So this is how 19th century artists felt when photography was invented, lol.

Honestly though, i'm all for it. I think this has great potential to making art not only more convenient but also much more accessible in general. It will save you so much time. It's the next step of after effects in a drawing program, or automated tween animation in Flash. Heck, think Xerox animation! No more inking of line-art, just let the machine do it.

It is far from being a perfect tool right now however. You can still see the limitations of the AI. I dabbled in Midjourney until my free trial ran out, I've gotten some amazing results and some that didn't work out at all. Pixel art is beyond it for one thing, couldn't get my cyberpunk 64bit megacity to look right no matter how i tweaked the prompts.

What this is really great for in its current state is getting inspiraion for your own art. Since most images turn out only halfway right (especially faces are too much to handle for the AI), the Uncanny Valley is staring you straight in the eye… No one's stopping you from taking the generated image and re-draw it in your own style though. I've seen some artists do exactly that. While under closer inspection you can tell that something is off about most pieces, the general mood of the lighting/color will often turn out fantasically.

On a personal level… If the tech gets more refined, what i would use it for is drawing backgrounds for example. I'm super lazy on backgrounds, i hate it frankly. If i could just tell a machine to do it for me, dam straight i would.

And i doubt it will make artists obsolete any time soon. The invention of photography didn't stop people from drawing portraits or landscapes either. And as i understand it, the AI does need images drawn by actual artists to pull from.

So i for one welcome our new Skynet Michelangelo overlord.

On that note, Stable Diffusion is an open-source Dall-E that lets you do a lot more things, and will also do NSFW, so look into that when you can.

Now, onto the subject at hand, as it is, unless you have absolutely no artistic skill, it oftentimes takes longer to get acceptable results out of these neural networks than it does to just… Make the thing yourself.
And oftentimes the result won't be exactly what you're going for.

I was looking into neural networks for backgrounds, whether it was viable, mostly out of curiosity, right?

So I tried getting it to do huts, manga-styled pages and things like that.
Unfortunately, it often messes up the horizon line and perspective, and if you do get an acceptable result (after over 30 minutes) it's oftentimes not in the right perspective to fit into the work you're doing, so either you'd adapt your work's perspective or start over.

Oftentimes there's also inconsistencies that any skilled artist would immediately notice, and it tends to just not fit in in, say, comic contexts, unless they were made from the ground-up with neural networks in mind.

Due to that, it's fairly limited in terms of comic contexts, or in contexts where the neural network's art needs to be expanded upon with other elements.

So unfortunately, I don't think it's ever going to get to a point where it's perfectly artistically viable.

It'll most likely end up used heavily in more commercial settings that most artists worth their salt would have disdain for, the kind where photography took over, and where stock photos reign supreme.

Due to that, I'm FAR from worried.

GeekyGami wrote:
On that note, Stable Diffusion is an open-source Dall-E that lets you do a lot more things, and will also do NSFW, so look into that when you can.

I gave it a try, doesn't give nearly as good results as Midjourney gave me though. I tried the same prompts like “Vintage photograph of a unicorn”, Midjourney delivered a deceptively convincing result. Diffusion's just looked like a badly rendered CG horse with a vintage filter slapped on it.

What i found interesting is that Midjourney worked much better the simpler i made the prompts. Only 3-5 words at best. The more i tried to narrow it down, the less it looked like what i wanted. Like i said, it's great for getting inspiration if you don't know how to start on a piece, but if you want anything specific ya still gotta pick up the pen yourself.

It all looks like crap. I hate robots! call me a robophobe, I don't care. I go out of my way to avoid shops that use self checkout. Hate that shit. And I hat vocoloids too. And all the algorythms and crap making the whole world into normy bullshit. I kinda hope the sun fries the electrical grid and we can all go back to the 70s. I wanna listen to the radio and play board games and eat pizza in a restaurant with the pitchers of root beer.

It's absolute garbage. Which means hordes of people will love it and in five years the number one comic on the web will be done completely by AI.
Even now the art is more original looking than the majority of what passes for a webcomic page.
People love garbage and don't know any better than to love garbage.

Either way I'm gonna keep drawing manually for as long as I'm able, out of principle. The principle being that we human beings are meant to use our mind, the ultimate computer, and our body, the ultimate technology. Use our powers, our gifts. It is by the same principle I still use mental arithmetic from time to time instead of always relying on a calculator to do the math for me.

Why I still use an old-fashioned bicycle, pedaling 3 miles to work every weekday and would never even wanna touch one of those power driven scooters that I see people use and then leave in a ditch somewhere

It is self-destructive in my opinion to always let machines do everything for you, because it is in my opinion to waste your own power. I say, may the machine always malfunction, and force us to take and make action of our own for all time.

TheJagged wrote:

So i for one welcome our new Skynet Michelangelo overlord.

The important question: Will it be pronounced “Robo-Mike-Al-Angelo” or “Robo-Mik-kel-An-Gelo”?

lothar wrote:

I kinda hope the sun fries the electrical grid and we can all go back to the 70s. I wanna listen to the radio and play board games and eat pizza in a restaurant with the pitchers of root beer.

