Skipping a step: Corridor Digital and AI anime
Almost 6 months ago Corridor Crew released an AI-drawn anime short (ANIME ROCK, PAPER, SCISSORS) with an accomppanying making-of video ( Did We Just Change Animation Forever?). It got... mixed reception. Some loved the new era of "democratizing animation" (meaning you don't anymore need a team of hundreds of animators which in turn means it's possible for smaller creative teams to make their visions come to life), others really hated it for blatantly just ripping off an existing anime (Vampire Hunter D: Bloodlust, 2000) and general disrespect over animation as a job and art form -- or at least that's how (some) animators felt. Having heard them talking about drawing each frame with such a passion (on Corridor's show!), I can understand the ire.
Now, almost half a year later, comes the sequel (ANIME ROCK, PAPER, SCISSORS 2) also with an accomppanying making-of (Did We Just Change Animation Forever... Again?). Things... have changed. Basically Corridor realized that stealing art is bad, so they hired a real artist to draw a model sheet as a base for the AI to draw from (instead of stealing others' work). They also hired a person to write a theme song and a team of online artists to touch up every frame of the anime (watch the making-of if you're interested in the details, they go through them very well).
Next, some personal opinions of mine, starting with the first anime. I liked it. It was a nice and funny short with an interesting, smooth style that comes with the territory when there are more frames crammed into a second. Overall, it was the goofy concept of rock paper scissors combined with the over-the-top life and death drama that was fun. Visual style on the other hand, nowhere near ready. The warping and "worming" between each frame were really distracting and it wasn't ready for more than a tech demo (or for some relatively out-there story where that stuff ties into the film, not as a distraction). But I was able to look past those problems because it was a pretty good video.
Most of all, I didn't like them using artists' work without permission (and not saying anything about it).
Now to the sequel. It's... basically same? Same problems, less warping but for example king's crown was changing its color like it was having some sort of multistage chemical burn, and the visual style wasn't as strong and at times more clunky than on the first one. Maybe that's due the fact that the AI style guidebook was a lot smaller or that they were only willing to spent X amount of hours and money working on this while aiming for the anime episode lenght -- I don't know. But the story and the writing were still the best parts. Interestingly also I think direction was a bit weaker and they used too many "cool moment" tricks which made it visually messy. It basically got in the way of the story.
(Also I really dislike that Niko still wasn't taking responsibility for stealing art from others, bit of a bummer since most of us knew better six months ago already.)
What they proved with the second anime is that AI is still not close to replacing actual artists and it's a lot of work to make them even this way -- even if the AI part worked smoothly! But most of all what matters is the content, the creativity and how it's translated to the screen. Not the AI. It's a tool, not a revolution.
Edit. For clarity and some additional thoughts.
I just don't think style is an IP, if you asked someone to draw a picture or make an animation in some shows style that it would be considered stealing.
Copying style might not technically be stealing, but it’s often seen as cheap, bandwagoning, etc.
It ultimately depends on what the copying artist adds to the equation though, which is why e.g. a new anime with a visual style that’s reminiscent of a classic manga artist’s work but modernized or otherwise has some twist is seen as an homage rather than a cheap knockoff.
This isn't an issue that's new to LLMs. Music has a long history of copyright cases where the question of copying style or melodies/chord progressions/etc has been a central (and incredibly fuzzy) issue. You're right, style isn't exactly copyright friendly. Copyright is a broken system. I think what's motivating a lot of the discussion is that we feel strongly about the ethical implications of LLMs, and copyright isn't a system designed to address our ethical concerns. It's designed to regulate commerce, and even to that extent it's an archaic system that's poorly equipped to handle modern creative/entertainment/art industries, at least in any way that's remotely equitable. Again though, not designed to address ethical concerns. TL;DR capitalism sux
For me it makes more sense to think of it like a software license. Because it's a tool, made by a third party, but what you make with it is your creation.
These AIs work in the same way. The art you feed it translates into ones and zeros, becoming its own part of the program, like a plugin, which you then use to create your own work.
I'm not in a place where I can watch this right now, but I did scrub through the video really quick to see how well it worked. And from what I could see, it's a marked improvement, though most of it appears to be because a lot more work was put into it compared to the previous one. The biggest problem is the faces. The faces actually looked worse than the first one. The bodies are actually mostly fine throughout it, minus some issues when dealing with props. If they had handled it like a real animation studio does, by having a junior artist drawing the faces, it would have been a huge improvement.
