AI Art

I don't see any logical rationale for this. AI isn't like 3D CGI rendering supplanting hand drawn 2D. Its the same sorts of art with the same expectations for detail, composition, handling of lighting and shadows, character expressions, etc etc. All the intellectual aspects of art are still there. Being able to draw on a tablet is maybe less valuable, but drawing is still usable and important for both the input and output angles.

I'll pose a question to you. How many of the skills used in AI art do you think would be translatable to architecture or industrial product design?
 
Last edited:
and I also think that their knowledge of copyright and derivatives is incorrect.
I mean in many ways this feels utterly irrelevant? Like copyright law is in most of its specifics simply a deeply bad and unfunny joke, and some people who complain about the AI utilizatizing artists' works may frame it as a legal issue, but i'd say to most of them at its core it is a more philosophical and moral one. Like even if you made an airtight argument that it was entirely legal under copyright law, do you think you'd actually change those people's minds- speaking even much more broadly than just here on SV Dot Com?

I do think that Baughn's point is interesting, though I would argue the core conclusion as far as their point in that specific post goes has more to do with the US copyright system and a lot of those systems more broadly being fucked than what most people in this thread honestly disagree about.
 
I mean in many ways this feels utterly irrelevant? Like copyright law is in most of its specifics simply a deeply bad and unfunny joke, and some people who complain about the AI utilizatizing artists' works may frame it as a legal issue, but i'd say to most of them at its core it is a more philosophical and moral one. Like even if you made an airtight argument that it was entirely legal under copyright law, do you think you'd actually change those people's minds- speaking even much more broadly than just here on SV Dot Com?

I do think that Baughn's point is interesting, though I would argue the core conclusion as far as their point in that specific post goes has more to do with the US copyright system and a lot of those systems more broadly being fucked than what most people in this thread honestly disagree about.
regardless, I've seen artists willing to go to court so the infringement definition has consequences. They're genuinely using copyright law to say that the model is derivative.

If copyright was fucked, it's in the other direction not in the direction that says its weak. I've never heard copyright was too weak from people until after AI Art. People and families were sued over copyright laws being too strict from massive corporations like UMG and nobody thought to expand copyright in that moment.
 
Last edited:
I'll pose a question to you. How many of the skills used in AI art do you think would be translatable to architecture or industrial product design?

How many skills you learn from drawing anime translate to architecture or industrial projects? How many of them translate to software development?

Like, this is entirely nonsensical argument. If you are artist, that doesn't magically make you capable of architecture or industrial product design. Hell, it might actually bad for you since you are likely to have certain biases.

You know Red Faction Guerilla? They had to hire actual architects to design their buildings because, as it turns out, being a 3D modeler does not mean you know how to make structurally sound buildings. Their in-house buildings just kept collapsing because the engine simulated actual structural weaknesses.
 
@firefossil If you have a moment, I'd like some advice on how to get better. Right now I'm trying to generate a fantasy town from TTRPGs setting. The idea is that they live in futuristic longhouses (Iroquois styled) Surrounded by wooden stockades.

I'm using 35 steps, DDIM, with Dreamshaper, and I'm using this for my prompts:
Positive: Skyscrapers made of glass windows surrounded by a tall stockade, in a wild west field. #Fantasy #RPG_Art #High_Quality #4k #Digital_Painting
Negative: #deformed #Bad #Ugly #Distorted #Poor_Quality #Photo

And I'm using the following two pictures for controlnet, the second one as my first model to generate a depth map, while the first one is being used for segmentation in my second controlnet.

The stuff I'm getting out is okay, and I like some of it, but some of the lines are deformed, and I'm not sure how to go about fixing that. How would I go about going the last 20% to make have these come out properly?
Here are some examples of outputs that are close, but just off:

I like 92 the best, but I'd like the central building to have a roof closer to 78 or 84. Meanwhile I'd like 92 to have the roofs for the other buildings from 107.
 
I can actually read it, and it's still not music. It's instructions. How to play an instrument, what notes to play. It's not music. Music is the audio generated. However, we are humans so we tend to be lazy and just skip to calling it music becuase it's easier.
Except that the entire tradition of jazz music acts as a commentary on how the sheet music is the map, not the territory. You are using the fact that sheet music can be colloquially described as "instructions" as the cover for a bit of rhetorical legerdemain fabricating a direct, absolute similarity between sheet music and the prompts fed into an AI art generator.