@Lothar- I am surprised! Your comic has always been one of the most futuristic ones here for as long as I can remember.

==

I would embrace comic-making AI.
It turns out my username “kawaiidaigakusei“ (and real name “claire”) both have “AI” in them (three to be exact).

I wouldn't worry too much. Art survived photography and the movies survived television. Not to mention we still have books (which you may have seen on television).

52 years ago I first picked up a Hunt 102 pen nib, which today is garbage. If you think a Zebra G nib is good I feel sorry for you.

I've been using a Wacom tablet for over a decade now. While I've seen some very fine digital art I've also seen the charm beat out of pieces because the creator thought the program made their work look better than it really does. It's more the person behind the tool than it is the tool itself.

You could give me the most expensive, up to date digital equipment and some one like Ironscarf would, to paraphrase Scott McCloud, outdraw me using a Snicker's bar.

Apropos of nothing but I keep thinking about that one scene of I Robot… “Can a robot turn a canvas into a beautiful masterpiece?” We're real close to finding out lmao.

Unka John wrote:
It's more the person behind the tool than it is the tool itself.

You could give me the most expensive, up to date digital equipment and some one like Ironscarf would, to paraphrase Scott McCloud, outdraw me using a Snicker's bar.

That is exactly why i find AI art exciting. Even now these algorithms produce better looking art than i could in my entire life with my meaty hands. What if i have an idea for this epic peice, but i'm simply incapable of getting it to look right? Just let Daddy HAL9000 handle it. Feed it the keywords, the style references, and presto, you got the piece exactly how you want it.

One super-dooper major flaw in all AI “art” I've seen (and one that non-artists never understand) is its total lack of design language.

The AIs are like kids who have physical artistic muscle skills and great ability to copy but no understanding of what they're copying and why the things look the way they do in the first place.

i.e Tell it to draw a space warship and it'll give you a generic grey blocky thing on a space background.
And fair play there are a lot of not very good artists out there who will give you that sort of thing…
A good artist wont though- They'll break down what you mean by “warship” and try an understand how the thing would work based on the type of propulsion it uses, the type of weapons it needs for the types of battles it supposed to fight (is is a solar system ship or interstellar or intergalactic?). What kind of race would have built it- were they aliens or humans? What are the cultural influences that directed the design? etc.


This applies to ALL speculative art, even to things like costumes and music. Bad designers and AIs will only ever be able to give to derivative pastiches with no understanding of why things are they way they are, so what they make will always be a bit bland and generic, or even down right silly.

The trouble is that while AIs are good at simple tasks, they're all stupid at the base of things- They can't understand why they're doing and they're limited by their programming. Which is why we get so many messups with big companies that have replaced expert help staff with chatbots, why spellcheckers are worse now, why the Google search algorithm is better in some ways but absolutely awful in others… AI doesn't yet live up the the promise. XD

That said, these art bots can replace bad artists at least.

Hey now, no need to get personal! *mumbles and goes back to playing with his blocky spaceships*

dpat57 wrote:
Hey now, no need to get personal! *mumbles and goes back to playing with his blocky spaceships*

Hahaha!
Fro what I've seen of your spaceships they're shaped the way they are for a reason. I can look at them and think about function and design. I can't do that for any of the ships that I saw on the AI I played with when I was testing my idea XD

Oz just explained it better than I could. While I don't see AI creating another Guernica, or anything by Paul Gauguin I DO see it winning a bad Hemingway contest.

Okay, thought I'd better test this out, so I asked the AI for a French girl on roller skates and here's the result.



Say what you will, but I found this strangely moving.

The skate shoes should be tricolor and she should be holding a Gauloise so I know she's French. But yes, nice composition.

That empty skate– intriguing.

It bothers me that the wheels are all messed up. You would think a robot could at least get something like that right. FAIL

That undoubtedly represents the inner turmoil of the French girl as she skates forlornly in search of her lost love, represented by the empty skate. She appears at first to be depicted lightly, in the manner of Chagall perhaps, but witness the gathering grey clouds of melancholy behind, which seem to reach out and form the very ground upon which she skates. Her tiny, frail arms are ill equipped to bear the all encompassing weight of her sorrow.

All ethical reasons aside, it's fun just messing with the prompts to see what the algortihm spits out. These were my personal favorites, with the according prompts.



Dino wizard doing magic

Cyber Warrior overlooking city, epic, in the style of Tsutomu Nihei

Show me Cthulhu


This 3rd one especially looks deceptively legit. All of them are Midjourney and i really notice a difference in quality as far as looking “convincing” compared to other AIs, genuine spot on composition and choice of color in a lot of results.

Teen me would have gladly hung those posters on his bedroom wall.

I remember that episode of Big Bang Theory when they visited Apple's offices and a woman who sounded just like Siri was fielding questions from customers. Are we sure there isn't a lone digital artist sitting in front of a dozen screens, quickly putting this stuff together?

