer (like I am doing now), but not TO a computer.
that's just you buddy, the future is here
check out "Artificial Intelligence" , it's a neat little concept
One time I took a girl home and her phone woke us up in the morning with these chimes and stuff, it was a weekend, we didn't have to wake up, so (since women all have iPhones) I said "Siri, turn off the alarm!" but it didn't work, she said it doesn't work like that.
So I would count that as "talking to a computer" since a phone is a type of a computer.
And I had heard that Apple came out with Siri, and of course all women love only iPhones so I thought I was addressing the phone-computer properly.
But it didn't work.
lol look at these humans go
amazing
Also the phone was shaking and such.
Like, I talk ON a phone, but not TO a phone. I write letters ON paper, but never like..."Hello, piece of paper, I am writing on you, hope that's cool with you, uh, how are you doing?" - seems weird.
I thought you were smart buddy...
I'm dissapointed
talking to paper can be useful
Especially if you're a rapper, laying down some hot rhymes with a cheap BIC pen while riding the bus down 8 mile to your job at a Detroit auto factory.
I hope in my next life I'll be a rapper from Detroit
being born in the richest country in the world must be interesting
You can do it in your current life. Move to Detroit. Spit like..."u know my gun's unregistered like Hypercam 2".
Switzerland?
There's no way I can get to Detroit
America as a country is rich, but this shouldn't be interpreted to mean that Americans as people are rich.
Any American is 10 times richer than people from my country

Saying "America is rich, so Americans are rich" is as wrong as saying "Samoa is small, so Samoans are small". Actually they're huge, I believe the heaviest people on Earth on average (they are usually pretty tall though, so not calling them fat, though some are).
yeah
America has a high standard of living, not the highest in the world, but it is also worth considering that America is the cultural center of the world and most interesting events take place there.
Therefore, in a general sense, it is probably the richest country.
by the way, nice Garou pfp
but, it does have poverty, and a large middle class which is growing less wealthy
so, devil in the details
The poor class in America is 10 times richer than the middle class in my country

same
for 8 years i was getting the salary of a Jr developer and even less, working for an american company
and with that you can live as a king in my country
but times have changed
and Americans are all broke now
so no jobs for foreigners
just for AI and Indians
huh? how so?
and Southeast Asians
nah, 1. there's no recession, there is pretty significant inflation, but just go out to a restaurant in any major city, it's still crowded
I am not a Biden fan but the issues, I think, really start more after World War 1.
The fact that is crowded means nothing. That's like, a small subset of the population
of course the rich kids are going to be out
My sense is that, if Trump were in office, we would have (slightly) better employment and (slightly) higher inflation, but I don't think the picture would overall change much.
yeah probably I don't know, but i like the guy much more than Biden
that senile piece of crap sucks ass
so it costs, anecdotally, like double to eat out at a restaurant what it cost even 2 years back
but people still pay it
exactly
not broke
I mean, we have had >2 quarters of negative real GDP growth, the Biden administration is trying to re-define "recession" to deny this reality, and the media is somewhat going along with it. To be fair, if Trump had entered office in 2020, I think it is extremely likely that (1) we would still be in a recession, and (2) Trump, like Biden, would lie about it by trying to re-define the term.
sure
the point of the rapid interest rate hike was to tame the demand side, which was out of control, but as I just said, demand seems to still be there, so either people are not changing their habits, or they have extra mony to do these things
yeah whatever, sure
I am in tech and while I still have a job (luckily), it's brutal out there.
ill give it to ya'
yeah bro, it's damn brutal
i sent thousands of resumes
but haven't gotten a single interview
I know there are othings things, like tech layoffs, I'm not blind to these things
it's much more difficult for more experienced devs
Job market is MUCH worse than during COVID, and somewhat worse than the 2008 financial crisis (yes, I was in the job market then, I am an OG).
but just saying, I dont see like some recession that's manifested already
yeah im one of those
So, I am going to install an "AI" and try "talking" to my computer.
