Part 23: The Jelly Donut: The One Thing the Art Refuses to Let You Localize

A boy is holding a triangle of white rice wrapped in a band of black seaweed. It is on screen. It is unambiguous. It has been drawn by people who have eaten one. And the voice coming out of the boy says: jelly-filled donuts.

This is the most famous localization failure in the history of anime in English, and it has outlived every careful decision made in the same decade. 4Kids Entertainment's Pokémon dub is otherwise mostly forgotten as a piece of craft. The donut is immortal. It is worth asking why, because the answer is not the one people usually give.

The category error

Part 6 established the hardest structural fact in this series: Japanese sound effects are drawn. They are not a caption laid over the art; they are ink, composed into the panel, part of the same gesture as the line. You cannot localize them without redrawing the page, because there is nothing to swap out — the word is the picture.

Food is the same problem wearing a friendlier face, and it is worse.

The localizer's basic instinct — a good instinct, the one that makes dubs watchable — is substitute a local equivalent. A joke that doesn't land becomes a joke that does. A pun on a Japanese homophone becomes a different pun on an English one. A reference to a variety show nobody's seen becomes a reference to one they have. This is legitimate work. Done well it is invisible, and Part 21 argued that invisibility is the job.

It is invisible because nobody can see the original. The English viewer has no access to the joke that was replaced. The substitution leaves no evidence.

Food leaves evidence. Food is drawn. The rice ball is a triangle, it is white, it has a black band, and no amount of confident voice acting will make it round or brown or filled with jam. 4Kids reached for the substitution instinct and applied it to the one category of thing where the substitution is falsifiable by looking at the screen.

So what happened in that scene was not translation, and it was not even localization. It was narration that contradicts the image. The picture says onigiri. The voice says donut. And the child watching — this is the part that stays with me — learns something true, which is that adults will lie to you about food, on television, while showing you the food.

Why they did it, which is not stupidity

It's easy to be contemptuous here and I want to resist it for a paragraph, because the reasoning was not insane. 4Kids' brief was a children's cartoon on American broadcast television in the late 1990s, in a regulatory and commercial environment that treated any unexplained foreignness as friction. The theory was that a child would trip over an unfamiliar food, and that tripping costs you the child. So: make it a thing the child already knows.

The theory is wrong, but it's wrong in an interesting way. It assumes the child is fragile. It assumes that meeting an unknown object in a story is a cost to be minimized rather than the reason children read stories at all. Every kid who watched that scene knew perfectly well they were not looking at a donut. What they took from it was not confusion about rice. It was the discovery of a lie, and — decades later, on the internet — the discovery that the lie had a name and a company attached to it.

4Kids later escalated: onigiri digitally repainted into sandwiches and cookies, frame by frame, which at least has the virtue of internal consistency. Fix the picture and the lie becomes unfalsifiable. That is the logical end of the substitution instinct, and it is also, precisely, Part 6's redrawn sound effect — except that the redraw is being used to conceal rather than to carry.

The honest version, and how strange it is

Now the good-faith case, because there is one, and it is the more unsettling half of this essay.

Oishinbo — written by Tetsu Kariya, drawn by Akira Hanasaki, running from 1983 in Big Comic Spirits — is a manga about food journalism that ran past a hundred volumes before going on an extended hiatus in the mid-2010s, following a storyline about radiation and Fukushima that generated real political controversy. Set that aside; what matters here is the shape of the thing. It is a hundred-plus volumes of Japanese people arguing, at length and in earnest, about dashi. About the correct water. About whether this rice is that rice.

A straight translation of Oishinbo into English, volume 1 through volume 111 in order, is close to unpublishable. Not because the sentences are hard — they're not — but because the work assumes a reader who has opinions about dashi already. It is a work rooted so deep in its own soil that pulling it up kills it.

So Viz did not translate it. Viz recompiled it. Oishinbo à la Carte is seven volumes, each one thematic — sake, ramen and gyoza, fish, vegetables, rice dishes, izakaya food, Japanese cuisine broadly — assembled by selecting chapters out of the hundred-volume corpus and resequencing them into a shape an English reader could enter.

That is not a translation. It's a curation. It is a museum's selection, hung by theme, drawn from a collection the visitor will never see. And it worked: it is the reason Oishinbo exists in English at all. Part 22 spent its length on the cost of works that never crossed, and here is a work that crossed only because someone was willing to take it apart first.

The uncomfortable comparison

So there are two responses to food that won't localize.

