Back to TheSmug1's Profile TheSmug1's Profile

Twins HInaHima is made almost entirely with AI. For some people, that's probably enough to drop it like a hot rock and try something else, but I thought it would be interesting to watch it and see if there was any meat to the concept. After all, regardless of all the internet moralizing, I don't think it's likely that this sort of technology is going to disappear, and it's best to see just how deep the possibilities might be.

Unfortunately, it isn't so deep as it initially appears, much like a puddle in the road after a storm. It turns out that "AI" was just used ... to draw 2D animation over 3D model animations, which were done by a human being using Unreal Engine. That said, Twins HinaHima does not appeal to me, aesthetically, in almost any way. The characters' outlines jump around between frames, and their movements (similar to 3DCG anime) are uncanny and only vaguely "human," but without the stylistic dissonance normally present in anime to offset it. It looks rotoscoped, in other words, which is pretty displeasing to the eye.

There is a story here, and they might have used the story to implement the technology in a more interesting and immersive way, but they opted instead to go all-in on it. Our stars, Hinana and Himari (whose voices one could be forgiven for thinking were also created by AI, as they seiyuu are not giving an inspired performance here), really want to go viral online. They try all kinds of silly videos but none of them catch on. Himari, the older twin, gives up and heads back to school to study, despite Hinana (the one really interested in going viral, to make life in her quiet town more interesting) protesting. While Himari is gone, Hinana finds a strange cat whose appears to flicker in-and-out of existence. She chases the cat, but falls/ambiguously gets hit by a truck, and awakens to find Himari, who is suddenly and uncharacteristically interested in going viral again, They practice their TikTok dance, and then another Himari shows up - the real Himari, and the other begins to flee.

The twins chase the fake older sister across a bridge into another world, despite her warning them not to follow and that it is dangerous, and the other world is very peculiar - signs, streetlights, and the scenery are all unlike anything in reality, and Hinana is interesting in further pursuit because she thinks it's way more interesting than reality (or so it is implied). This is the part of the story I think the studio could have used the AI to novel effect. AI animation is nothing if not uncanny, and using its inherent weirdness to offset the mundanity of the "real world," as well as to contrast fake Himari with real Himari, could have gone a long way toward making this more appealing, or at least give it more impact. As it is, the differences between the two worlds are only superficial, rather than fundamental.

Anyway, the twins chase the fake Himari to a cliff's edge, where she reveals she is the real Himari, from the future. They talk for a bit about the nature of reality, then future Himari says it's time to go. They hug, and future Himari throws the twins into a wormhole to send them back in time to the start of the episode, where again they try to go viral. Now future Himari comments on their videos - at least they have one fan. It's implied that Hinana is killed by the truck I mentioned earlier, but never outright stated. I think Himari went back in time to revive her, but that's just a theory of my own - there is no future Hinana, after all.

This story is functional. It isn't great, or deep, or noteworthy, but it mainly serves as a vehicle with which to show off this burgeoning technology. This is also true of the character designs, and the characters themselves, which are both quite bland (the designs could be those of mob characters in another series, and there isn't much to them beyond the bare minimum in terms of characterization).

The technology itself is still lacking. Even with the amount of coaxing and massaging which must have been necessary to make this look as it does now, it looks slipshod - not for the same reasons that an anime episode usually does (and, visually alone and regardless of whatever moral hangups you might have, it is superior to something like TamaYomi, whose production team couldn't even be bothered to draw backgrounds halfway through), but because there is an incomplete connection between two sequential frames. The AI simply doesn't seem to have the backbone necessary to process sequential images without invariably "missing something."

There is a niche of people who like to grandstand against AI as a construct, and while I am not one of those people, I do not welcome the total subsumption of anime as a medium using the technology. I also don't think this step really makes sense, though I am comfortable with using it to make 3DCG look a little more like proper 2D anime in those cases where the two media tend to clash.