We are at the brink of a revolution that will change anime and manga forever, it's time we had a thread about it.
What do you think the anime industry will look like in the stable diffusion era? How long do you think it will take for the first commercial manga produced using AI art generators to be released?
What about the first anime? I wonder how quickly the anime studios will start to adopt AI art generators into their productions, and what will happen to the rank and file animators on staff. Will the Korean animation industry collapse without the grunt work anymore? Or will there just be FAR more anime created because it's so easy to do?
it'll take fucking decades dude ai is nowhere near making affordable coherent animation
I guarantee you that it will take less than 2 years going from image to video isn't that difficult, in fact the ability to interpolate from still images to video already exists, not sure why you think that video would be much harder
>stable diffusion era you're born with a perfected version of it, it's called creativity
people think this stuff is going to kill independent artists
this stuff makes artists more valuable
A.I. is only good at generating images that superficially resemble artwork at a glance. The illusion of artwork. Once you look closer, the illusion falls apart. Deformed facial features, giant holes in characters' bodies, missing limbs, clothing fabric that makes no sense, limbs that bend the wrong way, fingers that have the wrong number of hands, etc.
A.I. generates images that vaguely resemble something you recognize, but after just a few seconds of examination, the illusion collapses behind any hope of repair.
A.I. does not understand the subjects it is creating images of. It does not understand things like "a hand has 5 fingers" - it just knows "I have 1,000 pictures labeled 'hand' and I will now blend them together" and that's what results in the deformed monstrosities you see in most A.I. generated images.
Until an A.I. exists that actually *understands* what it is drawing, rather than just blending images together from a library without any awareness of what it's doing, it will be completely infeasible to use it for any kind of commercial purpose.
Getting it to animate? A thousand times more complicated. A.I. can't even figure out what direction a leg is supposed to bend in, much less how a leg should look in motion.
As of now, this technology is - at most - a fun novelty. It will not be ready for commercial purposes for several more years at the least, and that's because we have to invent the technology that teaches A.I. to understand the human body as a three-dimensional shape, and not just as a tag in a gallery of images.
you guys are both right, it's going to accelerate the fuck out of artists. artists will be able to go from idea to complete work of art in hours, by getting an image out of stable diffusion and then modifying it to fit their vision
>A.I. is only good at generating images that superficially resemble artwork at a glance. The illusion of artwork. Once you look closer, the illusion falls apart. Deformed facial features, giant holes in characters' bodies, missing limbs, clothing fabric that makes no sense, limbs that bend the wrong way, fingers that have the wrong number of hands, etc. >A.I. generates images that vaguely resemble something you recognize, but after just a few seconds of examination, the illusion collapses behind any hope of repair.
there will also be a new concept in art and design of an AI driver
in 50 years you will be able to use a prompt 100 words long and make a 90 minute movie come out of it. adjusting this and getting the best result with the fewest words or other limitations, etc, will become like a sport.
it will be an entirely new thing that didn't exist before in the art/entertainment/math/gambling world.
do you know how many images a video has? connecting all of those and making actual animation isn't possible with current ai
Call me when this "revulution" thing will be able to emulate cel animation with the right shading and frame roughness, then we talk. Plastic digital tranny animetion images don't count
>>>>>>>>>AI marketing bullshit lingo that I wish people would stop using, like VR and hoverboards. I don't see it replacing human artists in our lifetime as things stand, however if standards drop far enough with the next generation of kids the ways it has for other forms of media (i.e. all those people who would rather watch people play videogames than play them themselves) then I may well be proven wrong interesting how half of those have watermarks and just about all of them have some issue, whether it's the disembodied hands or the assymetrical clothing or the hair dropping over the shoulder and vanishing. as far as >A.I. generates images that vaguely resemble something you recognize, but after just a few seconds of examination, the illusion collapses behind any hope of repair. that image just proves him correct
False. most people aren't capable of understanding how fast this technology is progressing. if you don't get the math behind this stuff at a deep level you cannot understand what's going on.
