The “First 95% A.I. Anime” Is a FRAUD (and also looks like garbage)
We officially have the first all AI anime and it looks horrible. I do have to give them some credit. They admittedly used the technology in a creative way that fits its strengths. Weirdly though, that’s not really a compliment. And in order to understand why, we have to take a dive into this confounded production, see how it came to be, why its marketing has been extremely misleading, and why it looks so bad. And watch as I get more and more heated the further I research this production and the company behind it. I promise I’ll chill with the AI stuff after this third installment in what has accidentally become a trilogy about AI and animation. But the AI hype train doesn’t run on coal. It runs on fear, speculation, and misinformation. All of which is easily dispelled with a little scrutiny. Made by the newly created and aptly named studio Kaka Creation. Seriously, why does it smell like kaka and weiz in here? And starring Yuri Eigoma, the voice of Ruby from Oshino Co. Twins Hinahhima is a single episode glorified tech demo about two twin high school girls, Himari and Hinana, who want to get famous doing dances on Tik Tok. Please excuse me for a moment while I cram myself into an old oil drum and roll myself down a quite steep hill until my brain is jumbled enough that I can bear continuing to live in this world. Did I mention that one of them is outgoing and and the other one is shy? Isn’t that just so cute and extremely kawaii? And it’s the world’s first 95% or maybe 100% AI original, whatever that means, anime. At least that’s what you would think if you believe the droves of articles and YouTube videos that uncritically parrot that sentence. This anime set to release in 2025 will be the first anime ever to be AI created. If you take a look at their website though, what they actually say is that more than 95% of the cuts are made with some sort of support from AI. It’s subtly different wording, but that’s a very big difference from being 95% AI. And by the way, positive reception where ah I see they found a few people on Twitter who said it was fine. Well, color me impressed. As far as I can tell, there are five main elite strategies that they use to implement the AI. So, let’s watch through the episode and collect them all as we go. Starting with what is by far the most prominent AI filter over CGI. And speaking of this anime being a fraud, fraud and a service that can help protect you from it. This video is sponsored by Aura. Being a VLlist YouTuber, one of my favorite pastimes is googling my own name. So, imagine my surprise when instead of pages and pages of glowing reviews of my pretentious and unnecessarily verbose videos about cartoons, I came across a page listing my email, phone number, names of my family members, and every address I’ve ever lived at, including the current one. That’s not a good feeling to know that my very essence is being sold for a profit. If you’ve ever wondered how scammers got your phone number and why your inbox is flooded with spam, this is why you won. and they are legally required to remove your info if you ask, but they intentionally make the process harder and more confusing than canceling your internet service. Luckily, Aura takes care of the hard part, automatically removing your info from these sites and keeping it off. Not only that, they also monitor your credit files, send you alerts if your personal info is found on the dark web, provide a secure VPN, and a bunch of other stuff. It’s all-in-one digital security, all wrapped up into one app. And if anything does happen, Aura also includes $5 million in identity theft insurance protection, plus US fraud experts ready to help 24/7. Try Aura for free for 14 days at aura.com/swampj. In those two weeks, you’ll discover where your data is being exposed and sold. I’m not leaving myself vulnerable, and you shouldn’t either. Thanks again to Aura for sponsoring this video. Now, back to it. This is the very first shot of the anime. You know, where you put your best foot forward to try to make a good first impression. It’s honestly baffling how many glaring errors pop up in the first 20 seconds. It’s immediately obvious that this is rotoscoped CGI. I would assume primarily driven by motion capture data. It even looks like whoever did the mocap mimed what’s supposed to be Hana nudging Heari to the side to line them up better in the frame, but I guess when they lined it up in post, it didn’t really work out to have them close enough to actually touch each other. So, she just does like a hover hand nudge instead. Also, I’m now noticing what appears to be a blatant tracking error on Heimari’s leg here. See how when she steps back into the middle, her leg doesn’t quite line up right. It’s angled too far inward, then suddenly in one