Yes, his job is to instruct people on what they need to do. He needs to tell guy in charge of lights how he wants lights to be set. He doesn't build the lights himself or adjust them. He tells other people. He doesn't act in front of cameras, he has other people doing that as per his instructions. He instructs editors how they should edit, how the timing goes... he goes around instructing people what he wants to achieve.
On a fundamental level, even the most authoritarian, megalomaniacal director has to accept at least a minimum level of feedback from the other people working on the film - the reason why there are lighting technicians, an SFX crew, makeup, choreographers, et al is because it allows for the production to incorporate a variety of different specialist perspectives that no single human being could ever hope to compile within their own psyche. The idea that a director must maintain such a granular level of control over every aspect of production - much less that he would be given such broad authority! - is farcical.

You seem to understand film as a medium wherein the director controls everyone else like automata, which is neither the norm nor a particularly effective means of creating good movies*. The original Star Wars trilogy, as a piece of creative work, was influenced by its matte painter and its costume department head (among others), not solely by the director, or even the director and the writers.

George Lucas was then given absolute control over the prequel trilogy due to the kind of Great Man Theory bullshit your post exemplifies, and we can see how the finished product had significantly more rough edges and untrimmed flash than it could have had under a less auteur-minded production model.




* In fact, it's literally killed people, and a lot of the people who are celebrated as auteurs, such as Stanley Kubrick and Alfred Hitchcock, wielded the dictatorial power they were given in an utterly disgraceful fashion. I haven't found any reason to believe that their abuses somehow imparted a level of quality to their films that couldn't have been achieved through other means.

Your mention of editors, specifically, reminds me of how the auteur behind cinematic bomb Heaven's Gate literally hired private security to roust the editors from the editing room and did all of the editing himself, which contributed to the film being a bloated, unwatchable mess.
 
How many skills you learn from drawing anime translate to architecture or industrial projects? How many of them translate to software development?

Like, this is entirely nonsensical argument. If you are artist, that doesn't magically make you capable of architecture or industrial product design. Hell, it might actually bad for you since you are likely to have certain biases.

I never said being a good artist was necessary to be a good architect or that it automatically conferred the ability to be a good architect. I simply asked how many of those skills from creating AI art would translate.

But I'll answer the question you posed in turn. Drawing can help to train your sense for the human body's proportions, and the scale of the human body in comparison to other objects, that's one example. Understanding of how lighting affects a space. The ability to visualise a setting or space in three dimensions. Of course, it's not like every artist develops these skills either, but these are some examples, since you wanted them.
 
Last edited:
I never said being a good artist was necessary to be a good architect or that it automatically conferred the ability to be a good architect. I simply asked how many of those skills from creating AI art would translate, and I notice none of the responses have actually tried to answer the question instead of deflecting.
I mean, wouldn't the 3d modeling I'm learning as part of being able to make more accurate prompts carry across?
How was I deflecting? What more of a response were you looking for?
 
Rule 3: Be Civil - 'Play the man not the ball'
Look, I'd be a lot more open to @Mandemon's argument's if he hasn't repeatedly shown himself to be uh really, really bad at discussing and understanding aesthetics.

Like sorry for not thinking the guy who literally believes subtext doesn't exist has the answers when it comes to philosophical discussions about art and technology.
 
Last edited:
I mean in many ways this feels utterly irrelevant? Like copyright law is in most of its specifics simply a deeply bad and unfunny joke, and some people who complain about the AI utilizatizing artists' works may frame it as a legal issue, but i'd say to most of them at its core it is a more philosophical and moral one. Like even if you made an airtight argument that it was entirely legal under copyright law, do you think you'd actually change those people's minds- speaking even much more broadly than just here on SV Dot Com?

I do think that Baughn's point is interesting, though I would argue the core conclusion as far as their point in that specific post goes has more to do with the US copyright system and a lot of those systems more broadly being fucked than what most people in this thread honestly disagree about.
regardless, I've seen artists willing to go to court so the infringement definition has consequences. They're genuinely using copyright law to say that the model is derivative.