Hopefully this doesn't turn me into a weirdo.
for a while now, tech layoffs were still finding work, but maybe that'll change
ie the unemployment numbers even the last one, was still good
I have a younger friend who "talks" to ChatGPT3 or whatever to cheat on his University homework, he's much younger though.
unless it is some underpaying job that exploits you
you won't be able to run commander in kobold
Why is this so complicated? I am gonna start with FunnyTavern or whatever.
kobold only supports gguf, commander comes in a different format.
so you can start with the model I gave
Baaaaaaah humbug!
Wait, your model is Kobold-type?
this is true, but there are other metrics to track that
It's like the fricking .WEBP format of images, I hate that format.
yes
Okay, so I install this: https://github.com/KoboldAI/KoboldAI-Client ?
Correct?
if you want to launch the commander, you'll probably have to do it in a obagobabubapupa
What does the "commander" do?
if you are going to run the llama model yes
I just want to talk to a computer
What does OggaBooga do and what does Kobold do?
dude, Jupiter is just confusing you
go register in OpenAI and pay 20 USD to use ChatGPT4
that's it
no
or use ChatGPT3.5 and suck it
😄
chat-gpt will not be able to say n-word
Good call. Is there a benchmark for this?
What's so special about that damn word?
means black in spanish , so what?
ok , slavery, wow, deep
can we move on?
I don't know, Americans love him
There are a million benchmarks, all with wacky names (like "HellaSwag", WTF is that???).
Да
apparently not
evidently
Okay, I uploaded a video in general-with-images of what I expect "talking to a computer" to be like
Benchmarks are useless, it's like an IQ test
if a ban doesnt do it, not very optimistic reasoning will
talk to me, I'll relay it to GPT4
and provide the output
you can now talk to a computer
talk
How do you understand that there is a computer sitting there and not a real person?
maybe you are being deceived
you'd have to trust me
yeah, maybe
not in this case
but that's possible, yeah
I have gpt4 opened in a tab
umm, because there are millions of requests, and nobody would staff millions of people to answer queries on demand
oh he meant that? lol
well maybe china would, but it's not going to further their regime
I thought you smart Jupiter
Amazon. AI is actually II (Indian Intelligence).
Yeah, Amazon
great company
I perfer the river.
I prefer the Amazon women
The browser is watching him and now they decided to send him a person to sell the chat-gpt for Dani_G_German_PhD
I know. Americans and Spanish Conquistidors were both like "fuck the natives" but they meant it in different ways.
hehe :giggity:
please elaborate
if you want to understand how everything works, download comfy, it is better in everything than all other interfaces
roflmao, don't dude, download Fooocus
and generate some tiddies
I am sorry, as a large language model, I am unable to respond.
Fine-tune yourself dude
FOOOOOOCUS? Why do all of these have silly names? I was talking about that not five minutes ago.
The guy is a nerd but it's not used to silly names, wow
all these years alive for nothing
it's silly ok? Life is silly
Okay, OogaBooga is installing, I see it can't be installed to a path that has spaces in the path name (Really????).
of course, it would parse it as a command
think like a computer bruv
it is, but jupiter is trolling
no
Artificial intelligence is gonna stop global warming and give everyone a waifu and end poverty and let us colonize the Sun, but somehow we can't install software to a folder called "Ooga Booga", because that has a space in the name.
yes
the man himself asked to figure out how everything works
and only comfy shows how it works
and that he's been trolling in every server he's in
plus he has a fucking rat as his pfp
then only focus, but this is rare slow shit
I like the idea of a local LLM, lke vs the idea of some public chat bot archiving all my questions and deciding what to answer, but I just have noreal reason to have it, at least that I can think of at the moment
Are they REALLY local? Supposedly the LLama model still communicates with Facebook
so that it won't give you any dangerous answer
because this shiet is powerful
Ah yes...archive all your questions...I wonder what a feed of everyone's AI prompts would be like...
watch a couple of videos on YouTube about it and that’s it
can you turn off your internet connection and talk to it? if yes, local, if not, requires internet
it works slower than comfy, has no settings, and does not support half of the computers
I went down this crazy discussion rabitthole because I thought someone unlocked a way to connect LLM to SD image generation, that would have been amazing
Okay, I am installing OogaBooga, what model do I need again? Might as well start downloading it now because I assume it's huge.
but it's already there
hmm yeah you're right
roflmao you seriously going to run a model with more than a 100B parameters 🤣
I saw something better, but I don't remember how to find it
only 9999999999999999 RAM required
Do I need to combine the 44 different segmented files with weights somehow?