frame, it snaps back to the right position. That type of little glitch can happen all the time with motion capture data, but you’re supposed to, you know, have somebody fix it ideally. In this press release, Kaka CEO Yastomo Takahhara describes their workflow and lays out two major problems with CGI and AI that he claims to have solved by combining the two. To cut down on drawing time, they animated much of the show in 3D with Maya and Unreal Engine 5. They wanted to use Unreal Engine because being a game engine, it has real-time physics simulation, which is helpful for flowing hair and skirts. But being a game engine, it’s also prone to clipping issues where parts of the model phase through each other. Meanwhile, AI generated animation is prone to this weird flickering effect because the AI has no understanding of continuity from one frame to the next. So, the idea was that if you combine the two using the CGI as a consistent base to paste the AI on top of, you could fix both of those problems. So, how well did it work? We’ll get back to the flickering issue a little bit later, but on the clipping front, bad. Really bad. What the hell were you talking about, Mr. Kaka, when you said that AI fixed this problem, Hinana’s hair passes through her arms like three times in the first 20 second cut, but it gets even more egregious in the second shot of the episode where Himari’s hair is just fully incorporeal. In fact, to my eye, the AI even adds to the issue because, as I learned from my inbetweening experiment, it has a really hard time when lines intersect each other. So, not only do her hands pass through her hair, they now also dissolve into a mess of putrid slime along the way. Once again, this is 45 seconds in. I’ve got news for you, Kakaan. They’ve been making 3D anime with filters and shading tricks for years. This isn’t AI animation. It’s a Snapchat filter with extra steps. The CGI is also just kind of badly rigged. Like, the hair in particular has this very noticeable helmet effect. There’s a line where it’s free to flow, but anything above that line, it’s like it’s glued to their skulls. Unfortunately, Hina and Hea suck at being Tik Tockers and are catastrophically starved of clout. Oh, I forgot to mention this is also an interactive augmented reality type deal because they have a real Tik Tok account where they’ve been posting unbearable garbage like this since late 2023 when the production started. And they really commit to the bit because in just under a year and a half, they’ve managed to accumulate very little clout because people can immediately clock this stuff as soulless garbage and have no interest in it. Anyway, Hinana eats a bunch of grass in frustration, and this is going to be one of the two sincere compliments I give in this video, so savor it while it lasts. I thought this was kind of funny. It made me giggle. Moving on, the two get some soda, chat a bunch of nonsense about multiverse stuff. and sketch the still image. For some simple dialogue scenes, they seem to have let the AI generate a single image based on a sketch and a reference sheet, then animated the mouth movements either by hand or let the AI do that, too. Or at least that’s what they claim. Though these stills are from an earlier phase of production, and they seem to have both been redone since, so I can’t be totally sure. AI is pretty good at generating passable still images, though, and these types of cuts aren’t exactly super artistic in the first place, so if this is what’s going on, it works reasonably well. Still, this is only one drawing, so I can’t imagine it saves all that much time in the long run. Anyway, inevitably, they’ll then cut back to one of the CGI cuts and jump scare you with the juxtaposition between the two. The twins make several terrible Tik Toks, and we’re treated to three more rapidfire epic strats. photo to background art. They also created the background art primarily by slapping an AI filter on real photographs. There isn’t that much to say here. I think you basically know what to expect at this point. It generally ranges from inoffensive to uncanny valley nightmare. The colors are inconsistent from one shot to the next. Everything feels at once weirdly glossy and weirdly scratchy. If you look too close, textures like these stone slabs stop making any sense. And there’s often little sense of composition because they’ve just dropped the characters into a realworld environment. They really want to show off this crane shot which is just like that one in Freren. But they don’t want to work too hard on it. So instead of drawing separate layers to create a parallax effect, they just let the diffusion model have its way with the initial image. So, it takes these branches in the foreground and just smears them across the background, leaving carnage in their wake, including