Sekiro, AI, and Auteur Directors – This Week in Anime

As Steve says, “A studio does not make an anime. People make anime.” But what tools should those people be using? Chris and Steve discuss.

Disclaimer: The views and opinions expressed by the participants in this chatlog are not the views of Anime News Network.
Spoiler Warning for discussion of the series ahead.

Chris

Truly, AI discourse is the Dark Souls of anime discussion. I’ve grappled with the scourge myself a couple of times, including with services experimenting with using it for translation. How’s that gone since? Ask Necronomico and the Cosmic Horror Show.

But this latest digital dustup is happening under its own unique circumstances. Specifically: a highly anticipated project getting announced, then hit back by fans merely for the possibility of using AI, to the point that, like an online artist posting their timelapse, the staff felt the need to put out a statement confirming this project was, in fact, made with human hands.

I’ll admit that I got my dunks in at the outset, though it was more on Qzil.la as a company, or, as you mentioned, touting their overall use of AI. They’ve also incorporated blockchain cruft before, just for an extra dose of techbro terribleness. It’s not an ideal look for an adaptation of a game renowned for the intentional artistry of the craft behind it.
The trailer itself, to my eyes, looks pretty good, and even features a credit roll at the end that seems to emphasize the sheer force of experienced artists and animators coming together to work on this. The company sucks, but they’ve brought in some good people here!

I’ll always be a little galled that they can use the same name as the other, more appreciable, anime-adjacent Azuki.

It’s a striking look that lends its own additional flashes of color and artistry to the base Sekiro art style. I love the painterly backgrounds too, which still feel like they carry that all-important intentionality.

It feels like there was some confirmation bias at work in how some viewers rushed to prove that the animation seen in the trailer was the product of generative AI once it was revealed that Qzil.la was involved in such technology. Thus, you had people playing the old game of grabbing out-of-context in-between frames as “evidence.” I get trying to be sure once the suspicion’s been planted. Still, it also demonstrates a lack of understanding about how animation works that I’ve been seeing from the sidelines, at least going back to folks complaining about Shingo Yamashita‘s work on the Pain fight in Naruto Shippūden.

One of my major concerns—and the one that fueled my intense response to this issue—is that any slightly deviant art or animation style will be subject to accusations of AI use. Anything that looks “off” will be stigmatized. If Masaaki Yuasa debuted in 2025, would people accuse him of using AI? I’ve seen similar witch hunts unfold in fan art spaces, and I don’t want to see them consume every avenue of art.

And I get it. I’m jumpy too. If I see a sus bit of art cross my feed, I start scrutinizing the details. But I don’t want us to forget that things are complicated. I saw this crop come up as potential evidence of AI use in the Sekiro trailer, because we’ve all internalized the heuristic that AI = extra fingers.

In reality, it was always far more likely that this was a coloring and/or lineart error on an in-between. Something that happens all the time.

It sucks, because in the before-times, that sort of momentary askew detail would be a neat reminder of the human touch behind this craft, rather than something for people to jump on and accuse them of outsourcing the animation to the plagiarism machine. Funny in-between frames are a mark of the medium, and no bad-faith generative animators can take something like Wide Superman from me.

There’s also an irony in that imperfection being a focal point, since a defining element of much AI imagery clogging up fanart feeds in the anime space is that it looks too “perfect”: immaculately airbrushed, sheen-finished anime babes with nary a hint of the handcrafted human touch.

Which is why I think it’s also useful to clarify what AI use in animation actually looks like right now. Because we have examples, and none of those remotely resemble Sekiro. And yes, I hate looking at this stuff too, but I suggest you follow my lead and direct that hatred towards popping the AI industry bubble that much faster.

Twins Hinahima is an illuminating example because they posted a video that partially reveals their process of creating it, allowing us to break down what AI is actually doing. That’s helpful when most companies handwave their sausage-making with platitudes about technology and the future.

Right away, we can see that they process the storyboard drawing with AI, and (imagine a buzzer sound here) whoops, looks like you needed significant human input on your very first step! You’re not getting a legible, consistent storyboard from the stable diffusion ether. You need a person for that.

It’s one of those things that you’d think would be obvious to anybody even adjacent to the animation industry. Storyboards aren’t just busywork that needs to be gotten out of the way before the “real” animation can begin—they’re recognized for the guiding hand they provide to the process, being able to dictate important elements like placement, framing, and pacing of movement. Yet, for some strange reason, storyboards are something even larger, more well-known animation studios are discussing wanting to offload onto AI.

It’s a perfect example of the pie-in-the-sky thinking holding up the entire industry. Because AI can (poorly) perform a basic task, its proponents argue that it can or will be able to perform a significantly more complex task. It’s inevitable. All we need is one trillion dollars right now.

A big ask from an industry where studios hardly want to pay their animators enough to live.

That video also shows them processing a photograph into a cartoon background, which I’ll admit is a far more realistic application of what AI can do.