The first is to lie over the top of it. It fails, visibly, immediately, and forever — the jelly donut is a meme precisely because the evidence was on screen the whole time. The lie was checkable, and so it got checked, and so it got mocked, and 4Kids' reputation among the people who grew up on that dub is what it is.

The second is to rebuild the container. It succeeds. It is respectful, careful, made by people who love the work. And it is invisible. No English reader of à la Carte knows what order those chapters ran in, what the connective tissue was, what happened in the hundred volumes that weren't picked, or what a hundred volumes of accumulated argument feels like as a single continuous act of reading. There is no scar. There is nothing to check.

Which is worse?

I don't think it's a close call on intent — one is contempt for the audience, the other is care for the work. But look at the accountability, because Part 21 set this trap and I'm going to walk into it. The translator's success condition is invisibility. The jelly donut was visible, and being visible is why we're still talking about it, and talking about it is a form of correction: no dub does that now, and the reason no dub does that now is that everyone saw the triangle. The careful thing is invisible, and invisible means unaccountable. The reader who has read à la Carte believes they have read Oishinbo. Nothing on the page tells them otherwise.

The lie got punished because it was legible. The honest reconstruction goes unexamined because it isn't. That is not an argument for lying. It is an argument that our whole system of noticing runs on damage being visible — and it is going to matter enormously in the next part.

The industry disproved itself, again

There is a clean refutation of the 4Kids theory and it took about fifteen years to arrive.

Shokugeki no Soma — Food Wars in English — is a manga and anime whose entire content is Japanese food. Not food as a background detail, not a rice ball held in one shot: food as the plot, the conflict, the dialogue, and the spectacle. Characters deliver lectures on technique. The drama is whether the dashi is right. It is, by 4Kids' logic, the single most unshippable property imaginable — a show that cannot function unless the audience is willing to sit still for a Japanese cooking argument.

It crossed to English without anyone repainting a thing, and it was popular.

This is Part 22's shape exactly. There, the industry decided the English-language reader was a boy who wanted motorcycles, and held that belief for four decades until teenage girls in bookstore aisles demolished it. Here, the industry decided the English-language child could not tolerate an unfamiliar food, and held it until a generation of viewers demonstrated that the unfamiliar food was the appeal. In both cases the model of the audience was not merely wrong; it was wrong in the specific direction of underestimating what people would happily accept. Nobody was ever protected by the jelly donut. There was no fragile child. There was a company's guess about a fragile child, and the guess got shipped.

The pattern is consistent enough by now that it deserves a name. Every time this industry has assumed the audience needed the foreign parts filed off, the audience turned up wanting the foreign parts. The strangeness was not a tax on the work. It was substantially what people were paying for.

The numbers

Onigiri reads Destiny 9, Heart 33, Personality 3 — Humanitarian & Sage, with Master Teacher sitting in the heart. Rice ball reads Destiny 8, Heart 6, Personality 2. The Japanese word and its English gloss share nothing. Not one number.

Which is at least honest about something: "rice ball" is not a translation of onigiri. It's a description of it, performed for someone who has never held one — the way "raw fish" is a description of sushi and tells you nothing you'd want to know. And "jelly donut" isn't even a description. It's a third thing, unrelated to either, asserted over the top of a picture of the first.

Then the engine does something I could not have arranged.

Rice ball: Destiny 8, Heart 6, Personality 2. Viz Media: Destiny 8, Heart 6, Personality 2.

A clean match — the seventh in this series. The largest publisher of manga in English has the identical numerological reading to the food that American television would not say the name of. I ran it because I was curious about the publisher, not because I expected anything, and the arithmetic handed me the joke.

It means nothing. Seven letters and eight letters happened to sum the same way. If Viz had been named anything else it wouldn't have happened, and Viz's name is a fact about a company in San Francisco, not about rice.

But notice what the engine is doing while it produces that joke, because it is doing something very specific. It takes a Japanese word. It converts it into a string of Latin letters — a transliteration, a thing built for English convenience, exactly as Part 22 found. It performs arithmetic on the letters. And then it announces, fluently and without hesitation, what the thing is: Humanitarian & Sage, Master Teacher in the heart.

It has never held one. It does not know the seaweed is there to keep your fingers dry. It has never tasted the rice.

That is the same move 4Kids made. Look at the surface, convert it into terms you're comfortable with, and announce with total confidence what it is — while the actual object sits right there in the frame, being a triangle, saying nothing, waiting for someone to look.