this little clump that originally looks like it’s part of this middle tree here, but then gets pulled into the foreground and becomes a lovely little cloud of leaves just floating there attached to nothing. And what’s happening to the path? It changes from light concrete to black asphalt. The AI suddenly decides to add some weirdass topology to it, and then it just sort of dissolves into a liquid by the end of the shot. I filter over liveaction footage. Now, this is a rare one. It only pops up a few times, but hands are hard. They’re hard to do convincingly in 3D, and god forbid you try to draw them. So, instead, they put the AI filter over liveaction footage. And boy, does it mangle this poor person’s hands. Raw CGI. Eventually, something actually happens when this kitty shows up looking all glitchy, and Hinana chases it across town into an area that she’s never seen before. Hear, who just said she was going to go to the library to do some homework, reappears and offers once again to practice their Tik Tok dances. The two walk off into the sunset in one of the worst looking shots yet. Have you ever seen somebody stretch like this? Cuz I haven’t. These arm movements actually look pretty reminiscent of what I got when I tested runway with my cat animation. So, I’m convinced they didn’t even use the 3D animation underneath in this shot and just figured if they shrink it down small enough and hide it behind some grass, no one will notice the wacky waving inflatable tube guy moves and that Hana’s collar briefly turns into a vest. Anyway, as it turns Oh, I’m going to spoil it, by the way. You’re not going to watch it. What am I saying? It turns out this isn’t the real Heari. There are actually two of them. This, of course, allows team Kaka to add a third character without needing a third character design or a third voice actor. Gee, it’s starting to seem like their only motive is to cut as many corners as possible at every single step. You say, “No, no, I, being highly educated,” respond calmly. They’re doing this all for the greater good. This is an act of altruism, empowering creators through the power of AI. It says it right there on their website. They want to foster a harmonious relationship where AI isn’t replacing animators, but acting as a tool to help empower them. At least the producers of this anime using AI are using it in good faith for the anime world. We are confident that the use of AI technology will make the future of anime production brighter and more sustainable. Isn’t that nice? So, who are these animators they’re empowering? Let’s see. Primary animator Yoske Kudo and oh, that’s it. Hm. Well, no. That’s a good start. Kaka creation empowering this one guy. Maybe with the next production they could empower like two or three guys. Well, okay, maybe I misunderstood. The animator’s creator, Isizuka Naomi, spoke at the Nigata International Animation Film Festival and explained that I never even thought of entrusting directing to AI. I thought it would be pointless to let AI do the interesting parts of video production, but that it’s totally fine to let AI do the jobs of the animators because they’re just following orders, mimicking. So, that’s not really a creative job anyway. That’s not the interesting part. Okay, I’m starting to get it now. So the line for which parts of the production are creative and worth keeping and which are robotic jobs we’d be better off without happens to be drawn just below the people who are making the decisions about where to draw that line. They want to empower the people at the top, the directors, the producers and executives to get more money and spend less of it on hiring people and empower the animators and background artists to find a new career path. Maybe one they’ll love even more. How selfless of them. “I’m sorry. I’m so sorry. I doubted Dr. Kaka’s intentions.” You whail while silently cursing the uninvited tears that are beginning to well up in your eyes. I say softly, “Be still. It’s okay. You couldn’t have known. Not with your condition.” “It’s really a shame because you know what? There actually are a few neat creative direction choices here, like the decision to abstract this dancing sequence by focusing in on little details like their shadows and close-ups of their shoes, or the way the camera whip pans back and forth following Hina’s gaze in this cut. I don’t really even know what it would mean to have an AI program direct an anime. But Naomi is correct that I’m sure if they had done it somehow, it would have been significantly worse. But imagine how much better this scene would be if in tandem with those direction choices, there were also cool animation choices, too. Like, I don’t know, some cool smearing as the shoe whirls past the camera instead of just the usual CGI run through an AI filter, then posterized down to 12 