30001st poster gets a cookie (cookie thread (Part 7))

[Daeron]

I would need to see data to evaluate this

[May]

And I suppose where our viewpoints differ… one, I don’t think a firm distinction exists between the AI “assistive” tools and “generative” ones, there’s a spectrum from upscaling to object removal to full on generation, they’re all part of the same object to me, they help bring images from your mind to the canvas quicker. The fruit of the tree (waifu2xing lowres images so my avatars are less crunchy) is good and therefore I love the tree. I don’t see the difference between the shortcuts of “drawing something and then using AI to make adjustments and finish it up” and “prompting the AI in to create the base image in the first place and then further adjusting it to create your vision”. Same thing to me.

The second point of disagreement: the AI art I love and find interesting is inseparable from the human decisions that make it. Have you looked at the links I keep plugging? The people in those links are doing art. The first person is using AI to create comics, they have precise character designs they’ve developed through iteration, they write a script for a panel, they prompt Midjourney for each individual character and splice them together, of course they’re thinking about compositions and storytelling and what have you.

You’re drawing a line around a part of this tech fundamentally inseparable from it, excluding all human control from the definition of AI art, and saying “well it has no human control in it so it’s bad”. That’s a bad argument.

The people using this technology at an advanced level want as much control as they can get, that’s the space most people are looking to innovate in, that’s where the skill expression comes from. Controlnet! Image-to-image! Great stuff! Inseparably part of the medium of “AI art”.

You appear to think, to my eye, that I am primarily arguing those random-ass search results that prompt an image gen for whatever you searched for are Good Art. They’re bad art. I don’t like them. They’re just spambots. People who type a random prompt into midjourney and take the first image they get are not Good Artists, like people who just point their phone camera at something with no thought for composition and hit the button are not Good Artists.

My point is not “the lazy AI art you know and familiar with is good and not bad”, my point is “this technology is more advanced than you give it credit for and there are lots of smart and creative people creating good things with that you don’t know about, that aren’t in your internet circles”. That’s why I’m constantly posting AI art I like here, I want to expand people’s horizons a bit.

The third point of disagreement: yeah an hour and a half isn’t long for a digital painting, I presented that as impressive not because it’s good for art in general but because it contradicts the idea that AI art is only typing a prompt. You are explaining this shit to a digital artist, I have more experience with this than you! Making fairly simple/small Flight Rising accents can take me like six to eight hours, and that’s drawing little accents on a base at a very low resolution, normal art takes sosososo long. That’s why I’m so excited about technology that can speed things up!

I love the part of the artistic process where you decide what to make, I love sketching compositions and deciding on initial rendering styles… and then I gotta repeat that rendering over and over again for every stupid fucking rose I sketched. That’s not part of the fun and thoughtful and creative art process to me, that’s gruntwork that just makes my wrists hurt. The majority of the time spent on art is often that kind of, like, repetitive tasks, more weaving cloth than designing clothing, and it isn’t where the mental and artistic decisions are all made.

Incorporating AI into normal artist workflows speeds that boring shit up. You brought up Spiderverse, great example. I just take it farther than you: I think it’s also artistic and ok to just draw flat colours and then tab through rendering styles and pick one you like for the drawing, I think it’s also artistic and ok to just sketch a composition and let the image gen fill in the rest, I think it’s also artistic and ok to just write a prompt describing what you want and start from there. Less impressive in many respects than super high-effort manual digital paintings? Sure! Not art? I disagree, it’s still art.

[Atlas]

…rich fuckers?

[notblackorwhite]

There’s basically never a point where it’s economically viable because it lacks scarcity. As supply approaches infinity, demand goes to zero. A hypothetical company that produces AI models could never afford to singlehandedly pay artists to produce a better model than the ones we currently have because the ones we currently have were trained on a significant portion of their medium’s corpus. The art it produces is fundamentally never valuable and each generation of the model that produces better and better images from the jump would just devalue itself further and further.