IMHO the problem of machine-generated art is that, art being a form of communication, we expect that there is another mind behind the piece of art, and knowing that there isn't diminish the experience. For example, suppose that I see a painting of a beautiful girl that turns in a cockroach: this clearly would have a strong emotive connotation. But now suppose that I learn that the artist meant nothing by it, it's just that he made a market research and discovered that the trendiest sublects for art were girls and cockroaches. This would certainly diminish my interest for the painting, because the emotive connotation would feel fake - but only after I knew it. The same would happen if I discovered that the painting was made by an algorithm, as I would feel no emotive connection with it.

marcorossi wrote:
IMHO the problem of machine-generated art is that, art being a form of communication, we expect that there is another mind behind the piece of art, and knowing that there isn't diminish the experience. For example, suppose that I see a painting of a beautiful girl that turns in a cockroach: this clearly would have a strong emotive connotation. But now suppose that I learn that the artist meant nothing by it, it's just that he made a market research and discovered that the trendiest sublects for art were girls and cockroaches. This would certainly diminish my interest for the painting, because the emotive connotation would feel fake - but only after I knew it. The same would happen if I discovered that the painting was made by an algorithm, as I would feel no emotive connection with it.

kinda like pop music

We have a Quackcast on this going up tomorrow.

This kind of AI generated imagery has a use- you can use it to generate the most basic, starting level version of an art piece because like @marcorossi says it uses searched imagery to develop its images.

The problem (and it's a HUGE problem), is that people think it can do far more than it's able to. This is why you have so many people calling the images “art”.

People fundamentally misunderstand even the most simple principals of the concept of art and design.
It's one of those things- we see art and design everywhere so everyone thinks they understand it far better than they do, so when it comes to this stuff they mistakenly think it's the same thing.

But in reality is like mistaking the voice of the Swedish Chef Muppet for the Prose Edda in the original Icelandic.

lothar wrote:
kinda like pop music

This is a good analogy.

People always say that modern pop is terrible, while others counter that because people have always said that it can't actually be true and it must be just faulty perception…

While actually it is more true now that it has been in the past because the tools for creating sound-alike stuff are better so that when people say that music tends to sounds the same it's often true. They can also use programs to search out what's popular and create music with it…
There's more to it that than but modern digital editing means that now more than ever music really does sound the same and it's not very good…

It reminds me of a few years ago when a lot of digital art started to look the same- it's because everyone was using the same brushes and the same digital art tools. This is like that but worse.

I would propose that text-to-image AI programs are just another medium. As with all new media, it takes a some time for people to figure out what artistic product is truly unique to that medium. (thereby making it conceptually meaningful)
Some comics, like mine, use 3D models just as an underlay, which I use to draw 2D art on top of. The 3D modeling part of my comics plays a minimal role in the conceptual intent of my end product. But in many other comics, the synthetic quality of the 3D render is an intrinsic part of the visual experience of the end product. It makes full use of the medium.

A comic that uses AI generation algorithms could conceivably be valid if it offers an experience that is only possible to create with the tools contained within that particular medium. Perhaps a hypothetical AI-generated comic could come as a script of all the spoken dialogue. Where there should be a “panel,” there would just be a cluster of verbal prompts specifically tailored to make the AI procedurally generate the appropriate image on the fly. It wouldn’t be exactly the right image, but it would be statistically close. Once you turn the page, the generated image would dissipate back into the cloud until it needs to be regenerated again. Everyone reading the comic would get the same story, but the visual experience would be unique to every reader. Similarly, re-reading the comic would result in a slightly different comic than the one you read before.

Of course, the hypothetical comic that I just described above would probably be really lame-o. I just offered up the concept for the sake of argument.

It could indeed be lame-o, but with a human editor to pick the good stuff you now have a comic generator that requires a lot less person hours to reach a finished result. A creator who writes a lot faster than they can draw, which probably applies to most of us, could get their work out at a much faster pace, even if they still do a lot of ‘hand’ finishing.

If I had a program that would eat my thumbnails and spit out finished panel versions in my chosen style, until I'm happy with the result, should I really keep on doing it the hard way?

Ironscarf wrote:
If I had a program that would eat my thumbnails and spit out finished panel versions in my chosen style, until I'm happy with the result, should I really keep on doing it the hard way?

Frankly I'd kill for an AI that auto-colors my comics. I don't mind line-art but coloring is the most tedious, repetitive thing in the universe. I wouldn't even care if it looks a little crappier than my usual color job (which is not that great to begin with anyway) if it saved me the hours of sufffering it myself… a bit of quality downgrade would be acceptable.

This is emblematic of the whole AI art fantasy bubble:

FIRST GRAPHIC NOVEL PRODUCED ENTIRELY BY AI!!!
https://beincrypto.com/ai-art-worlds-first-bot-generated-graphic-novel-hits-the-market/

Yeahhhh… It's not what you think.
The reality is that it's a 700+ plus page picturebook full of unrelated non-sequential images.
That doesn't put anyone out of a job!

These kinds of books made with copyright free art, or old illustrations that have been licensed, and then had a story added to them loosely linking things are an old concept.

The main difference is that licenced images and copyright free art have far more meaning and can stand alone as artwork, while none of the AI stuff can.- it's also not really copyright free because it's made with digitally scraped stolen art.