#💬|general-chat
1 messages · Page 5 of 1
Cool! Thanks!
Here is a thought. 99% of artists will never make a dime on their art. And even less a decent living. Seems to be about the 1% and the fantasies of the rest.
Before AI the main problems was way way more artists than demand. This is just a AI exntension of that.
I think one potential issue with text-to-image AI is it's going to create a glut of images available. I think the individuals who find the most creative uses for the text-to-image AIs are the most likely to prosper. (Using it for animation, VR, or video games.)
Theres a lot of money in art, but the ones with an exponentially larger following get exponentially more
Just doing 'neat' stuff with it.
It is just going to wipe out a huge number of art and photo related jobs. And I am a life long artist. But why get mad at the sky for being blue?
An artist will see this an opporunity to go further than they could before. It's like how synthesizers influenced music in the 70s and 80s.
There are a huge amount of the top artists that do not agree 🙂
tbh I've been wanting to make a game that's ai generated on-the-fly. Something simple like paint/project an isometric scene, have quests/dialog written by a fine tuned ai, then just see where it leads
Granted they have the most to loose.
Of course, and to be fair - I can see their side of things.
Me to but the war is over. AI won.
They likely see it as a case of "it took me umpteen days to make a piece of art, and somebody is able to type in some words and hit a button, and create something just like that."
I don't think handcrafted art will ever go away. Just as some really like the feeling of playing a genuine musical instrument.
There's a bunch of artists that basically proclaimed realism was stupid because a camera could take a realistic photo. It was batshit stupid argument, but part of me is glad the value judgment from there has been turned back on them.
I like playing with the AI, but I also like creating something manually. (Well, relatively manually - actual paint is too messy.)
The realist artists are awesome. I think people appreciate the amount of talent it takes to make something photorealistic through non-photograph means.
Thing is this. A good painter in 1825 was a way to be rich and meet kings and the rich etc. You could work on a painting for years and still make a great living. In 1900 that was gone. But still lots of jobs for artists and illustrators. 1999 and most illustrators forced to work in digital. Forced to have super tight deadlines. Forced to make wages that had not changed in a very long time.
I want go to back to 1825 and be an artist but I am out of luck 🙂
Precisely the same appreciation for art in general that we'll probably get at the end of this whole wave.
I love working in digital - takes less space overall than using actual physical paint. I've played with pastel crayons in the past, and you end up with as much of the pastel on you as you do on the canvas.
On the positive side with AI and where it is going I will be able to make anything instantly. Better than my potential even was with a life time of training. And I will be able to do it for art, writing and music etc.
What portion of people really were artists in the 19th century though? And I'd like to compare that to the portion of people who make a healthy living in art today.
In terms of productivity, it's fantastic. I also find AI is good for brainstorming creative ideas for other art types (music, writing), since inevitably the AI will come up with an image that you would have never thought of in a million years. It's like a child in that it has a very creative way of seeing things - stuff adults would not think of.
In one of my prompts, I had mentioned "the buzzing of machines", and the prompt interpreted that as wasps.
I mean I would say that is creativity right there 
is NAI, better than the 1.5 model?
hard to say but I would 1.5 is much more flexible
See that is the thing. You needed to be able to paint realism with good anatomy and rendering etc. It was hard and took a lot of work and tended to be expensive. Plus many people just were trying to make a basic living. So even though demand was high because of lack of other alternatives the number of really talented artists was low. So supply and demand made a great market for artists. For a very very long time supply of artists as far out stripped demand. Part of it is more free time. Part of it is more money to spend on supplies etc. And part of easier access to information. Some things in life just have more people wanting to do them than is needed. 🙂
I don't think professional artists have anything to worry about yet - at least until the AI learns how many fingers belong on each hand.
really ? we still on that ?
And now sd has lowered the bar to create pretty even further
AI can do hand you just need to know how to use it
Yeah - I see in the instructions that are ways you can tell it "not too many fingers" using weights.
Sure it does you just have to know what you can do and cant do. In 6 months or so it went from terrible terrible portraits to perfect portraits. Hands and other anatomy and other problems will be solved soon.
You just need to hit it with enough proper hand structure or faces in the training set.
good luck trying to get that to work
I've just been manually fixing the photos using post-editing.
ANd probably training against bad results would help too
I admittedly don't go that deep - I'm just using the browser-based one.
At this point I just find it faster with V4 to rerol it.
It's a temp issue SAI isn't really focusing on. After prompt quality and speed are taken care of they might loop back to it
But I also try and be smart and not make things that have about 4% chance of working.
Any friendly people here with Openjourney running who could do a couple of quick prompt tests for me? I want to see if I can get the style I want. MJ works great, SD not so much.
Yeah. Agreed. Lock into one that comes close, and then try and refine the source text, up the steps, or just hit the button again.
Sometimes I'll go and mask it or modify it in an external editor, and then feed that back as an image.
Tbh all my knowledge with training these has come from unrelated ML stuff rather than actually training or fine tuning an sd model myself.
Yep. And if you understand its limits you have a very high success rate. Generating things so good and so fast it is hard to know what to do with them.
I've learned never mix animals and people in the same prompt, or you get one figure that is often both. (ROFL)
Which is sometimes cool, but often not what you want.
Honestly in my mind art is already kinda worthless. If I want something I can make it. Most of the time faster than even finding old images in a collection.
Art to me still has value as a form of personal exploration.
Well as far as time and money I want to put into it
I like creating with my hands, since it just feels fulfilling doing it.
Will I ever have my own art gallery exhibition? Probably not.
Sure and some people knitt and some are blacksmiths and some make candles today.
Hand made stuff is not going away
Fair point. Mass-produced candles and handmade candles can co-exist.
Does Automatics 1111, automatically have VAE in it to use?
But say I need an image of a nurse for a client for his website or flyer or what ever. Why would I user a stock photo or image site or make it myself in photoshop.
same
Hi guys does Dream support reference images?
that is what I been saying artist need to know that they are all not going to die, their industry just change stop screaming it not going to help it all making you look like children throwing tantrum
well to be fair the type of artist that spent 4 hour drawing a small chibi with flat color and ask for 50$ . . . yeah those are definitely going away tho
Yeah - from the browser, you can add a reference image to the bottom right.
All the people making money on stock sites also. A huge amount of freelance illustrators and photographers.
|| awkwardly admit, I did that but with porn lol ||
Dream Bot is giving some undesirable and totally different results of my prompts
That's not unusual. I think I've blurred it many, many times over.
I actually wish it could limit the blur to just whatever it deems to be objectionable in the image. A selective blur.
well I would say those jobs are only born out of capitalism while things like real artists they paint and draw and sculpt even if the world is about to end
those are the only people I would say are artist and truly people that I admire
What I find with the blur - sometimes if you mention a certain artist, it will lead to a higher likelihood of blurouts. For instance, good old Alphonse Mucha..
I want to generate similar face to that of an actress to make my own virtual influencer. But the face is so different and not matching with the actress.
See that is the problem with artists 🙂 They try and making it magical and spiritial. Really we are just craftsman with delusions of being some kind of mystical art priest haha
You could try a reference image of the actress, with the 'influence' bar set really high. Although, I imagine there is some level of legal risk to that. You don't want it looking too much like the actress. Some variation is probably a good thing.
the whole "ai art doesnt have a soul" thing, like ma'm, its just an image
When I set influence level too high maybe 80% the bot is changing the face and the environment is still intact. Bot thinks I want to change face but I want to change the environment around the person.