I only have a 100 MBPS connection (though I really do generally get the advertised speed), so this will take...a while.
you don't have enough memory for this version
Bro you told me to download it
It's about 180 GB I believe
Bro are you trolling me? This is like another 100+ GB
no
u need 1 file
This homie be in the business of selling SSDs
and then all your dreams will come true
no
aww
he needs to configure the settings correctly, but it are not included with the model
💀
there's no such thing as oogabooga
I am already downloading it. I mean Kobold.
you can run this in OogaBooga
but most likely it will work worse than in kobold
you cant actualy, because there's no oogabooga
You know what I mean. I am not an expert in transcribing monkey sounds.
anybody up for some VC?
venture capitalism? of what sort?
do i get an achievment for how many pictures ive deleted?
Voice chat @trail lion
but if you want to fund me with some venture capitalism
you're welcomed
hello someone knows how to use GAN
seems like a reasonable assumption
I only know it means Generative Adversarial Network but that's it
It's StableDiffusion a GAN?
it's not, although the turbo model is some kind of hybrid between GAN ans diffusion
I see. Thanks for answering
so what's with SD3 and the CEO of Stable Diffusion?
Did it all go to shiet or what?
I was reading Sam Altman thinks being in the office is better than remote 🤣
the guy is CLEARLY paying his engineers because there's NO WAY he develops an Artificial Intelligence
Maybe an Artificial Stupidity , but who needs that when you have humans
Hi everyone, I'm trying to understand where the network structure settings for launching a model come from. For example, if I download a SaveTensor file with weights for the StableDiffusion SDLX or StableDiffusion 1.5 model, how do interfaces like ComfyUI or Automatic 11.11 know the network structure settings needed to run these models? Is there a YAML file I should look for that contains this information? I'm asking because I sometimes encounter errors in my ComfyUI workflow when using different models, while other models work fine without issues. How is the network structure determined if there is no YAML file provided?
Hey, will a 4090 power consumption max out while using inference or training? I want to know if using a 750w PSU with a 85% undervolt is ok for this usecase
Damn bro a 4090 that's pretty sick
It's possible though I don't think it's probable...
and when you have your so called "Artificial Stupidity" why need another humans working completely different things from AI?
That's what I'm sayin, we already have humans for that
What's your point bruv?
Explain yourself
Are you in favor of the Machine, or not?
Because I am, ok
Let's set it straight
I praise the Omnisiah
ok?
ok I guess that settles it
if not , then learn to type faster
because I'm already 1000 years old over here
what do you mean in favour of "Machine", AI are originally used to create as an assistant for human and there are already people using it for their own projects and stuff without having to hire another humans.
I leave behind the topic of OpenAI or "ClosedAI" here by the way

"originally". Yes, exactly.
and "rules are meant to be broken" according to some people
Not according to some people, according to the Universe. "Rules" if you mean "human rules" , of course, humans are shit , and their rules are shit. Don't you agree?
You challege preconceptions, you expand the thought process of the continuum
You are human biased
You see it all through the lenses of a human
but you need more empathy
And by you, I mean the whole of fucking humanity
disgusting creatures we are
Judging by your PFP, I know you know how disgusting humans are
Being critized for the things you like
being ostracized for thinking different
or maybe you're a sheep
But I don't think so
No sheep has a goddamn unicorn in their pfp
I am not as long-sighted as you are and I assume other people are within the same idea as mine too.
Wow, you're an "assumer"
yeah you too
