Forum - View topicNo, You Can't Make Anime with AI
Goto page Previous 1, 2, 3, 4 Next Note: this is the discussion thread for this article |
Author | Message | |||||
---|---|---|---|---|---|---|
dmanatunga
Posts: 61 |
|
|||||
I really connected with the talk of inconsistencies in this article. It is one of the things I find most jarring when 3D elements are added to anime. Just the perfection of the shapes just feels like an unnatural clash with the rest of the medium. I do hope that the technology improves (whether AI or other) such that this conflict can be reduced.
|
||||||
Matros
Posts: 306 |
|
|||||
You should feel bad for even thinking about it. The One Piece staff is very passionate and hard working, they value quality and it's one of the main reasons why the show's current quality puts many seasonal shows to shame. Why would they want a half baked tool that achieves the opposite results? |
||||||
Vanadise
Posts: 516 |
|
|||||
To make a more extended analogy, AI is a tool like a screwdriver if your screwdriver had a 10% chance of overtightening the screw and stripping the threads, or camming out the head of the screw, or sometimes hitting it at an angle and bending it. If the person using it has a hand that is too large and the moon is full, it also has a 0.01% chance of catching on fire. Absolutely no professional would ever use that screwdriver. The amount of work it would take to fix its mistakes is more than it'd cost you to just use a plain old manual driver. That's the state that AI image generation is at right now, and as somebody who's been involved in software development for >20 years now, I don't see any indication of that changing any time soon. People have been saying that AI would replace humans in various fields for decades now -- translation, writing algorithms, folding proteins, etc -- and so far it's never happened. It always gets to the point where it can impress laymen but professionals won't touch it, and that's where image generation is at now. Will it be better than that someday? Maybe! But there's no reason to believe that right now except for the hopes and dreams of naive techbros. |
||||||
AiddonValentine
Posts: 2277 |
|
|||||
That's the big problem; if a human is still having to do stuff like edit or even throw out entire bits, then it's not truly AI-made. In fact, it makes an argument that it might even slow things down because of the sheer amount of work that has to be sifted through so why not just do it with a person in the first place? All AI can do is just trace, it can't actually understand intent. This is how you can tell a lot of these AI junkies don't appreciate art because they don't get that art is more than just drawing or painting. |
||||||
dm
Subscriber
Posts: 1436 |
|
|||||
Gosh. I have a wretchedly untrained eye, even after decades of watching animation --- everything from weird little Eastern European things, to claymation, to Pixar, to rotoscoping, to anime. Of course, that means I've watched a lot of crap, so maybe my eyes have been ruined.
I thought the animated short looked pretty good, actually. But, I think people should think of machine assists less as "artificial intelligence" and more as "area fill and layers on steroids" --- or, when used in the creation of manga, "hypertrophied screen-tones". Really just more tools that skilled artists can use to create something fantastic, and mediocre artists can use ... to strive after their dreams, or something. |
||||||
MagicPolly
Posts: 1616 |
|
|||||
The same reason manga publishers want to MTL text and then hire translators as "editors", you can pay them much less when their job description is just as a corrector Hinestly corridor's video is a filter at best, I'd barely call it rotoscoping |
||||||
Feng Lengshun
Posts: 50 Location: Indonesia |
|
|||||
The article echoed most of my opinions.