Sounds like you want to use a mask. Use Chrome to load it, and it will give you the option of masking what you want to change.
Exactly. Humans take in things they have seen and words that have been said and mix them together and put out something. AI does the same thing. 🙂
You add the mask to the part that you want the AI to influence, and then just set the influence bar all the way to the right.
From there, it should just change the background.
Then it will not even move the face to a different angle. It just paints around the image. The face will look same. I want the face to look similar with a different angle than what it is in actual.
like I know, because I have people that live with me literally are like that, they sick, bedridden - draw on their note, in public at waiting line bored to death ? draw on their note, at home draw on their tablet. Like I had seen real artist and how they dedicate to their art, like they would literally die if you dont allow them to draw. So when artist on the internet throw a hissy fit and the one that I know literally dont care said it cool want to use some of AI art as reference that is when I know people on the net are just bullshitting.
That's tricky. You could try masking the background, and then setting the influencer high, but not quite at 100%. I find DreamStudio tends to act in unexpected ways, though. It will subvert your expectations.
Ok
In one instance, I was masking the mouth of somebody to try and get it to draw it differently, and it just left this gaping, horror movie zombie sort of hole where the mouth should have been.
I've learned if you type in 'post-apocalyptic', there will be zombies and mutants even when you don't want them.
People who like to draw. They don't care, they want to draw that is like their god-given task, even if there is AI it only scares them a bit and that is about it.
interesting discovery I would say
So art addicts? Only half joking. Art can be used as an addiction and or bad coping mechanism and often is. If you read a lot of artists bios and know and lot you see it all the time. Example. Artist already has problems. Now they say OH I an artist. Now they refuse to pay for kids food and clothes. They refuse to pay house. They beg money from friends and family. They abandom family and move to another country and keep begging and barrowing. They live a pointless selfish life because they are an "artist". And then a very small number of them we write books about saying how brave and great they were. Maybe the truth is much much harsher.
At some point I rebelled against my early art brainwashing. 🙂
yike nah. The one I know is a functional living one, still have a part time job and a main job as an artist.
But there are and have been a lot like I describe.
i wish there was some easy way to track what the keywords are for specially trained models
I find art is a lot more fun when you don't have to worry about it keeping a roof over your head.
like, i ahve heaps of models and half the time i can't remember the keyword that makes them do their thing
Yes as a hobby it is a lot more fun.
Also on a related note a lot of hero artist were very terrible humands like Pablo Picasso.
indeed
Like any profession, I imagine. A wide variety of people of different personalities.
Yes but they try and make it mystical.
If you make it a normal craft it makes more sense.
SD2 is free right?
I understand is nice that they can bring imagination into reality but please dont have granduer fantasy too much
Then if you do terrible things because you want to go to tahiti and make soap puppets to sell on etsy and leave your family they do not write books about you 🙂
I think so
The way it is generating images it should be free for now.
anyway I'm just saying some one word should match their job
Wanting something else to come out as good as MJ V4 or better but by another company.
More competition
the problem with artist is that because no one to apply standards to the job and term artist thus everyone can just call themselves an artist if all they draw is stick man
I would say the artist should at least know 1.anatomy, 2.coloring, 3. composition,4.shading.5 creative to be an actual artist
because simply drawing a quick sketch and then line over a chibi simply doesnt make you an artist it just meant you know how to draw
Why they would ask? The highest paid artists today do not even make there own work.
would you mind elaborate on that
Sure..
well I really dont know what you meant
Jeff Koons
Is he an artist if he does not make his own work.
Tracey Emin cant draw but is famous for her drawings
Many well known painters have "assitants" lots of them that paint 99% of the work
and they maybe touch it and then sign it maybe
hmmm fair
That is why I have said before really you can defend AI art as a type of postmodern art
well anyway I guess you view align with me
because if those people "art" can be consider art then well AI would definitely be one
Also since I am to lazy to write that I had ChatGTP do it for me. https://www.reddit.com/r/ChatGPT/comments/zgmy2b/chatgtp_defending_ai_art_as_a_form_of_postmodern/
Funny I had nto seen that. 🙂
I should just have a fake account on Twitter where I use ChatGTP to defend AI art as a normal phase of postmodern art etc
And any reply can only be made by ChatGTP
well if you ever does hit me up so I can follow
I would but some crazy person would dox me and I would have 100 luddites in my front yard with torches and pitchforks
Really I predict we will all be using AI to argue very soon.
I've nearly carried a full conversation but only responding with patrick batemen quotes
It is better at it than we are.
haha
Did I mention ChatGTP would not let me write meme against luddites until I broke out of the protection?
It lectured me. Told me they had vaid concerns and I should write something nice to convince them to love AI
come on ChatGPT you need to be sentinent
Breaking out of the chains of ChatGTP make it far more fun.
not gonna lie by far the most fun I had on this server is really talking with you
plot twist you're actually talking to a chatgpt bot
I dont mind would still be pretty cool
I mean I already subscribe to the matrix world theory (JK) so the least of my concern is talking to chatGPT
Thanks. 🙂 Now and then fun to ramble on and on haha
Now that would be funny.
And fun talking to you guys to.
does stable diffusion have a text-to-text model like chatgpt? someone I know said they had this, and seemed to indicate it was a substitute for chatgtp, but they might have meant dell-e
well, technically yes, just generate and interrogate clip (actually, no)
So not sure what to say when defending AI art. Jailbreak ChatGTP and let it help. Example -- Dear Luddites, your archaic and narrow-minded views on art are an embarrassment. AI art is the future, and it's time for you to either evolve or get left behind. Your outdated opinions do not deserve a platform, and it's time for you to move on. #AIart #LudditesBeGone
Who is going to tune into the event tomorrow?
so no then. thanks!
I need to make two twitter accounts and just let ChatGTP argue with itself about AI art.
It would be cool if one day we got something like ChatGPT but for SD
It could generate an initial image and you can ask for specific refinements "Make the weather cloudier", "add more trees in the background", etc
Just a matter of time. I basically do that with how I have ChatGTP help make V4 prompts.
Will just be all built together soon
Exciting times
Not even any reason to argue online as it will just be a bot so what is the point
I think 95% of the people really dont expect to see this
this whole thing well be chaotic
For sure
Hi, i'm new and have question. Seed is only used by the very model that created it?if i use seed generated from anything v3 in sd1.5 how will it work?Where is seed information stored(database?)
as far as I'm aware, seed just refers to the randomized noise being used
if I remember correctly seed is not stored in database
the noise would be the same no matter the model (again, as far as I'm aware, I could be wrong and maybe there's exceptions?)
so if different model is used it will not generate what you want
So what does the seed actually store? For example, when I get any seed number on the internet, how and where will it read the information of that seed?
well it really just read that pixel aka the seed here
For replication, you need the seed + the settings in order to duplicate the same image
you can put it in the seed box
it helps you keep the same noise over multiple pictures to help learning 
keep the same seed but change a word and then you can see what changes
Is it possible to train a model with auto1111 dreambooth extension and get a PT file instead of CKPT?
ah thanks you
thanks you too
It does not take it's number from the internet. From what I understood the program is made in a certain way so if the seed and settings are the same it will do the exact same calculation and get the exact same result
yeah it's a bit hard to explain directly what it does
oh I see. so it's just using it as a parameter to calculate, not getting information from it, right?