If copyright was fucked, it's in the other direction not in the direction that says its weak. I've never heard copyright was too weak from people until after AI Art. People and families were sued over copyright laws being too strict from massive corporations like UMG and nobody thought to expand copyright in that moment.
My perspective is IP infringement should ban only things that do not add value. Someone sells a bootleg of your book that matches the original word for word, they aren't adding value, so society has every reason to stomp on them. I'm ok with banning plagiarization like that. Someone writes an original book that has a cover and authorial pseudonym designed to trick people into thinking its your critically acclaimed work, they might technically be 'adding value' but are still mostly trying to mooch off of you. I'm ok with banning or at least restricting that sort of trademark infringement.

Someone writes a fanfic of your story but does not literally plagiarize it nor falsely advertise their work as your own, that strikes me as a transformative work that has value. It does not infringe on your value in any way other than rejecting your right to have a government enforced artificial monopoly on an idea that makes it your personal property to rent-seek off of. Since valuable IP is by and large owned by wealthy capitalists, this fencing off of ideas does not benefit the average creator. I think the ability legally sell fanart of major IPs is far more valuable to struggling indie artists then the ability of them to legally monopolize their own IPs on the incredibly low off chance said IP becomes wildly popular.

I know that artists tend to be uncomfortable with this idea. Still though I think backing modern IP law in the hopes you'll be the next Harry Potter is like backing tax cuts for the rich in the hopes you'll be the next millionaire, rather than accepting high taxes on the rich as necessary to provide welfare for the poor, most of whom will never have a chance to be rich.

I'm using 35 steps, DDIM, with Dreamshaper, and I'm using this for my prompts:

And I'm using the following two pictures for controlnet, the second one as my first model to generate a depth map, while the first one is being used for segmentation in my second controlnet.
If you are using dreamshaper and controlnet you've already advanced beyond my level of expertise. I've been putting off learning them because I figure in a matter of months there'll be more powerful more user friendly versions of them anyways. Then again perhaps now would be a good time to take a look? You wouldn't happen to have a link to where to download those two or guides for how they work?

I'll pose a question to you. How many of the skills used in AI art do you think would be translatable to architecture or industrial product design?
I'm not a fan of leading questions. Its obtuse and gunks a thread up. Make your point or answer mine.
 
Last edited:
That traditional forms of art require and help to train up skills that AI art doesn't, because you're delegating a significant portion of the process?
I never said being a good artist was necessary to be a good architect or that it automatically conferred the ability to be a good architect. I simply asked how many of those skills from creating AI art would translate.

But I'll answer the question you posed in turn. Drawing can help to train your sense for the human body's proportions, and the scale of the human body in comparison to other objects, that's one example. Understanding of how lighting affects a space. The ability to visualise a setting or space in three dimensions. Of course, it's not like every artist develops these skills either, but these are some examples, since you wanted them.

Does traditional forms of art make you better programmer? Better at designing algorithms? Give you skills you need to understand how to parse data? No?

AI art does, as you need to understand how these things work in order to get results. So there is skill that can be transferred. A skill you develop. It's just not exactly same as with traditional arts, but it works both ways. Traditional arts do not teach skills that AI art does.
 
If you are using dreamshaper and controlnet you've already advanced beyond my level of expertise. I've been putting off learning them because I figure in a matter of months there'll be more powerful more user friendly versions of them anyways. Then again perhaps now would be a good time to take a look? You wouldn't happen to have a link to where to download those two or guides for how they work?

I'm guessing you already have A1111. I'm using the Controlnet with Paint With Words extension right now, though I'm not too impressed with the Paint with words part of it. It does have a few alterations to the base controlnet that I find useful, specifically allowing me to mess around with the resolution it extracts depth maps at. That said, the base controlnet version is much easier to install, with a clearer walk through.

Also, while in A1111 make sure you check the available extensions from the tab. There is a bunch there that plays well with Controlnet, including things like a 3d character image poser made to export straight into controlnet.

If you have any questions or such, feel free to ask. I suffered through this over the last 2 days, so it's still fresh in my mind.

Edit: Right now I strongly recommend using segmentation over Paint With Words. PWW is harder to use, and gives less useful results, and I'm not sure how well it even plays with multiple controlnets.
 
Last edited:
Look, I'd be a lot more open to @Mandemon's argument's if he hasn't repeatedly shown himself to be uh really, really bad at discussing and understanding aesthetics.