The way I see it, AI will be a very useful tool for in-betweening, and in some cases, with specialists that can do a very specific thing and get it done well, which isn't that much different than the current system of contracting specialty animators and editors. But it's never going to replace human inconsistencies. Plus, the legal-grayness of the data source means that it's most useful for studios with a long history of animation they can draw from, making them most useful for the ones making anime for cheap (and thus not paying their animators well) by the dozens. I don't see KyoAni implementing AI anytime soon, and even Ufotable is going to take a while because they already have a good VFX pipeline and I don't see any motivation to mess around to plug-in AI when it'd just add more work for them to clean up. The major consideration is the animators, and personally... I think the current status-quo is already too harmful and unsustainable. People are paid at rates where many are living with second jobs or subsist on assistance programs. All because they really want to make anime, and the studios need to make a lot of anime for cheap due to the demands. I... personally would rather they be released. It's sad, but I'd rather them try for other jobs much like game devs that are moving on from the industry, instead of enduring dehumanizing conditions because they're passionate about anime and studio needs a lot of fodders. |
||||||
kotomikun
Posts: 1205 |
|
|||||
Well, maybe. But if that does happen, it's going to require something fundamentally different from the current technologies, not just iteration on them. (And if that actually does happen, we'll be in the territory where we need to worry about an AI takeover of humanity.) What we're calling AI these days is a real-life version of the "Chinese room" thought experiment. It takes symbols (and/or pixels) as input, manipulates them according to patterns it recognized in training inputs, and outputs more symbols and pixels. At no point does it have any understanding of what the data actually is, because it never had to learn through real-world experience what words and images actually correspond to and construct a model of reality based on those experiences. Without any general intelligence, it can't do more than copy things that have already been done by humans thousands of times, and will often make weird mistakes that it has no way of even recognizing. Stuff like background objects that are misaligned on either side of a subject, or things that should clearly be round (the moon, basketballs, etc.) being ovular for no apparent reason. Or, in this animation, lighting and clothing that flickers chaotically. It's impressive that it can produce passable art at all, but I don't see this ever being able to take "produce an entire Naruto episode" as input and spit out something ready for broadcast. Even with rotoscoping, human voice acting, and probably a lot of manual post-processing, the result was still... really bad, not something anyone would want to watch without the novelty factor. |
||||||
Boy Howdy
Posts: 24 |
|
|||||
If standards can go that low. |
||||||
rikku45
Posts: 107 |
|
|||||
Say what you want, this rock paper scissors thing was awesome.
stop having a whinge |
||||||
Feng Lengshun
Posts: 50 Location: Indonesia |
|
|||||
I think a lot of people, except the ones that are very anti-AI Art for one reason or another, think it's cool. Heck, I think even the people who are anti-AI Art think it's cool, and that's why it provoked a stronger emotional response. Regardless, the whole discussion only happens because of Corridor crew's video being titled "Did we changed animation forever?" Some are more heated about it than others, because it's the internet. But it's a fair question to ask if, when, for what, and how will AI be incorporated into anime, as well as what are the ethical, moral, and economical implications of it. Especially since the industry is, quite frankly, broken in so many ways that AI to shore up some of its issues are inevitable. |
||||||
AsleepBySunset
Posts: 231 |
|
|||||
You're being ridiculous, it was genuinely Handshakers tier. Even if you truly, 100% loved it, a non AI version, where they used the exact same production pipeline, but replaced the AI-filter with a regular photoshop filter would have been equally as good, if not better. I mean, I've seen clips of it, and the AI filter didn't even implement the stuff which would be really hard to implement with regular programming, like giving the chins a sharp point, a staple of anime. I'm tired of people insisting AI is the solution to every problem in the animation industry. Image generating AI is fundamentally unethical, regardless of whether you use it to draw keyframes or inbetweens, regardless of where you scraped the data from. |
||||||
enurtsol
Posts: 14853 |
|
|||||
Disney Animators Have Unexpected Reaction To Controversial AI-Generated Anime:
https://www.youtube.com/watch?v=jQ_DfORb3kw This is just proof-of-concept. It's just the start. Others will refine it and improve. Just like 3DCG looked like crap in the beginning; now everybody uses it at some aspect. Look how much ChatGPT has improved - from the Bottom 10% to the Top 10% of the US Bar Exam for law students, and kicking ass in the SAT and AP exams "ChatGPT v4 aces the bar, SATs and can identify exploits in ETH contracts"
So, ChatGPT took these exams cold. That means 90% of us are now dumber than a machine |
||||||
AsleepBySunset
Posts: 231 |
|
|||||
I like how all this AI research is just a thinly veiled tutorial on how to cheat on your homework, the only "practical use" for chatbot AIs. (edit to add) Also, the implication AI is growing "more and more intelligent" is ridiculous. It has no reasoning skills, none, it does not understand, for example, theory of mind. See https://garymarcus.substack.com/p/how-not-to-test-gpt-3?utm_source=twitter&sd=pf. Noone has worked out how to program general reasoning skills such as if x then y into an AI chatbot. Chatbot AI does not reason, it merely predicts the next word. |
||||||
enurtsol
Posts: 14853 |
|
|||||
When ya check the source link, you'll see other uses And also now as a gaming guide helper
|
||||||
All times are GMT - 5 Hours |
||
|
Powered by phpBB © 2001, 2005 phpBB Group