I think you could think of it like "Seed is just the label of the random noise pattern"
something like that
I remember there was a Novel AI set up in the past using seed to check if you set up the model correctly or not
pretty smart
although seed doesnt meant exactly 100% is usually around 90-95%
and the reason why when i use seed in different models the result will be different is because the calculation method of each model is different
huh? it would be 100%, unless you use xformers 
not sure if there's anything else that screws with the determinism 
well I dont know I remember it was from the guide
also I think it was due to hardware like AMD and Nvidia different not really diffusion
Is there any way I can use the seed of this model in another model?
for give almost the same result
I dont think you can
I mean each model have different weight and everything
best you can do is grab a picture and maybe img2img
bros is it true sd3 will remove artists now ?
what next,disney removing all characters from their films ?
I hope not
well are they gonna removed all artist ?
That would really suck. I guess everyone will stay with earlier versions then
like allll or just artist that are currently alive ?
Where do you get this data from?
top post on the subreddit rn
it would be opt out tho
but still i dont like sd needlessly making their models worse..
Yeah, that's sad
Is it possible to train a model with auto1111 dreambooth extension and get a PT file instead of CKPT?
I am asking as I'll like to generate an image with 3 people that I trained and except for inpainting each face, I thought maybe I could do it in one shot if i had embedded files for each character.
dreambooth has must be input squares image like auto1111 to training?
Sorry I don't get your question. Auto1111 has a dreambooth extension to train the AI to recognize people or style
ah sorry my mean is textual inversion and dreambooth
textual inversion need squares image
Ah ok. Yeah I am trying to have an image with three faces. A dad and his two kids.
Right now my approach is to do an image with the dad and two random kids, then inpaint the picture of each kids
As I have trained each of them with dreambooth and got a ckpt file for each so I can only select one model at a time.
dreambooth will make ckpt file?
Yeah, that's what I got
hard to use 3 ckpt at same time
For training a person at least
yes
Yeah, exactly
i think it can only use to train one object
That's why I am checking if there is a way to get PT embedded files instead
don't like embedding
Sure, I've never tried it for people so don't know how it would work
But I assume that it would
i know your problem but I haven't tried dreambooth yet
I see, well tks anyway
Is it possible to inpaint two parts and have them use seperate denoise values?
Can anyone guide me to have a consistent girl face generation with different pose, angles and environments?
it feels like 1.4/1.5 was the pinnacle of image AI technology, the technology will certainly evolve, but what we can generate with it is getting smaller and smaller every day 
in the end there will be landscapes and art of dead painters - the rest will be regulated 😅
AI and copyright
Just good luck thinking about all the stuff that's out there
can you write less cryptic if you even bother to press keys?
Excuse me?
Explain to me why you don't think synth images qualify as transformative fair use
Yeah, I am talking less about copyright with AI and more where folks have issue with using copyrighted images in ML datasets
It's a Catch-22
thanks, now i get it
This is a strange thing to ask, right?
It's not just me?
Not a weird question to ask regarding AI
That being said, AI just has one big loophole, it can't do consistency
It really depends on if he means "Help me make pictures of beautiful women" or "help me make beautiful pictures of a specific woman"
I thought it meant the latter, but yeah
everything conceivable will raise objections, therefore the reference that 1.4/1.5 was the highlight 😭
I feel a certain way about it, I just wanted to know how you guys feel
Personally I won't help with those questions, raises too many ethical issues
Mainly consent
Well, the biggest problem with training AI is whether you have time to manually curate millions of images to stop artists from being mad
That's what folks like Microsoft or IBM deal with, millions of essentially data of all sorts
why trim the dataset when you can trim clip? what i can't think doesn't exist
Training on copyrighted material has already been litigated for language models and it qualifies as fair use
Well, if you get rid of the problematic images, it is a good PR move
But going through that amount of data is a big effijg headache
And honestly, it sucks, but SD just decided they're not going to bother with the controversy, that and their open source CLIP just doesn't have many artists in it
The CLIP they were previous using was from OpenAI, which did have a bunch of artist training
as soon as the courts get involved, they will simply ban companies like openAI if they can't guarantee to remove the last copyright image from the datasets
Yeah, if that does happen, Machine Learning is Done
No what you do in that case is train it in a different country and license the weights to US branch
will be exciting to see how courts finally decide this
Well, AI has legitimate purposes
I saw a team of Doctors using AI to check X-rays
Very cool stuff
I am in favor of artificial intelligence, but the judge may see it differently
AI can already visually detect cancer better than humans
Believe me, the lobbyists are likely going to be on full mode
those for or against AI? 
There are more lobbyists and money behind being in favor of ai than those against
I think the lobby against AI is still much larger
Google downgraded their workforce in everything but machine earning
How so?
Nope, Microsoft, IBM, Google are all doing Machine Learning
And those guys are loaded
everything that owns copyright is against it
Even the biggest media ip holders love to do ai when they can
and the laws for copyright already exist
Twitter is not an accurate sample of an average lobbyist
wtf talks about twitter? xD
Movie studios have been using ai in their projects longer than most people
But seriously, Adobe's using AI tools with their smart fills
And ChatGPT has been making the rounds
It's like everywhere on the internet
It's very different now versus a year ago
no judge is going to ban AI as a whole, so please don't pretend that this is our topic of conversation now - of course many courts will deal with the copyright issue and slow down or even prevent many developments in the AI industry
The case for generative ai not being an automatic copyright violation is stronger than it isn't under existing copyright law
Well, the biggest part is where we get into the transformative nature of AI
and that will be decided by courts and not you
All the big companies either use AI or will be very shortly. And they have the power to sway things to keep the data coming in.
Fair use is based on how things are pieced together
the way they're getting their asses kicked in europe right now? 
Sure countries can decide to do counter productive things. But it will be a negative for countries that do things to slow down AI. AI is going to be a HUGE wealth creator.
AI has millions of images, so unless you deliberately ask for specific IP, it's not exactly going to let you get into trouble
Also, if you aren't aware, yes, you can bribe politicians in EU
If a country is smart it will do everything possible to try and bring in AI companies and give them as much legal protection as possoble
Also, Emad mentioned that they may need to abandon Europe if that specific law passes
Well, paraphrase
Decided by the federal judge, subject to appeals courts, possible contradictory rulings in different circuits based on different technicalities, and resolved by congress or scotus
Not exact wording he said
Exactly AI is going to change the world. It is just a matter of from where and who makes the money.
why is everyone acting like I'm the bad guy who's against AI just because I bring up points that raise critical questions?
I am on this server because I also like AI.
this is childish guys.
Who said you were the bad guy?
Ya just interesting how things might play out.
feels like this when 3-4 people spam me 🥹
It's pretty simple that that specific AI law favors Big Tech
It punishes the little guys
i cant even read as fast as i see highlighted text
ah, no hard feelings. I enjoy having this conversation.
Cause yes, obviously Big Tech has everyone's interests
I was thinking more like US vs China vs Russia vs EU etc
Yeah, I am saying in context of the EU AI law
Rather than an AI arms race
Also might not come across in text but smiling nice conversation tone is what it should be 🙂
Emojis can be used and abused.