Like sorry for not thinking the guy who literally believes subtext doesn't exist has the answers when it comes to philosophical discussions about art and technology.
I have to repeat again that this is the whole problem here—you shouldn't have to have any understanding of aesthetics to discuss how society and the law ought to treat art and technology, because a foundational principle of liberal pluralist society is that aesthetics must be segregated into the private sphere and should not matter a whit to public morality and law.

As far as the law is concerned, the question of how to regulate art is entirely a matter of politics, not of aesthetics—what matters is that the art that is produced, the process by which art is produced, and the way in which the art is received is politically and philosophically acceptable, and the purpose about any law regarding artistic production is to promote this. To this end, one needs a firm understanding of art criticism, the skill of analyzing the messages and meanings that art is trying to convey—and one can be justly faulted for lacking such an understanding, as in a case of failing to perceive subtext—but an understanding of art and aesthetics itself, the eye for beauty, an artist's experience, etc., is completely irrelevant. Law is of the empire of literacy, and it is literacy and literacy alone that ought to guide it—it's concerns are only for the just, the moral, the necessary, the permitted, and so forth, and no one is or should be allowed to smuggle aesthetic considerations into legal decisionmaking under the ostensible guise of "mere" economic regulation.
 
Last edited:
That traditional forms of art help to train up skills that AI art doesn't, because you're delegating a significant portion of the process?
I don't see why that would be the case? If you want AI to, say, do lighting and shadows right, you need to know what 'right' looks like. Preferably well enough that you can give instructions in advance to whatever AI modifiers would control that so it does it reasonably close to right the first time, rather than having to constantly redo and rehash it in image2image. If you don't, you'll probably just have a vague sense of "this doesn't look quite right" with no idea how to fix it.

If anything I'd think AI would help teach people by allowing them to do quicker mock-ups to get a sense for things like "does this pose from this camera angle look good?". While different from going to a formal art school to teach you the underlying principles for such composition, learning through doing is still learning.

I'm guessing you already have A1111. I'm using the Controlnet with Paint With Words extension right now, though I'm not too impressed with the Paint with words part of it. It does have a few alterations to the base controlnet that I find useful, specifically allowing me to mess around with the resolution it extracts depth maps at. That said, the base controlnet version is much easier to install, with a clearer walk through.

Also, while in A1111 make sure you check the available extensions from the tab. There is a bunch there that plays well with Controlnet, including things like a 3d character image poser made to export straight into controlnet.

If you have any questions or such, feel free to ask. I suffered through this over the last 2 days, so it's still fresh in my mind.
My model says "Easy Diffusionv2.5.38" at the top, I don't think its the same. The actual downloaded folder for my version of SD is titled stable-diffusion-ui-win64 whereas your link is for stable-diffusion-webui-1.2.1, a download which itself seems to require me to download several other things like Python and Git. It'll take time to figure out. Thanks for the link though! Good to know where the latest version is at.
 
My model says "Easy Diffusionv2.5.38" at the top, I don't think its the same. The actual downloaded folder for my version of SD is titled stable-diffusion-ui-win64 whereas your link is for stable-diffusion-webui-1.2.1, a download which itself seems to require me to download several other things like Python and Git. It'll take time to figure out. Thanks for the link though! Good to know where the latest version is at.
No problem. A1111 seems to have the most cutting edge tech in it, but it's also only one step up from having to learn python to get running at times. I hope a most simple install method for it comes along soon.
 
I have to repeat again that this is the whole problem here—you shouldn't have to have any understanding of aesthetics to discuss how society and the law ought to treat art and technology, because a foundational principle of liberal pluralist society is that aesthetics must be segregated into the private sphere and should not matter a whit to public morality and law.

As far as the law is concerned, the question of how to regulate art is entirely a matter of politics, not of aesthetics—what matters is that the art that is produced, the process by which art is produced, and the way in which the art is received is politically and philosophically acceptable, and the purpose about any law regarding artistic production is to promote this. To this end, one needs a firm understanding of art criticism, the skill of analyzing the messages and meanings that art is trying to convey—and one can be justly faulted for lacking such an understanding, as in a case of failing to perceive subtext—but an understanding of art and aesthetics itself, the eye for beauty, an artist's experience, etc., is completely irrelevant. Law is of the empire of literacy, and it is literacy and literacy alone that ought to guide it—it's concerns are only for the just, the moral, the necessary, the permitted, and so forth, no one is or should be allowed to smuggle aesthetic considerations into legal decisionmaking under the ostensible guise of "mere" economic regulation.