[Atlas]

they still value having like five swimming pools you just hear it less nowadays

also nobody new is getting like five swimming pools because all the older people have wealth

[May]

Like, I also held similar opinions re: AI art when I knew less about it, never as extreme but you can see old posts of mine on here like “it’s not good art because it’s always derivative”. It was learning more about 1. the extent of control available within the technology and 2. what artistically-inclined people are doing with the technology (rather than people who want to skip out on actual art) that changed my mind

[Daeron]

if they ever kept making new iterations such that there was a perfect or near perfect model, then yes artists will disappear
that being said if the model really was that boundless then I would consider ai imagery as art, and the prompt engineer as the artist

and then your horses and cars comparison becomes relevant. except in the horses and cars case, horses still exist. artists just wouldn’t exist (or there would be very little “traditional” artists.)

I think this is a net positive outcome and therefore a moral one

[Italy]

i’m pretty sure the main controversy regarding ai art is the fact that most of their reference images are taken without consent

[ElizaThePsycho (Willow)]

Holy shit

[ElizaThePsycho (Willow)]

This sounds suspiciously like an RP…

[Daeron]

never

[Italy]

which is also why the “poisoning” movement started where artists would rig their images to corrupt any machines using them as a reference

[Arete]

the boundary between text-based roleplaying and text-based collaborative fanfic is thin

[May]

Well that’s what a lot of people say it’s about but there’s also a toooon of arguments about Artistic Souls out there, it’s clearly not all plagiarism, also there exist models which draw only from public domain images and people still hate those and don’t make exceptions for it, so that’s clearly not the whole issue

[May]

Most tools to do that just straight up don’t work lol

[May]

Like, here’s the set of control tools available in the Stable Diffusion for Krita thing I downloaded and messed with and was really bad at using:

This plugin is for an existing digital art tool, so it’s designed from the “Spiderverse perspective” up - the base idea is “you can use AI to speed up your existing digital art workflow”, and the tools for more intensely “pure AI generated” images are built on top of that

Top to bottom, that’s:

1. Providing a general reference image - you can, for example, put a character model sheet in this and then prompt for that character and their features will generally be transferred over, I’ve messed around with this to see how good it is at making MajSoul characters (it works well in terms of getting the literal features down but all my images are so ugly in style cause I’m bad at prompting :’()

2. Providing an image for style reference - say I’ve made a painting myself and I want to create another one in the same style, I can do that here (or I can make an image in the style of another artist, but I do not think that is plagiarism! You can do this by hand, and it’s a very common artistic exercise!)

3. Providing a composition - exactly what you were saying AI art doesn’t do

4. Providing a face for reference to keep faces consistent (very useful for fanart or comic art)

5. Scribbling down a drawing and having that completed

6. Putting in lineart and having that coloured

7. IDK what soft edge is lol I’m bad at AI art. Maybe just drawing out where soft edges are?

8. IDK what canny edge is lol I’m bad at AI art. Same thing but with sharper edges?

9. Providing a depth map (e.g. from a 3D model), similar to composition stuff but even more precise

10. #normal (I am bad at AI art)

11. Posing a movable doll whose joints you can drag around and having the generator create a character in that exact pose (this is SO SO COOL)

12. what is a segment help

13. Unblur just unblurs your art which is very nice to have

14. Stencil fills in a stencil (obviousl)

15. fix hands pleas

This is a pretty basic AI plugin. It’s just some shit I downloaded for free for a free art program, it’s not especially cutting-edge. The people creating art with this care about the things you talked about, they care about character design, they care about composition, they care about style choices. They’re just using different methods to make those decisions. It’s art! It’s just art.

[ElizaThePsycho (Willow)]

I do love it when people bring back like, random practices for like, Ancient Greek gods, or something
I don’t think anyone actually believes that Zeus exists but the rituals are fun. I mostly think this because it’s a giant “fuck you” to Christianity, though, and have not thought deeper about this

[tutuu]

do u guys wanna do fol roleplay?

[May]

Apologies for being BAD AT EXPLAINING them there’s so little documentation about this shit (this is why I’m bad at using it)

[ElizaThePsycho (Willow)]

Ok so I can call it basically roleplaying to anyone who asks, got it