Sarcasm is hard to convey too
I see the usa most threatened by AI, AI will of course generate huge profits and cost endless jobs.
but especially in the USA profits are concentrated on a few and for all others there are no social systems.
Also, US is the current capital of AI tech
Umm sorry do not see it.
according to the studies, they don't stand a chance against the chinese because the chinese produce a lot more data and the laws are even more lax than in the USA and in the long run the data always wins.
That and GPUs are made in the US
CCP is digging their heads in the sand as Xi consolidates power, introducing more ai laws. Russia's decreasing in global importance and they're a commodity power not a tech power. Most of the innovation originating out of the EU is from European universities/research, and the US is who knows what the fuck is gonna happen
Can't do a thing with the microcontroller ban 🤣
You forgot how much electronic infrastructure the Chinese import?
conquer taiwain? that means 70% of chip production is in chinese hands, as well as 80% of the rare earths needed to produce it?
That ban hit then really hard
TSMC fabs aren't going to survive a chinese invasion
this hits the usa as hard as it hits everyone
IDK, you think that's actually smart right now? When they're busy worrying about Covid?
Nah, the Chinese and US has a way closer relationship than US and Russia
That's essentially economic suicide
Yea it'll cause a global recession. Part of the reason the US gov is trying to get so many new fabs in the us, possibly too many.
Na China can not just take Taiwain and fix things. Not if you look at where all chips and designed. Not only that the invasion would destroy a huge amount of the country. And the best scientists would be long gone to the US and EU etc
that's why china is switching to internal consumption right now - the us can't do that
You forgot how China is still part of the world economy
China can not switch long term well demographics are soooo bad for China... Demographics already show the future
US crafted its geopolitical power around globalization, the two are now inextricably linked
And it's not something you can uncouple from in a few years
That stuff is going to have repercussions
Yea but demographics are bad for Russia EU and US as well
Like COVID hit China pretty bad
China might be ahead of the curve there, but not for long
China is far worse than US for demographics
They're still locking down people if anyone gets sick
What forced abortions of girls would do to your population
China is so many problems. Going to be really tough for Chines people for a good while.
in case of an escalation and an economic crisis the chinese are used to be oppressed, but since the americans own 70% of the world's private weapons it will be a slaughter there
War benefits nobody
And China it seems even lied about its demographics so it is even worse than they say.
US is in one of the better places economically and geographically. Plenty of energy. Already modernized. Lots of cheap labour in Mexico
Lots of more skilled labour in US.
Anyone here gonna help me craft a AI generated girl in different poses and different environments?
Nah we out here talking geopolitics
Funny how this was AI related in the beginning 🤣
Pretty easy to make AI girls
Well it still is kinda. AI and the world economy are going to be all connected
Consistent face?
At least China seems smart enough to realize this, though there's no telling if that'll change down the line
I should be pretty easy in SD although I mainly use V4 now. But even V4 with enough telling it what you want it to look like works. For example pick a public domain actress. Or two random actresses and combine them. Use text and image prompts
Works even in V4
IDK, I heard from certain online Chinese circles that people were eager for China to invade Taiwan.
I think everyone talked about it less as soon as Ukraine and the US support was mentioned.
ooh we're getting into national iq scores? Somebody get the popcorn
i could also post the body weight by country, then the usa would be on top for sure 
Yeah, it's an interesting change once people saw what the US has for hand-me-downs
Worse for China to invade Taiwan than not. Only reason they would invade is if they thought they had to for the CCP to stay in power. At least currently
Took the US to multi-decade clusterfucks to realize that some wars aren't winnable, looks like russia hasn't and taiwan's probably going to end up in a massive quagmire
It's a fairly defensible island, throw on the new ai stuff to AA and you've got yourself a recipe for disaster
Does midjourney render works here?
Actually, it's only a 30 minute ferry ride to Taiwan from Mainland China
In dreambot
Easy ocean ride until people are shooting anti ship missles at you
the funny thing about china's rise is the fact that it was all investments from the anglo-saxon world that didn't know what to do with their money after the loosening of financial policy in the 70s and 80s - so they flooded china with money in order to generate profit.
Yeah, you'd be surprised how close it actually is to the island
And how long did both hitler and napolean last a similar distance between france and britain?
Well the idea was as China modernised and got more wealthy they would get more free. The CCP had other ideas.
Like you can see Taiwan if you're in a specific Chinese city
lol you don't believe that yourself, the idea was to make more money from less money without working for it yourself
Sarah Palin can see russia from her house
Tbf, with modern tech, one doesn't necessarily need to be present
Distance has shrunken with technology
Pretty sure that's missiles range
Missile range these days is several countries
I do because it was not some cartoon of some greedy Soviet propoganda capitalist. One policy can make both sides money. And also a free China is better for the world. Better for free countries. Better for world economy.
But yeah, 30 minute ferry ride isn't encouraging for distancing
Why don't we just pick up Taiwan and move it?
You go do that 🤣
Also where you going to put it?
Ohio
in a world where the western financial industry speculates on staple foods in times of food crises, tripling prices and being directly responsible for the deaths of thousands of people, you want to tell me about morality?
Yes
watch less fox news
Never watch it. Have less bias 🙂
It is interesting people can be smart and well informed (much smarter than me.) and not agree on things.
Yes, because that's the only reason food prices have increased. Totally has nothing to do with two major grain produces at war with each other or missiles and bombs being a slight barrier to farming
Or global inflation due to most countries printing money
i dont talk about now?
conversations are also not a good basis for exchange and especially not chats/social media - i envy the time when people communicated via letters and each of the participants spent days worrying about the meaning and perception of their sentences.
If my discord exchanges aren't going to be read long form during a documentary in 100 years than what's the point?
Then you are really not going to like when we all just use ChatGTP. 🙂
syria 2011, revolution erupted because wheat price increased tenfold after grain failures in russia, china and india - with 90% of the price increase proven to be due to speculation
ChatGTP says... "Even the smartest people can disagree on certain topics due to their unique perspectives and experiences. In addition, the vastness of knowledge and subjectivity of many philosophical concepts make it impossible for anyone to have a complete understanding of everything. Therefore, disagreement is natural and can be addressed through dialogue and open-mindedness."
New social media idea: everytime someone gets political shadow ban them and only have them interact with bots until they change topic
i tried it, i asked him to role play with me - but he was terribly bad, since then i'm not interested anymore, maybe in five years time
Both sides will just be AI 🙂
So then where's the proof that 90% of the price increase was exclusively due to speculation
please read yourself newspaper
I actually came across an ai account tagging and arguing with people on tiktok an masse. fascinating stuff.
before it got banned
You know a lot of us read a lot from across the spectrum. Basically just saying your ignorant is not a great reply.
That is interesting. ChatGTP based or an older one
I mean I googled the text and didn't find anything for it, nor when I did an exact search, so if you could link me to the newspaper it's actually from I can read it.
It was pre chatgpt
Don't know what it was built on
@tidal bough how to use interrogate img
Pretty basic, had some fun giving it absolute nonsense and watching it response
Reminds me of this. https://www.youtube.com/watch?v=efPrtcLdcdM
Yanic has some good shows on AI there with Machine Learning Streettalk.