fps, which makes it so the foot just appears on the screen for one frame, then vanishes, making it virtually impossible for your eye to make sense of what’s happening in real time. Let’s come back to the flickering issue that they claim to have fixed because yeah, that’s not true either. I mean, to be fair, it’s probably better than most attempts at AI animation I’ve seen. It’s present throughout most of the episode, and a lot of the time it’s subtle enough to pass for wind or regular old boil, and there are even several cuts that would be totally acceptable in an okay traced 3D anime. But it often does get pretty rough. And this cut with Hana looking back and forth between the two Hemis is one of the best examples. She gets a new ear with each frame. Her clavicles are flapping all over the place. The bow on her uniform keeps changing colors. That was another problem I also had in my tests. I see that they haven’t been able to fix that. And you can very clearly see the lines on her head where they masked out a section with a paused single frame, presumably to prevent her face and top of her head from wobbling around like crazy. It has a really tough time in particular with the details in their shoes and their ears. Like in this cut where Heimari blinks her eyes and her ears. The second Heari runs away and the two twins chase her. The world gets pink and also crazy and sick and twisted. And suddenly I’m back on board. Here comes the second compliment you were promised because this is the good old days of AI video. Give me an airplane on a crane, a bus that that’s like furry and stuff and like Will Smith eating spaghetti. It’s a cleverly appropriate choice to use AI for its biggest strength, making weird, unsettling, freaky stuff, and set your anime in a world where weird, unsettling, freaky stuff happens, like this type of otherworldly melting effect. Only AI would do it quite like this and make it look so inhuman. And when they finally catch the impostor, that flickering problem is as bad as it’s ever been. Look at the way the shadows just completely pop in and out on the underside of her hair here. Imagine if they had only used the AI for this part, this side of the world, and actually capitalized on the contrast between the look of the real animation and the uncanny AI stuff. Although, if what you’re trying to do is market your technology and make lots of money by cutting costs while producing all kinds of anime, this doesn’t really sell me on the versatility of the tool. A blowtorrch is really good at making that brittle, sugary crust on the top of creme brulee. But I wouldn’t make you a creme brulee and then say that I’m going to revolutionize the restaurant industry by cooking everything with a blowtorrch and making it so no one has to buy an oven or a stove anymore, then ask you for several million dollars. The other he explains that this world is all like a simulation man and she, the future he created it. I swear this was made specifically to make Elon Musk bust in his ill-fitting jeans. Upon learning that their entire existence has been a lie, instead of having an existential breakdown, the twins are like, “Okay, yeah, sure, that makes sense.” And do the insufferable thing where you point out the unoriginality of your own story without actually fixing that problem. They then have this completely unearned emotional heart-to-heart moment, and future Hear yeetss them both into the void, where everything gets reset back to the start of the episode. Well, that was not very good, huh? It’s obvious that Kaka’s goal is not to make something good, but to make something as cheaply as possible and generate as much hype as possible so they can secure investment money. No, this is not a 95% AI original anime. It’s a motion capture CGI anime with an AI filter pasted over the top. The AI didn’t create anything or do almost any actual animation except by the looks of it, this ending credits sequence. And all over the place, you can find evidence of somebody doing a sloppy job of masking out sections to make what the AI did look more acceptable, which sounds like a soul crushing task. It’s only one episode, and they did work on it for almost a year and a half. But they did obviously manage to produce this with a lot less manpower than usual. There are a lot fewer credits, but most of that savings seems to come from regular old cost cutting measures. only two character models and two voice actors and using whichever technique is easiest for any given moment or element, which makes the final product look like an inongruous mess with no cohesive style tying it all together. It says right here on their website, their goal isn’t to solve the problems with overwork in the anime industry. It’s to pump out an even higher volume of anime with the same resources and make even more money while returning