"As far as the law is concerned, the question of how to regulate AI is entirely a matter of politics, not of programming—what matters is that the AI that is produced, the process by which AI produces content, and the way in which the content is received is politically and philosophically acceptable, and the purpose about any law regarding artistic production is to promote this."

Do you see how you can dress this same argument in the same fake lawyer suit and flip it around to AI? Because I think we'd both agree that if I said "I don't need to understand how the AI generates this images, merely that it does, and I have moral, political and aesthetic questions about that" you and many others would roast me to kingdom come.

So yeah, when someone is making sweeping arguments about all of art, when they're constantly framing things with a bizarre "us vs. them" mentality where artists are snooty effete snobs keeping out the oppressed blue collar AI artists, I'm gonna roll my eyes because that same person (and many others) literally don't understand how art functions on the most basic level. Thinking subtext is a plot by leftists to be mean over the internet is what children think, so I'm gonna look sideways at anything that follows after that.
 
Alert: Action taken, thread reopened.
action taken, thread reopened.

If you are going to argue over this I am going to stress very clearly:

Rule 3: Be Civil

Be courteous and decorous to everyone you deal with. It is possible to disagree, even vehemently, with people (both on and off the site) in a constructive fashion.
  • Respect the participation of others. Attack arguments, not people. Don't post personal attacks or insults, or join a thread for the purposes of attacking someone else.
The Fine Print
[...]

At its most basic, incivility is trying to crowd out others. Personal attacks, insults, extreme condescension, threats (particularly those relating to the report system or the ignore system, which are often grandstanding) and a feeling like their private thoughts will be used against them all push users away from Sufficient Velocity, and are not tolerated.

This doesn't mean a vigorous debate is prohibited. But it means you need to "play the ball and not the man". Posts need to focus on opinions, positions, and ideas - and not on other users.

---

I have issued multiple Rule 3 infractions as well as a Rule 4 infraction - if you wish to debate the merits of Traditional Art vs AI Art it is imperative that you do not descend to personal attacks. This is well known to be a thread where tempers flare, it will also be here tommorrow and likely the day after. This debate will last further into the future than any of us probably care for. Take your time when posting and consider the arguments directly.

Arthur Frayne, Flibspeak and ChineseDrone have all recieved 72 hour threadbans.

 
If someone came up to me and asked how learning scales translates to making EDM or something I'd have an internal chuckle at the least. We've had this unproductive discussion about the importance of process already. If you don't get it you don't get it, but you can't also dismiss it as unimportant if you don't.
 
I have very, very personal reasons for being pro-ai. And I know people on both sides of the argument got banned here, but seeing people in charge and against it also get in trouble restored my faith in this website when I really needed it.

I will work to argue in as good faith as I can on a very personal topic for me here.
 
I don't see why that would be the case? If you want AI to, say, do lighting and shadows right, you need to know what 'right' looks like. Preferably well enough that you can give instructions in advance to whatever AI modifiers would control that so it does it reasonably close to right the first time, rather than having to constantly redo and rehash it in image2image. If you don't, you'll probably just have a vague sense of "this doesn't look quite right" with no idea how to fix it.

You are literally arguing that one can develop artistic skills purely through observation. Which is just not how it works.

Does traditional forms of art make you better programmer? Better at designing algorithms? Give you skills you need to understand how to parse data? No?

AI art does, as you need to understand how these things work in order to get results. So there is skill that can be transferred. A skill you develop. It's just not exactly same as with traditional arts, but it works both ways. Traditional arts do not teach skills that AI art does.

People aren't declaring programmers obsolete over this or pushing for them to be obsoleted, though. If anything they're going to be even more in demand.
 
Last edited:
You are literally arguing that one can become a skilled artist purely through observation. Which is just not how it works.



People aren't declaring programmers obsolete over this, though. If anything they're going to be even more in demand.
I would say on AI art and traditional art that the amount of mediums and skills 'art' already falls over means that I really just see AI art taking a place amongst the rest of them.

Maybe it's harder to see its skills directly translate - but then I think that's a matter of how you employ it. Things like Adobe's new AI tool for PS can just be a more advanced editing tool for photography and tools like Inpaint can be the same to a digital artist, letting them focus on the general structure say of a form while utilising Inpainting for details within it.
 
Back
Top