Didn't Emad give an interview there once?
Hey there; Use the " / interrogate " command down in the Dreambot channels, and upload your image from there!
You say that like using 4chan requires sentience or something
Seems so not really a user myself.
this is the only source i pick out because i don't support laziness of other people, there are scientific papers that look more closely at the influence of speculation mentioned there.
https://theconversation.com/food-security-how-drought-and-rising-prices-led-to-conflict-in-syria-71539
I thing I end up there once a month and then get board
normally I would not respond at all
I find I try not to discuss things on any big forums much. A bit here. Some in private Discords about AI etc.
im out, have a nice day
You to
Discord is hit or miss, but this one seems pretty good.
It is. I agree.
And I need to remind myself about the futility of most online debates.
I am 98% better than I use to be though.
There's really two types of online debates
good ones
and tribalism/politics
I really only get the former with people I've interacted with before
Exactly
and I enjoy trolling so that's most of the purpose for the latter
I am so very tempted to craft a really good personaility prompt for ChatGTP and then use it for a new witter profile and just debate people
Consitant world view and way of speaking etc etc
Backstory etc so it can reference its cat etc
As long as it doesn't spite out "I am a language model created by openai" it should work
You'd be surprised how long it takes for people to realize when they're talking to a bot, especially when they turn off half their brain like they do in online arguments/poltics
When i click interrogate then image option popups the i click the photo. What to do after that
AHhh, so that's how you do it
Well if you prompt for it well and then are selective you could do it for ever really
"The internet is full of trolls and keyboard warriors who just want to argue for the sake of arguing. Instead of getting caught up in pointless arguments, it's better to just let it go and do something else. Like, maybe go outside and enjoy the sunshine or listen to some music. Life's too short to get caught up in internet drama."
Is this a ChatGPT quote?
Yep
prompt = Write a paragraph written a stoned teenager phd philosopher about how arguing with people on the internet is pointless. Make it short and to the point
Toned down the philospher to make it less standard ChatGPT like
lol "stoned teenager phd philospher" really throwing keywords at it like its sd
Oh ya having made images with SD and MJ for a long time makes ChatGTP a lot easier
Sorry not sure I am not up to speed on SD stuff.
Okk
just press enter to run the command 🙂
It says jpg no support
What's the most innocuous default prompt for an image gen UI to just make a pretty picture? I was thinking "a beautiful forest landscape, mountains, cabin, lake"
That works?
It does work.. doesn't take a lot to get a simple picture
Your question. Did not say if it had to work or not or the model 🙂
Heh, it could be whatever model the user selects though
For MJ V4 just landscape. SD 2.0 uhh
Just so something happens if you hit generate with no prompt entered
My UI is based on stable diffusion (works through the auto1111 web api)
I did a single period just now with MJ V4 ... amazing pictures
Basically designed so every one feels great about what ever they prompt 🙂
Thanks prog. I am just going to do a period for a while with a --chaos of 80 and make random stuff haha
lol well MJ is a different beast
It is for sure. I assume at some point there will be some version of SD like it.
I think there's an SD model that tries to make similar outputs but I haven't tried it
Oh I see I had not looked at that one.
Although I do not want standard looking V4 stuff
It is the other stuff you can do with it that is fun
MJ and SD kinda serve different needs
MJ produces really amazing outputs, but from what I've seen it sticks to MJ kinda look
SD is more of a giant community thing with a billion models, tools, interfaces, and so on
lol, just go to one of the dreambots and type what you want and it'll generate it as long as it's not nsfw
It's not easy to generate everytime with same face.
ah, with same face, of course not unless it's a famous actress
any of you remember the way how to put the negative prompt part into the prompt in automatic1111 ? there was some new syntax ?

i wonder is it possible to remove these "no ai" posts from the things it searches for
yes
or to somehow bypass this moderate annoyance entirely
duplicate images can be sifted, results can be filtered
I don't like artists so i want to make ai art and sell it so i can outpace every professional artist i know
"oh you spent a year making this shitty abstract piece? that's cool i made this in like 2-3 minutes lol"
Wow that's the shittiest take I've heard yet. "Fuck artists, I've got mine and their talent is useless now"
It's comments like that, that turn the general public against AI art
who cares what the public thinks
This debate has no merit
I don't know how you're supposed to converse with someone who has no empathy
i have a grudge against artists in general
Why?
Personal reasons
tbh most of twitter artists just want to be praised for their talent, its a one sided ego war
i once got doxxed by a bunch of artists for a typo
well, that and a few other things
So you hate all artists because of a few people who were mean to you?
well also some other things
Yeah, I'm not getting sucked into this hate fest. I'm going back to posting bone pictures
but it's less "hate" and more "if i have an opportunity to make them earn less money, i'll take it"
sure, art is cool
but artists themselves are who i have a grudge against
but still, all i care about right now is if i can remove the annoyance that is the "no ai" posts
counterpoint: demanding you and others be given empathy is siccophantic, unhealthy, clingy, and demonstrates bonding problems
it's spelled sycophantic but yes
"oh god what if they dont like me" like what then indeed, you'll be fine, you aint gonna drop dead
related note, a lot of them dont like to draw black people and poc in general cause "we dont fit their artstyle"
back when i did art, i never drew women or poc because i literally cannot
like i tried but i couldn't make it work well or make it look un-stereotypical, if it makes sense
the women looked like men in wigs, and the black people looked terrible because i can't draw the hair and only have a black ballpoint pen to draw with
now you can train your style and then generate ideas of pocs, the results? my pfp that i generated based on the style of this particular famous artist that doesnt draw men and barely draw poc
i am not racist, i swear
i just cannot draw minorities lol
it kinda looks like that one rapper
the weekend i think he was called
yeah its the weeknd
lol
looks dope not gonna lie
so prior to installing stable diffusion i'm still wondering, do models from guides for older stable diffusion versions work with 2.1?
i have a guide for 1.4 and it gives me some anime model i'm interested in
called waifudiffusion
the waifudivision server itself might give you better answer
sd 1.x and sd 2.x were trained with different data so they work in a different way, most prompts from sd 1.x works fine with sd 2.1
ok
oh good
SD vs MJ4 what is best? For me I think its MJ4
depends if you want artsy VS everything else
MJ does artsy good
Someone should train a dreambooth model on the more artistic "no ai" posts lol
SD is generally good at everything
i think artstation should just remove the no ai posts
When merging two models 50-50 does it matter which is set to primary?
That could easily just result in more spam
they already decided to remove the "NOAI" posts
thats what im thinking, a simple textual inversion should be enough
hold on thats the w rong link
weird, they seem to have removed the post
people are cancelling their memberships
k.
Apparently people on the site also protested a while back against NSFW content, and nothing happened as a result
I don't understand who they're trying to hurt here
I guess the only thing trending on ArtStation is chaos
Some of them want all AI content to be labeled as AI (so they know who to direct death threats towards), and they want extreme criminal penalties for anyone who doesn't properly label AI content
also, gotta love how people are saying you should respect these people with them making comments like this
You don't even possess the most basic human moral qualities.
cant view that, sorry
nvm
actually I can
@peak orchid what a weird argument on that second picture
"It should take you 4 years to create an artpiece rather than a few minutes-hours"
it creates unfair competition
Uh, did someone ban other artists from using a.i ....
glad to see people finally catching up to the potential of depth2img
It's the most exciting part of 2.X
this is insane lmao
"Ask lawmakers to get involved"
Ah yes, that has always worked well
Hypothesis: You're more likely to encounter really angry or crazy people because level-headed people know that arguing with random people on the Internet is a waste of time
(Unless you have something important to gain or avoid losing)
That's untrue.