To be honest, there's a whole barrel of worms I could open about why this approach still isn't as good as traditional style animation, but frankly I don't think the conversation would be very productive because the people who have opinions on it aren't likely to change them at this point.
It does seem a little early to call it stealing. It's still unknown if AI art is considered derivative or transformative in US copyright law at this time, and it will likely take a few years to resolve that question. The answers may also vary in other countries, which could result in a large shift in industry and brain drain towards cooperative countries.
Though personally, I feel even "derivative vs transformative" is a bit too black and white, and it may not be so simple. Generative art can be programmed to closely model its source material, or to generate as differently as possible. It seems to me that thresholds for certain allowances would be necessary for using models trained on copyrighted (and unlicensed) works. Or possibly, some sort of quantization pass may be necessary on the data itself to reduce the resolution on the original works.
Another consideration is the size of the database. In the same way that information can only be anonymized in a sufficiently large sea of data, an AI training database needs to be large enough to avoid creating too similar of works to those that inspired it. Models may need to be rated for different categories based on their training data to ensure there is sufficient variety for unique results.
Most of this only applies to models trained on copyrighted works. I can see a secondary market emerging for content licensed specifically for AI use, at higher resolutions and CFG Scales, and with highly-detailed metadata tagging.
My opinion remains in flux as the technology is evolving so rapidly, but I do think it's a reasonable and fair goal to allow the creation of generative art up to the point that it impedes on a creator's rights. This will have democratizing effects, but also disruptive effects to existing industry. I would hope that policies and programs would emerge to soften the landing for those negatively affected.
I'm looking forward to the barriers of entry lowering, but also wish we lived in a time when folks cared to make sure the industries affected could shift into a new configuration without leaving folks in shit positions. I'm also concerned with whether this will make good work harder to find; the field absolutely will get flooded with the most low-effort stuff imaginable and it's totally possible the audience for that is enormous. It's possible to just go check out the shit you like now, but there's a layer of advertising and cultural noise which is inescapable. And of course, endless, inevitable examples of creators starting out sincere and ending up part of that noise. That whole..thing... is what I'm not looking forward to.
All that being said though, I am also excited for those inevitable stories of folks who otherwise would never have been known, getting seen because the tools got easy enough. Imperfect as it is, it's stuff that I just didn't think would actually happen while I was alive.
I thought the anime was impressive considering where stuff was at even just a year or two ago, but I also look at it and think, beyond the technical issues, there's nothing there, in a way. Let's credit the work to the "AI", which in reality is a complicated set of statistical models, crunching through fucktons of data in specific ways. They're not sentient machines, or at the very least, they lack the ability to participate in things such that folks begin believing they are people at scale (random people on the internet are not representative). So, that means the work was not made by a person. There was no conscious thing directing the choices; the choices emerged through computation. The technical achievement is astounding, so my appreciation of it is similar to seeing something like a raytracing demo. But in terms of me as a person, experiencing this, I can't call it "art", and I can't connect with it in the way I would something artistic. At base, what's "artistic" is pretty simple - something a person made. I don't think this would really be contentious, but I also don't expect anyone to agree with me. "Person" is what's flexible - anything that can participate in society sufficient for folks to regard it as a person, counts. It's "good enough" at that point. If we can develop a machine which can do that, and that machine produces something like this, then I would be all about looking at it more closely. What makes art interesting to me, is the set of choices people make, and what they were wanting us to know from what they did.
So I'm not appearing hypocritical - I thought the anime was a fun little thing. Kinda funny, absurd, sure I'm down. But if the deal is that the tools make the creative aspects tremendously easier, then my judgment is more focused on what choices did exist, and so I can't help but compare, and it does nothing special at all. That sounds harsh, I don't mean it to be, but we've all seen things of this nature a thousand times. What makes each special to us, are the specific choices made by the people who produced those things. Sometimes that comes from a more cynical place, but not always, and (I would argue) not most of the time.