In the immediate future this will make storyboarding And pre production much more efficient, then it will be universally used for inbetweening and to make frames drawing quicker, but the industry is slow to change so I don’t see any radical use this decade
Being honest, I think japanese animation and manga are too conservative and entrenched to be the first to take advantage of it, they’ll be able to stand just by their popularity, but other countries’ animation and comic sector will be able to outproduce them
>comic that's all scenery with no characters
quantity is pretty meaningless.
this will make quantity entirely meaningless.
now only quality matters at all.
AI art generators have been out for like 5 months, let's wait until they are more mature before critisizing them so harshly. I don't even think there's a polished "complete" one released yet.
When the "photoshop" of AI art generators comes out I think those issues will be fixed. If you specify 5 apples it will give you 5 apples. If you specify human it will give you a picture of someone that looks human with eyes in the correct places and hands with enough fingers. Google's Imagen can do this already. Give it time.
google says their implementation is two levels above what is being discussed or shown in the thread.
so compare generated Elsa images to those images from a few years ago with spirals of fractal dog faces. and then apply that going forward.
Dude, you just proved my point. Every one of those Elsas has glaring flaws.
>Misshapen clothing fabric >Weird, jagged blush on cheek >Down syndrome eyes >Torsos that make sense out of the corner of your eye, but once you look closer you realize the anatomy is fucked >Hands with too many figures >Eyes and pupils in weird shapes
most of them look really fucking good i think you are nitpicking a bit too much if a mangaka was capable of creating work that consistently good we would all be perfectly happy with it
just imagine what doujins will be like in 10 years
>those images in any way appearing similar to what you were saying
one of the really, really sad things about the state of the world is that internet-only interaction and the malfeasant desire to "win" has made people like you only capable of disingenuousness. look at you pretending this is anything like what you said, fag. fuck yourself.
not that you were ever taught not to be this way. sigh.
Sure but it needs to do that 8 to 12 times per second and stay consistently on model, fro mwhat I've seen I don't believe we're remotely close to that no matter how much Google might brag. And even then that's only the animation. There's still story development, writing and voice acting to worry about
>Sure but it needs to do that 8 to 12 times per second and stay consistently on model
you.........think it needs to generate the images at playback speed?
Fuck off, crossboarders.
Why did you single out this line: >A.I. generates images that vaguely resemble something you recognize, but after just a few seconds of examination, the illusion collapses behind any hope of repair. then post a set of images that vaguely resemble something you recognize, but after just a few seconds of examination, the illusion collapses behind any hope of repair?
One thing to remember: everybody using it recognise that Text to Image is fundamentally a meme, it’s normal after all, using a prompt to generate an image is inherently inneficient and lacking, and will only be used for very early concept art generation or to look at an "alternate google image"
The real power of this is using it alongside images and videos, integrate it in your drawing software, feed it a sketch to get a picture, an uncoloured frame to colour it
God damn, that's ironic. I was just thinking the exact same thing, but about *you*. Your desire to "win" over the Internet has made you blind to the flaws in those Elsas. You're hopeless. There's no point attempting to explain anything to you.
Of course not, I mean for a full length anime episode it's going to need to generate at least 10,000 frames with consistently drawn characters, backgrounds, props and so on. Look at Presumably the same AI tool with the same prompt but they all look completely different. You can't animate that.
>not being able to see that he is right the art is really fucking good, and it was made by a computer. it's not perfect.
youre talking about something first generation like it stops there
when a year ago we had spiral fractal dogs
you lose, loser.
>vtranny i'm taking the other guy's side. fuck you, stop nitpicking
>Look at (You) >Presumably the same AI tool with the same prompt but they all look completely different. You can't animate that.
? why are you assuming this tehcnology is not advancing? it doesn't apply context to image elements now, it will in one revision, consumer-available in less than 2 years.
telling it to keep the same character will be effortless in a single hand number of fingers years.
the current industry is pushing cgi shorter cut down seasons and wants to milk more money out of this business . there is at best a few insividual artists who think ai generated has a chance and also work in the industry . non of the suits tho.
I'm willing to be proven wrong but I don't think that's a priority for the people developing this stuff
dude let me tell you a story
i try not to think about this. i knew about bitcoin the day you could get it. okay?
and i said "that will never be a thing"
that is you right now. try not to be stupid. you have myopia.