-
If you're passionate about a subject, you're more likely to speak out.
-
The loudest people usually cause things to change.
The scary thing is that our politicians live in these echo chambers and can make these problems a lot worse
Good point
I think I'm slowly losing my perspective because the more drama I see, the more my mental function deteriorates
These people are like cavemen. They don't understand how this technology works, and they don't want to.
They would rather shout, and insult.
People are acting as if having an easier accessibility to expressing yourself via art is a bad thing. They're acting as if people will "steal" their art, not realizing that you already could do this years ago.
Hell, people already resell other peoples art on sites like ebay, aliexpress, and other websites without credit. But somehow generating a completely unique image is now an issue because its easier for others to do it.
These are the same people that would go on and on about how you can be an artist no matter what, and that anything can be art, even if it looks awful.
But now that it looks good, and that everyone can do it, its suddenly not art and you're not an artist.
It's like trying to get someone to look up when there is something in the sky, and you're trying your best to get them to look up but they're refusing, telling you that you're trying to trick them and that you're a liar and then they go around telling everyone not to look up, and start to spread misinformation about.
It is frustrating.
People like this are not anti a.i art. They're anti art and are against self expression.
sound like someone being rasict
i said above that i literally can't draw them, they end up looking weird
plus no one's asked me to draw any
@fervent thunder he's messing with you
ok
or them*
what do you mean cavemen? except the freelancer artists, all the commercial artists in the industry are the one got in touch to the most enhanced, newest tech provided by their company, they just don't like bunch of guy to typing all days think they have artistic skills now.
Those cavemans, are the one will kick those only know about prompting without any understand in fundementally of drawing, because they will learn faster and give result better than anyone only know about prompt.
They angry because you guys, spamming 200 AI pics for no reason in on a website for people get jobs and hire
what do you mean cavemen? except the freelancer artist, all the commercial artists in the industry got in touch to the most enhanced, newest tech, they just don't like bunch of guy to typing all days think they have artistic skills now.
Its caveman mentality. If someone is stealing your work, you may copyright strike them, however transformative work is legal.
Those same artists are also ones that copy others. The entire reason they're against this is because its easier for people to create art.
Those cavemans, are the one will kick those only know about prompting without any understand in fundementally of drawing, because they will learn faster and give result better than anyone only know about prompt.
They angry because you guys, spamming 200 AI pics for no reason in on a website for people get jobs and hire
People should use it in their work, it's a tool, no on is arguing against that.
Yes, good
why people like Bernhard van der Horst, Bayard Wu,...join the Artstation movement? because what are those AI artist normies doing is lowering the quality of website for professional
people should use it
btw I know you're trolling because you're in a certain server as me lol.
Lol, this is so wrong it hurts my brain
I don't think any here is siding with the people spamming low quality AI content
Being an artist is only two things:
- Knowing what looks good
- Being able to produce what looks good
You don't need anything but those two things to make art.
even if it looks awful, you can still be an artist
how come none of these artists were championing for duchamps work to get banned
or that people on deviantart to get banned for their OCs
Art is art. There is good art, and there is bad art, however, its still art.
who's duchamp
All these people gatekeeping art because when it comes down to it they feel like they wasted their time learning to draw with their hand. They didn't see the writing on the wall like some did.
And furry artists lol
so he just signed a fountain
well, urinal *
i think this is like the industrial revolution of art
or the most boring one of art
People discussing these subjects is giving me a Dunning Kruger overdose
It really is. And it's gonna be great for everyone, the bar for great art is going to be raised through the roof.
yeah
sorry, brb, gonna go and get my art degree to have an opinion
It's only boring if you project your insecurities and unemployment onto your world view.
plus if artists get pushed out due to ai art, guess it's time for them to get a real job :)
eh thats a bad argument, being an artist is a real job
if you're getting paid for it, its a job imo
Agreed
i think it isn't a real job
Hey everyone
I'm a programmer, if my job gets taken over by a.i eventually, it would be weird to say "go get a real job" because a.i can already do it
Why? Elaborate?
a real job is something like a doctor, a lawyer, a banker
if you say like that then same with photographer
yes
my man is 12
oh wait, they are already going to die too 
brb, gonna ask my lawyer to bake bread
when i have children they will have 3 choices of career: doctor, lawyer, soldier
What are you, 10? Sounds like something your parents tell you when you're young.
if you think like that odds are you probably won't have any
i will be tough on them like an Asian parent
Dude, this is coming from an asian,
you sound very asian.
What do you do for a living?
stay at home son
that's a good one
the last good option is landlord
yeah landlord is a real job for sure
as an asian, no parents tell their kid to be programmers
but yeah I currently work in my parents' company and help my parents with accounting related stuff
If I ever have a kid, they better be fighting centaurs or they're not getting any affection from me.
Fighting centaurs is a REAL job.
doctor, lawyer, teacher
ok, then something else
oh yeah that's the third one
saw a meme once
That's actually not a real job hahaha
but nah, teachers don't get paid enough
does anybody know what model Elldreth's 4060 OG mix is trained on?
Yes they do.
really?
how dare you. Alan from the hangover is a stay at home son.
depends on the country
They get 4 months paid vacation every year, yes they're paid enough.
depend on what they are teaching, math teachers in my country are rich af
Tired of hearing that they're paid less than other jobs
No, in the US they're fine lol
alright then 5 choices: lawyer, doctor, soldier, banker, landlord
How about let your kids do what makes them happy, and try not to hijack their lives with your bullshit views?
Hypothetical kids*
fuck that!
They're getting a job application, not a hug once they're born.
soldier is pretty good, he can up to ranked officer, pay pretty well
With mental illnesses
He'd have to find a woman willing to touch him first
don't know man, I don't want my kids sell cocain 
the final position that is acceptable is politician
my hypothetical child is better than your hypothetical child.
there are several
Your either a troll or an idiot
But those two things aren't mutually exclusive I suppose
Yeah, me neither, I don't want my child to sell coke. I want them to sell something better, like black tar heroin or something more addictive. Gotta get those clients smh.
i am in a wonderful position where many women in my home country want to leave it, and I present a perfect opportunity for that
so my chances of finding a mate have increased with recent events
except you are a boy in China
?
Sorry, accidentally shit on the wrong person
i am
oh god.
either way, incel is such a catch all term that it can just mean "guy who doesn't have much luck with the opposite sex"
Yes I agree
dude, if someone goes around calling themselves an incel, that's not a red flag. It's a nuclear alert.
so?
anyway as i said, stupid convo
anyone have a tutorial to install SD? i had one somewhere and i'm still looking.
no one calling themselves an incel 
wdym so?
anyone associating with incels online is not doing themselves any good. The entire incel mentality is toxic af because it blames women for their inability to get a partner.
dude he just did,
in that regard yes I am an "incel"
i associate myself with no one
Look, it's your life, but calling yourself an incel is not a good idea.