So this - some light humor and silly situations really any of us could come up with. That kinda tracks too, since the models are all drawing off stuff we all did. The only creative choices were stuff like this; the premise, the jokes to keep, the direction to let it go in. I'm not really impressed, I've seen some tremendously good shit from people over the years. Since they didn't have to apply themselves to it in the same way, I feel less restricted in that sort of criticism, because again it's all that's there from my perspective. In a time where it can be produced "perfectly" (enough that you don't realize it was made that way), there's no one to credit besides the person who chose what would get shown. Is that something we can one day give the same sort of reverence, as the sort of paintbrush someone uses, or the type of guitar they play? Perhaps it's the same as any other medium, and what it takes is someone making "the right" choices for folks to look at it like that. I think for now, if someone is going to put this sort of work out into the world, and attempt to credit the machine, I'm going to use it as an opportunity to be very particular about what it makes. There's no artist to offend, after all.
Anime has already been revolutionized (I think for the worse) by technology used by VTubers such as Kizuna Ai (OK) and Hololive (I think a disease)
With the whole fanservice thing, self-referentiality and after Evangelion and Lucky Star, endless anime set in an isekai sekai and Touhou and fan art of Touhou and the fact that almost all the images on Danbooru that aren't of VTubers come from video games (where you are the event) anime doesn't really want to be commercial, it wants to be by the fans and for the fans.
Unfortunately a lot of people don't like anime because of certain trends in it that are seen as having gotten worse over time. Both my son and wife think it is a disease and my son believes that exposure to anime seriously damaged one of his childhood friends.
Sometimes I find something that they like but often it turns bad, Gin Tama for instance has a good mixture of yin and yang but my wife thinks it is too much like The Simpsons.
I'm in the same boat. I loved anime well into my college years but as time went on, I found the same tropes and themes being rehashed over and over again and found watching anime both uninteresting and tedious. I'd hop back in whenever something was receiving a lot of buzz and acclaim (Attack on Titan for example) but I couldn't stomach it. At the same time, US animation was getting better and exploring ideas anime just wouldn't/couldn't touch. I'm thinking Adventure Time, Avatar/Legend of Korra, Over the Garden Wall, hell even the Spider-verse movies.
I did try the "next Miyazaki" route as well and watched the majority of the works of Makoto Shinkai (Your Name, Weathering You, etc) ) and Mamoru Hosoda (Belle, Wolf Children, etc.) and while better, they still felt too hollow and anime trope-y. At this point, it's really just waiting for the final Miyazaki film to come out stateside and from there, I'll probably ignore most if not all anime.
That said the last anime I truly enjoyed was , funny enough, the Rebuild of Evangelion tetralogy. As I got older, Eva's place in my brain went from a ground breaking, mind bending anime into this melodramatic, nonsensical series covered in a thick layer of pseudo-psychology and meaningless edgelord religious references. However, given how much Eva reflects the Anno's personal journey, the Rebuilt tetralogy really shows his growth as a human being and as creator. It has some flaws (fan service for example) but ultimately, it's a coherent well thoughtout story, has proper character development, and tells the best version of the world of Eva. It's really good and if Eva still has an impact in your life, I highly recommend doing a binge of the movies.
Yeah my partner and I hated the rebuilds. Felt like trying to cash in on the creeps still waifu-ing teenagers in need of serious therapy. They even did the thing like "eva pilots don't age, so Asuka looks 15 but she's actually 30," but I'm not even sure that's the worst part.
Mari spoilers, more creep details
They introduced a new pilot, Mari, who gets about 4 lines and half of her character is boobs. She's actually a friend of Shinji's mom but didn't age bc eva pilots. And in the end they pretty strongly imply she's Shinji's romantic partner, and this is after Shinji got time frozen for 15 years!So it feels a lot grosser than the original and like they're tacitly encouraging the waifu-ing. Which ties back to the original thread about trends in anime -- gotta pump up that fan service apparently.
I'm probably biased though since I disagree with takes calling it a "nonsensical series covered in a thick layer of pseudo-psychology and meaningless edgelord religious references," but that's a whole other discussion.
Ah fair enough, the third movie is the one that deviates the most from the original story and to me is what really sold the tetralogy for me but then again I did enjoy or at the very least tolerate the first two movies. It was one of those things that I realized I enjoyed the more I thought about it, then again that type of reflection maybe distorting how I actually felt while watching the films.
Maybe talk to your son about not thinking that way? The kind of thought process that decides that someone chasing their own harmless, personal joy is damaging usually goes hand in hand with voting to take people's rights away, I've found. I hope your son is an exception to this observation.
You've got a point and I've certainly floated that with him.