profits to creators, which we now know means only the people at the top. Personally, I think there’s already a lot of anime being produced. Maybe they should try making a little bit less of it and focus on making it good. Oh, and did I mention they also have a page on the website for this nonsensical interview with Makoto Tzuka, Osamu Tzuka’s son, who is not involved in this project, but did cannibalize his own dead father’s work to create an AI manga, where he basically says, “Yeah, AI is cool, and people won’t care if we just don’t tell them about it. Western animation doesn’t convey emotion.” And yeah, my dad did say that the eroticism of anime comes from people drawing things that flow into each other over time, but uh we can fake it with AI. And I think he would have used AI anyway. I assume they put this interview here in an attempt to vaguely associate themselves with the name of one of the pioneers of Japanese animation. Gross. I don’t fundamentally disagree with the idea of finding ways to use AI as a tool to assist rather than create. And if I had to guess, I would guess that after the hype bubble bursts and the dust settles, we’ll be left with a handful of limited tools that can increase speed and efficiency when employed judiciously in contexts where they don’t encroach too much on actual creativity. But to use the phrase empowering creators as a smoke screen while you cut out animators from your process, cut corners at every possible turn, and cynically pump out the absolute bare minimum of what can technically be considered art with the sole goal of further concentrating profits in the hands of executives while misrepresenting what it is you’re actually doing. Well, that’s pretty cringe. Pretty kaka if you ask me. Luckily, we don’t have to be too worried because once again, it looks like absolute garbage and nobody likes it. I don’t hate this simply because they used AI. I’ve used AI motion tracking before, but I didn’t claim to have made a 95% AI generated stopotion short and raise a bunch of corporate funding off that. And I’m also no stranger to making things that look kind of cheap and thrown together. But I’m also not claiming to revolutionize an industry. I hate it because it’s generic, boring, ugly, and each cut looks like it’s from a different show than the one that came before it, and you couldn’t be bothered to hire anybody to fix any of these problems. Instead, you just hoped nobody would notice. But it’s got cute girls in it, so I guess that’s good enough. Boy, this one started out light-hearted, but the more I researched this company, the more I was possessed by a rancerous fiend. Let me know your thoughts in the comments. And if you found this interesting, you can subscribe to the channel for more videos like it, or check out this next one that’s going to pop up on the screen in just a second. And a big thank you to my generous supporters on Patreon and their immaculate $3,000 haircuts. Smacktosh, Scorpione, Katy L, Retro Heime, Serial, Leapico, Nutaf, Patrick Davidson, Arbor, Ginger June, Ginger June, Dr. Yolo, Mcwag, Grav G, Squid Goo, Joe, Fugnugget, Annie, and Sammy Z, as well as all the other names you’re seeing on the screen right
You can go to my sponsor https://aura.com/swampjawn to try 14 days for free. That’s enough time for Aura to start scrubbing your personal info off these data broker sites, without you lifting a finger.
Patreon: https://www.patreon.com/SwampJawn
Tumblr: tumblr.com/swampjawn
Bluesky: https://bsky.app/profile/swampjawn.bsky.social
Music from #Uppbeat (free for Creators!):
https://uppbeat.io/t/simon-folwar/neon-signs
License code: DK8XEFXNWFSHQPRB
https://uppbeat.io/t/soundroll/all-the-things-you-love
License code: YUCYLJ3WLCUTJIWV
Sources:
https://anime-hinahima.com/
https://www.animenewsnetwork.com/encyclopedia/anime.php?id=34172
https://www.animenewsnetwork.com/news/2025-02-28/frontier-works-kaka-creation-twins-hinahima-ai-anime-reveals-march-29-tv-debut/.221769
https://screenrant.com/japan-first-ai-anime-defends-criticism-before-release-twins-hinahima/
https://www.kakacreation.co.jp/
https://fandomwire.com/anime-production-is-made-up-of-imitation-iizuka-naomichi-thinks-japans-first-100-ai-anime-is-a-beacon-of-opportunities-for-animators/
https://prtimes.jp/main/html/rd/p/000000001.000132834.html
@marcusyellsanime THE FIRST AI CREATED ANIME EVER! “TWINS HINAHIMA” WILL CHANGE THE FUTURE OF ANIME FOREVER! #anime #animerecommendations #newanime #anime2025
_____Video Contents_____
00:00 – Intro
00:46 – The World’s First 95% AI Anime!!!
02:32 – …but not really. (technique 1)
07:44 – technique 2
08:34 – technique 3
09:48 – technique 4
10:07 – technique 5
11:07 – The company behind it
13:21 – Choices in direction and animation
15:13 – Let’s Get FREAKY!
17:23 – Why calling this an “AI Anime” is misleading
19:37 – Conclusion