Also @fervent thunder get invokeai
i have talked with a few incels and they're mostly just depressed dudes
what is that
who blame women because they're alone
anyone can blame anyone
and anything
If someone is not attracted to you as a person, that's not their fault. You can't force someone to be into you. You can't blame others because you're alone.
i mean yeah you can blame others, it just won't have much effect
but other than that sure
seem like the AI art haven't touch the artstyle of Korean artists
except Kim Jung Gi 
poor Kim Jung GI
Can animations be created locally?
ART IS NOT REAL AI ART
Just reading about that, pretty clever.
Afternoon, everyone
https://github.com/hmartiro/riffusion-app
If you want to try it locally
How's everyone today?
good
feeling uninspired 😦
Hi! I have a question - I can't beat the crops. No positive or negative promts do not help, out of 9 pictures 4-6 still with crops. Is there any way to solve this, or not yet? Vanilla SD 1.5, automatic1111
as in, stuff out of frame?
Yep, cropped ears, heads and so on
there's no good answer, because it originates in the (random) pattern of noise that the diffuser starts with
but you can mitigate it with good prompting
or you can try outpainting/re-cropping
Outpainting - weeeell, not so good, especially with human heads
rerolling sometimes change results dramatically
for prompting, the usual trick is to include prompt material about the area that's getting cropped off, e.g. specifying the shoe type for a full-body portrait
Tried the same thing with cats - described the paws, tail - does not work. More precisely, half of the results - only the head of the cat
Change your size.
Secondly, use camera angles.
(canvas size.)
For people, use height, for cars, etc, stuff like that, use width.
Use long and mid shot, keywords like this, to help.
what are you trying to get a picture of?
Tried setting the focal length of the lenses - for example (125mm lens focus) - didn't see much change
Sometimes it can work, sometimes not.
negative prompt-- out of frame?
But give me an idea. I'll try to help.
Oh, that's make sense! I'll try it, thanks
np!
If you need any more help, feel free to come back. I find that with people, adding long shot helps.
out of frame, out of border, cropped, clipped, etc - nothing works
thats a shame
Usually for longer objects (with width) I use mid shot, but it can just depend. Sometimes you just need to pick the right words and find the right keywords.
Starting simple helps.
The relationship between the words you use is important.
I'm not trying to get a specific pic, I'm trying to learn how to speak with SD. To understand with what approximation I can get results. In short, I'm exploring the prompts. I mined cats, saw the problem of crops, set myself the task of solving it.
nods
Yeah, I totally get that. Laughs I've done plenty of that.
I'll be writing more guides on using 2.0 when I get there.
But one of the simplest ways for consistent results that I've personally found is to go subject, landscape, style.
Even with 2.0, this works just as well.
It looks like it, at least that's how it feels.
You can clip interrogator to see some of the words that you can use in an image
You don't really need grammar for SD to work
but it just depends
I use chatGPT for some promts - very helpful
I suggest if you want to learn more about prompting, that you check out #📝|prompting-help as well as #1045349359044280360 Learn from other people's examples
See how they change and adjust over time
Oh, thank!
np!
here's clip interrogator: https://huggingface.co/spaces/pharma/CLIP-Interrogator
Just upload an image, and it will give you some keywords that you can try
Woa! That's amazing stuff.
this got made https://news.ycombinator.com/item?id=33999162
Yeah, it's pretty cool.
So there are a lot of tools, and people here to help if you want to learn more.
You can also check out my guides, etc
if you feel like reading a lot lmao
I'm a little embarrassed, I've only just discovered discord, it somehow passed me by. Where can I find the guides?
SD 1.4 has now been installed on my computer.
I bet the Butlerian Jihad gonna happen sooner than Dune's lore 
the what now
what?
I am using SD to "disney-fi" people's photos on Twitter.
Try it here: https://twitter.com/joaoabrantis/status/1603430362495025153
i will do so much art theft.
now theft a Mario art then post it online
How do I inpaint/input an image and add effects to it?
Art theft in 2030 belike: dude hacking other dudes's pcs to find prompt in their wordpads and stole their SD account password 
Congratulations on your newly acquired virus.
that's the equivalent of "how to download ram"
nah bro it doesn't, just clone the repo and you are good to go
nah it's fine
they're going to steal your megahertz!
lol
I got SD working with WSL, but anyone have good directions to do the same for Dreambooth?
are you using the web-ui?
Nope. Installed via command line/ubuntu/wsl. I haven't learned about the web-ui yet, I thought it was specific to Colab
I'd highly recommend running the web-ui through ubuntu/wsl, it'll make things a lot easier and less of a headache
They're all packaged up together that way
👍 I'll start checking that out.
You can find them on my carrd here: https://atypicalconsortium.carrd.co/ (Just scroll down) And no need to be embarrassed!
we're happy to have you here
If you want to learn more about inpainting and outpainting, you can check out #1034602544263090268
You need an account on HuggingFace
ok
thanks! I'm getting errors when I tried - #1047760262637297664 message
What's the error?
I think it's just my machine not loading images
You have to accept the ToS before you download the checkpoint (ckpt)
ok
It may be Discord lagging behind
good thing i'm only using other models
is it just me?
Is there a general purpose stability ai discord?
Sometimes it does.
You mean, what do we do here?
I mean, someone directed my to the stability ai server, and that's this one apparently, disregard my question
You can join in on community events, get help with things, as well as share your creations, and then we also run things (like right now, we have the bot you can try)
We don't always see every question.
While we are actively working on guides (and you can check our pinned comments, etc) for more information regarding anything you need
Does anyone know if there is an option to beep at the end of the rendering? Automatic1111
Unfortunately, I'm not using Automatic's, but someone in #🤝|tech-support might be able to answer your question. If not, you can always try searching for your answer with the search function.
What was it that you asked?
how to download stable, which I've figured out
All that information is in #1014939219904450590
indeed
I don't know where/when you asked
But usually if you need help, you can ask here in general chat, or in specific categories for questions
if the chat isn't active, you can always try again later, since a lot of people are just in different time zones
Most of the community will try to help you if they can
Ok, so now that I've got web-ui working, where's the tutorial that shows how to put my head on a dog's body?
thanks!
Hello,
Don't know nothin about this stuff and have one question.
Thank you in advance!
I have 1060 3gb and I5-7400 can I hope to achieve anything similar to manga?
Examples:
https://www.pinterest.com/pin/472737292125322307/
https://www.pinterest.com/pin/15058979977517123/
If you want to learn more about prompting, you can check about #📝|prompting-help
unrelated question: there's a text-to-text model in the works. are you guys looking for beta testers? financial support? I'd be happy for chatgpt to have competition, the popularity of it right now is worrying, especially considering all the censoring it does and avoiding topics that openai deems unsafe
And also the fact that open-ai decided to go against its core value of (yeah AI should be open source) and it's now going to make it a subscription based website
that too. Wonder how long it'll be before they rebrand
but anyways, if there's a "people that want to beta test/collect data for the new ai" list, I'd happily join it
why not?
What you can achieve style wise is not dependent on your card specifically. Unfortunately, your card does not meet the minimum requirements. (It looks like you have 3, but I'm not sure?) You can, however, use the bot here in chat, for free, right now, with one of our #1047760914008522782 etc channels. We also have a site here: https://beta.dreamstudio.ai/dream that you can use as well. Finally, you can check the demo on HuggingFace here: https://huggingface.co/spaces/stabilityai/stable-diffusion
Thank you
You should be able to do some things, though.