Amazing that even just filtering still utterly mangles the contents of that vending machine there. You know I’m jonesing for what appears to be a dozen irregularly shaped boxes of band aids.

You don’t need AI to do that! Anime have been filtering photographs and compositing those into their backgrounds for ages. Here’s a recent example from My Dress-Up Darling. No AI required. Just the good old computer.

Furthermore, Hinahima admits that it relies on human touch-ups and quality checks. So, how much time, effort, and money is truly being saved here?

It’s also worth noting that they made one standard-length episode in this fashion. Not a full season. One episode. Almost like this was more of a stunt than a workflow, you can scale up appropriately.

It’s certainly a proof of concept more than anything else, and the concept it’s proving is that using AI for animation isn’t really worth it. To say nothing of the fact that, as the Sekiro situation showed, audiences are energized to outright reject anything they think even might be AI. They haven’t bothered getting Hinahima streaming anywhere official here, and between you and me and the TWIA readers, there’s not even a proper English fansub immediately findable. People don’t want it, and they seem to be aware of this.

I found one fansub, and it was a joke one done using AI transcription and machine translation. I respect the bit. Even if you look at those trailers, they have an abysmally low likes-to-views ratio, and they’ve turned the comments off. The people have (not) spoken.

Moreover, despite every criticism I’ve laid on it, I think Hinahima is by far the best current example of AI integration in anime that I’ve seen. It’s not good, mind you, but it seems like the most honest one. They’re at least somewhat cognizant of the technology’s limits, they utilize other techniques like motion capture, and they put in an effort to make it look better than slop. I think that’s all a waste, but credit where it’s due.

And I mostly say that because the bar for AI animation is buried somewhere in the Earth’s mantle. It doesn’t take long to find another AI project that has garnered recent headlines, and this one makes Hinahima look like a Vermeer painting.

I felt bad for even indirectly contributing to people getting the wrong idea about Sekiro and criticizing the people behind it, so part of me wants to hold off on making incendiary remarks again in this same space. But honestly?

This looks like shit.

It is utter slop. Ugly hyperreal nonsense that even has the gall to be narrated by a soulless AI voiceover. I’m sorry you all have to see this. You can’t afford to look away. You need to understand that this is what the world economy is riding on right now.

Incidentally, I love that no two shots of the grandma look remotely alike.

Same with the monkeys. There’s no art style because there’s no art. It really does look like a loose collection of shots thrown on social media that merely serve as examples for whatever Sora or other soulless corporate automaton being pimped this week can do.

These sorts of smeary sliding cutouts with lip movements that would make Clutch Cargo weep are an insult when they’re frontpaged on DLsite asking for money. And these doofuses dare to blurt them out in a feature film? I am not sorry they seemingly lost their play for a proper release in Japanese theaters last year.

Scientists so preoccupied with whether or not they could that they never etc etc.

Again, you mustn’t look away. Especially if you thought, even for a moment, that you saw telltale signs of AI in the Sekiro trailer. AI companies, including Qzil.la, want you to think that they are capable of making the Sekiro trailer with just a text prompt and a dream. They’re relying on that. But generAIdoscope skews far closer to these programs’ actual capabilities.

It’s a night-and-day comparison that makes the artistry behind what we’ve seen of Sekiro shine even brighter. And as you’ve said, it’s a teachable moment for folks on either side of the screen, as the Qzil.la suits have seen what the response will be if they actually do use AI for animation. I hope they’re also seeing the enormously diminished market reactions to Twins Hinahima and generAIdoscope to give them additional pause.

This applies to any other studio as well, as even a giant like Toei is reportedly exploring AI use for tasks such as creating storyboards for series like Pretty Cure. It’s shocking not just because of the value of the human touch on storyboards, which we have already covered, but because Pretty Cure itself has regularly been a proving ground for animators at the studio. Seeing a well-considered backlash might encourage them to reconsider.

And that’s fine! Believe it or not, I’m not anti-AI experimentation, and I do think there may be limited use cases for it in the future, particularly in a creative environment controlled by artists, not executives. But the burden of proof is on the corporations, and, so far, they haven’t shown me anything besides greed and delusion. And a vacuum of taste.

Exactly. Companies like Qzil.la claim to be “maximizing value” and “evolving production processes,” but in reality, they are poisoning the well for creatives, making them more susceptible to less-informed viewers hurling accusations at them. The sheer irony is that, as what we’ve seen of Sekiro so far shows, the path to success really is as simple as hiring exceptional artists and letting them create to their own standards—probably for ultimately less than they’re throwing at genAI startups and processing.

Sekiro may be based on a game known for being hard, but this part seems pretty easy.

Indeed. And hopefully, the next time we discuss Sekiro, we can focus on the aspects that are actually germane to the narrative, like spamming the L1 button.

Less horrifying AI-generated monkeys, more giant monster monkeys getting their heads cut off, please.