Your argument does come close to a very glib thing I hear repeated on Mastodon to the effect that "What's right and wrong in your religion does not matter at all to me" which unfortunately can be immediately turned around to "Your science says you can't put unlimited amounts of CO2 in the air without consequences but mine does".
If there's a single reason for the cultural crisis today I think it is a form of extreme individualism where on the right a billionaire deserves every penny of their money and can do what they want with it (and it's none of your business) and on the left you can invent any identity you want whenever you want to (and it's none of your business.) See
https://www.amazon.com/Rights-Talk-Impoverishment-Political-Discourse/dp/0029118239
What technology?
https://en.wikipedia.org/wiki/Live2D
(disclaimer: I am a Hololive fan)
I'm pretty sure Kizuna AI was only ever 3D, and I'm not aware of L2D being used to make any actual anime shows, nor does the wiki article list any such shows. How did L2D make anime worse?
You're free to have your own preferences but here are mine... (as a person who deserves the title "Weeaboo" because my interest in anime has gotten me in trouble)
There have long been people who think anime has something unhealthy about it and that includes Japanese commentators like Saitō Tamaki and also American commentators like Thomas Lamarre both of whom have written important (and not entirely negative) books on the subject.
As I see it, Hololive is the belly of that beast. Other anime has plot, sometimes excellent art, brilliantly choreographed fighting scenes, complex situations, etc. Hololive has a running gag between very heterosexual Mori Callilope and Takanashi Kiara (who can't decide if she is a chicken girl or a phoenix girl) who has a crush on Mori. Mori speaks in sexual innuendo and the whole joke is that Kiara doesn't get it, the trouble with this is that Youtube has gotten increasingly prudish and you'll get demonetized if you have any real sexual innuendo so it is increasingly "coded" and reaches for anything that provokes squicky feelings as a substitute for feels of arousal. (I'm kinda embarassed to admit that I've seen this!)
A friend of my son became a "shut-in" around the time of the pandemic and has gotten into what we think is a very bad place and this person has learned to talk like a VTuber and is surrounded by people online who see everything the same way and doesn't get any mainstream perspective. We could be wrong (although my son wouldn't admit it) and it might not really be the fault of VTubers but this person's interest in VTubers have been part of a number of changes that we think are unhealthy for this person.
And insofar as I make the claim "revolutionized" I'd say that "anime" is practically the hub of various media such as manga, video games, light novels, etc. and you could say usually an IP has "made it" when it has a media. I'd point to whatever is being posted to Danbooru at any given time is the center of gravity of "anime" for many people, and by that standard things like Touhou, Kantai Collection, Hololive, Fate/Grand Order, Genshin Impact and now Blue Archive punch way above their weight even if some of those things haven't had a commercial anime of them or if the commercial anime is insignificant compared to the property.
Hololive is not anime. It is at most anime-adjacent, so I fail to see how it could possibly play a part in ruining actual anime. Hololive has
7580 streaming talents; interactions between two of them do not define the company. Calli has never outright stated her orientation to my knowledge. The gag between Calli and Kiara was definitely something they played up at first but has not been a thing for about 2 years now, possibly more. Kiara has specifically addressed this. (I'm only familiar with "squick" meaning "gross" as in causing feelings of revulsion, like body horror might, so I'm not sure what you mean by that.) People definitely develop unhealthy attachments to online personalities and communities, but this is hardly unique to vtubers. I'm sorry for what happened to your son's friend and I can see why you might blame the content he consumes.As you say, we both are entitled to our preferences, so I think it best we agree to disagree here. Have a pleasant day.
I'd maintain that today anime isn't really the center of gravity of anime. Looking at the fan art site Danbooru (definitely focused around anime style art as opposed to manga style art or 3-d style art),
https://danbooru.donmai.us/tags?commit=Search&search%5Bcategory%5D=3&search%5Bhide_empty%5D=yes&search%5Border%5D=count
I'd say out of the top 20 everything is primarily a video game or vtuber except for #18, which is a straight up conventional work of fiction (I'd include something based on a manga or an light novel as "conventional fiction". I'm not going to take a Robert Ebert position that video games aren't art, but there is an element of collection as noted in https://www.amazon.com/Otaku-Database-Animals-Hiroki-Azuma/dp/0816653526 that is intrinsic in most of the top scoring games such as FGO, Pokemon, Kantai Collection and even arguably games like Touhou or Neptunia or Dynasty Warriors that have a crazy number of playable characters.)