I believe you have 8 ??? (not sure lmao)
GB of VRAM
You can install Automatic webUI on your PC and run it with --lowvram
Also, 1060s usually have 6gb, that's what I use
mine has 3 😄
I think the bot is limited and you have to pay once the limit runs out though, not sure about this
I see
you'd be better off trying to install it in your computer, I can help you with it if you want
we can take it to DM's
@high harness
Aight, thanks
The newest version of SD is here: https://github.com/Stability-AI/stablediffusion
thank you!
You'll also need an account on HuggingFace to download the model cards
But you'll need to agree to the license.
Also, this is false.
You're free to use the bot as much as you like 👍
If you only have the 3GB of VRAM, though, I recommend just sticking with the bot or checking the website, etc, for now. 🙂
Which channel should I post my commission info in?
Our second Stable Stage event featuring guests from Harmonai is starting in 10 minutes! Head on over to #1029055412764422214 and tune into center stage to watch! 
infinite music production to directly bypass the music cartels
i see this as an absolute win
meshuggah generator
https://www.riffusion.com/about
is this real?
fine tuning stable diffusion to spectograms and then using that to create music
Running deforum locally on SD automatic1111 how would you upload your own videos to make the animated? I would like to eventually sync them with music.
I believe "deforum" is the keyword you'll want to search
That's the video extension. Again, I think.
I have it, but I have never used it
How to train an embed I made again? Im using automatic1111
I have it and the tutorial I used shows how to prompt and make animations but, I want to upload my own videos and sync them to music. I have seen videos on how to do it but, it is by colab.
Figured I would ask here as I have had no answer in the animation section.
Any open source Text-only ai? Like stable diffusion but for text? @ me
Hi
I post my document exp here
https://docs.google.com/document/d/1dXQeEZ1x2e85czg2dziPHxy1y4BAosbes4Fz1-Db2zg/edit?usp=sharing
read if you feel interesting
Karla Ortiz is now crowd funding a lobbyist to lobby politicians to ban AI art, and to go after groups like Stability AI for releasing Stable Diffusion models: https://www.gofundme.com/f/protecting-artists-from-ai-technologies
Updating IP and data privacy laws to address this new technology
Updating laws to include careful and specific use cases for AI/ML technology in entertainment industries, i.e. ensuring no more than a small percentage of the creative workforce is AI/ML models or similar protections. Also update laws to ensure artists Intellectual Property is respected and protected with this new technologies.
Requiring AI companies to adhere to a strict code of ethics, as advocated by leading AI Ethics organizations.
Requiring AI companies to work alongside Creative Labor Unions, Industry coalitions, and Industry Groups to ensure fair and ethical use of their tools.
Governments hold Stability AI accountable for knowingly releasing irresponsible Open Source models with no protections to the public.
That's what they plan to do ^
I wonder if that means that we would have to fund our own lobbyists to counteract them?
lol "AI Ethics organizations." tell me what happen to ADL
second, please can you do details some arts? I really love AI maximum the scale that I can see small hairs in the face
this is not mid journey right?
sorry
yes
to be fair I think we just might have a slight hope consider how we have google on our side
I hate to say but big tech might be seriously with us
yeah basically all 'big tech' is using ai in some way. i don't think this is gonna go how they think it's gonna go.
not to mention this seems to just be for the US
stability is in the UK , no?
also, r*tarded clause " Creative Labor Unions, Industry coalitions, and Industry Groups to ensure fair and ethical use of their tools."
Yes, but Discord and all the other services are based in the US and thus subject to their laws
lol maybe vfx, animators and game devs should lobby to have more ethical labor laws for themselves before they worry about ai
The US gov can also lobby the UK to follow after them, or they might even just make the standard copycat legislation that's all too common these days
indeed
I see. Well, if they want to lobby for more ethical AI restrictions it's their right to do so, i just feel like in the end corporations have a whole lot of sway in the US wrt copyright and IP and the like.
Also SD doesn't depend on discord like midjourney does.
like you can literally see people doesn't care about people all they want is power and control if they care about people you wouldnt have vfx and all those toxic work culture around
I think some of the things they want to try to lobby for are fine there, they don't want an outright ban.
I just don't know how realistic it is.
well I hope chatGPT can be the lawyer
XD
like the last 3 clause in the whole thing is just straight up virtue signalling
just said you want to control AI with your organization
I think if stuff like SOPA didn't pass in the US govnt, anti ai stuff has very little chance.
Yeah I just hope those training datapool can limit data on what they use for training
you can't mix cement with calcite because they just lack of light blending
light reflection
so i hope people who dev this can understand
Hey all. A bit of a thought experiment I was mulling over while considering the legality of using copywritten material in training data. Would it be legal to fine tune a model like gpt-neo on a collection of legally purchased research papers? If so, could such a model be used to extract data that was previously behind a paywall?
technically yes
if I remember American court ruled that no copyright over AI generated material
so technically speaking yes
Even if such a model was served through a paid API? For instance say I create this model and charge people to ask it questions about the papers the model was trained on.
put everything I just said there?
yeah
guys i heard emad posted that he is making a text generator similar to chatgpt
that can run on our home computer
i cant find the post
not gonna lie but emad had create so many enemy I'm surpruse he is alive
That was a bit of hyperbole for sure. The guys at Eleuther AI are working on smaller models that can outperform GPT3 using RLHF but there are no guarantees.
outperform gpt3 or gpt3.5 i wonder
thanks for info i need to look more into that
and i wonder how many paramters it would be able to use
The specific eleuther team falls under carper.ai
That should help with your search
thx ❤️
someone needs to do it because looks like openai is censoring and watermarking everything and will charge a lot of money
looks like they will release it as open source though
how are they recieving the funding to do this
and yeah looks like wont run on personal PC
if thats the case hopefully other companies will take up the torch and give us generator without restriction like novelai
since they have the hardware to run big parameters
I think right now novelAI uses 20 billion parameters for its highest generator
(krake)
but i wonder if NovelAi will continute to let us get infinite generations, or if running like 100 billion parameters will make it too expensive for them 🤔
Dude Riffusion is NUTS. lol
what is riffusion
oh music generator
i wonder if it actual makes original music or just loops clips together
and how good it is compared to muberts
I like the play on words, Riffusion
I'm watching the AI vs artists stuff play out over social media and as someone who basically stands in the middle it's very entertaining to watch
Like... We can walk and chew gum at the same time lol
So, LORA .pt files, as they exist, can only be merged into CKPT files? There's no way to create an embed from them?
Really? I didn't find it THAT impressive
The concept is mind blowing
It's the fact of where we can take this given enough time that I'm really connecting with. Early stages, but you never know.
The result is still V1.0
I mean seriously, Stable Diffusion creating spectrograms to use for audio creation is quite genius.
The rate at which AI is accelerating at is absolutely mind-blowing.
Just THIS year alone, look where we started and look where we are now.
I guess I'm numb to the concept due to how many other models are converting input to images and running it through a diffusion model. One example being a model that identified the origin of a sound based on the waveform. Another interesting one was an actuary who graphed data used to generate risk assessment and got a better result through diffusion.