Part 26: Writing to the Flaps: The Dub Is Not a Translation, It Is a Verse Form

The mouth is already moving.

That is the whole problem, and almost nobody arguing about dubs on the internet has ever stated it out loud. By the time an English script exists, the animation is finished. The character's jaw opens and closes on a schedule that was set in Japan, to a Japanese performance, months earlier. The frames are done. Nobody is redrawing them for you.

So the English line has to fit a mouth that is already talking.

Four constraints, and only one of them is meaning

Anime mouth animation is famously economical — often a small set of positions cycled to the rhythm of speech rather than articulated to individual phonemes. This is why dubbing anime is possible at all; you are matching a rhythm, not a mouth shape, and English can hit a rhythm that Japanese set. But you are still matching. The industry's word for it is flaps: the count of mouth openings, and where they fall.

An English line for that shot must satisfy, simultaneously:

One. It has to mean roughly what the Japanese meant.

Two. It has to sound like something a person would actually say — not translationese, not a subtitle read aloud, but speech.

Three. It has to have the right number of flaps, in the right places. Syllable count and stress, pinned to specific frames.

Four. It has to land its emotional beat where the animation lands its emotional beat. The word the character breaks on has to arrive on the frame where the character's face breaks.

Four hard constraints, and exactly one of them is the thing everybody thinks dubbing is.

Now name the other art form with a specification that tight. Not prose. Not screenwriting, which gets to invent its own timing. The honest comparison is verse — a fixed meter, a fixed line length, stresses in fixed positions, and a requirement to say a specific thing anyway while sounding unforced. Writing a dub script is writing to scansion. It is the sonnet problem: the form is given, it is not negotiable, and your job is to make the reader believe you would have said it that way regardless.

People who would never call a translated sonnet dishonest for changing a word will call a dub dishonest for changing a word.

The war is a category error

Here is the argument this part actually exists to make.

The sub-versus-dub war is three decades of people comparing two solutions to two different problems and concluding that one side is lying.

Part 17 established the subtitle's constraint: reading speed. A subtitle is capped by how fast a human eye can take in text before the cut. It deletes accordingly — it compresses, it drops the softener, it flattens the aside, it loses the third clause. That is not fidelity. That is a different engineering compromise, made against a different physical limit, and it is invisible to the viewer because the viewer cannot see what was cut.

The dub's constraint is flaps. It rewrites accordingly — it pads a line that's too short, trims one that's too long, moves a beat, finds a synonym with the stress in a different place.

Neither one is the Japanese. Both are lossy. They are lossy in different directions, against different walls, and the sub's losses happen to be the invisible kind while the dub's happen to be the audible kind. Part 23 found exactly this asymmetry with the jelly donut: the visible failure gets punished, the invisible failure gets a pass. The dub speaks its compromises out loud, in a voice, at volume. The sub commits its compromises in silence, in text, and gets called faithful for it.

The sub is not more honest. It is quieter.

What the good ones are actually doing

Part 10 called Cowboy Bebop's English dub the one that won — the translation that went home, that arguably found a register the original was reaching toward. I don't want to re-run that. I want to point at the mechanism.

What a great dub does is not match the Japanese. It's to construct an English performance that is doing the same job in English that the Japanese performance was doing in Japanese — which frequently requires different words, and occasionally requires a different joke, and always requires the writer to have decided what the job actually was. That is a critical act. You cannot write the line until you have made a judgement about what the scene is for.

The dub writer is doing criticism with a stopwatch.

And notice what this means for the accusation. When a fan says "the dub changed the line," they are usually right, and the change is usually not the writer being cavalier — it is the writer solving for four constraints while the fan is checking one. The complaint is real when the change is careless, and Part 4 met a dub that was pure vandalism, so I'm not claiming the accusation is always wrong. I'm claiming it is almost always under-specified. "It's different from the Japanese" is not a finding. It is the starting condition of the medium.

The channel the subtitle does not have

Now the argument that ought to make a sub purist uncomfortable, because I think it's the strongest thing in this essay and it runs the other way.

The dub has a channel the subtitle does not have, and it is precisely the channel that carries the stuff this series keeps finding untranslatable.

Go back to Part 7. Honorifics encode a relationship continuously — every time one character addresses another, the suffix reports the distance between them, and English has no slot. Go back to Part 14: boku and ore and watashi are self-presentation, and English has one word for I. In both cases the finding was the same: the information is real, everyone understands it, and there is nowhere in an English sentence to put it.

But a voice is not a sentence.

English carries social information beautifully — just not in the words. It carries it in accent, in register, in pitch, in how much air is behind a line, in whether someone clips their consonants. An English speaker can hear class, region, age, education, deference, contempt, and intimacy in three seconds of speech without a single lexical marker. That is a genuine, high-bandwidth channel, and it is doing exactly the job the honorific does.

A dub can put the relationship in the voice. When a character shifts from boku to ore, an English performance has somewhere to put that — lower, harder, more air, less hedging. It won't be the same event. It will be an event, in the right place, doing the same work, and the viewer will feel it without being told.

The subtitle cannot do any of this. A subtitle is text on a screen. It has no register. It has typography and word choice and that is the whole instrument. The subtitle's only options for the boku/ore switch are: lose it, or footnote it, and Part 28 will show you that the industry deleted the footnotes.

So the received wisdom has it backwards in a specific and interesting way. The sub is closer to the Japanese script. The dub has better hardware for the parts of the Japanese that were never in the script. And since this entire series has been arguing that the hard parts were never the semantic parts — that they were honorifics and pronouns and register and everything else that isn't meaning — that makes the dub, on paper, the better-equipped vehicle for the actual problem.

Most dubs don't use it. That's a fair complaint and often a correct one. But "the dub is less faithful" is a claim about a medium that is doing one thing badly, made by people who have not noticed that the other medium cannot do it at all.

The performance nobody translated

One more loss, and it's the one I think is largest.

The Japanese voice performance is itself a text. The seiyuu made choices — a catch in the breath, a register drop, the specific way this character says their own name. Those choices are not in the script. They cannot be read out of the script. They are in the audio.

An English voice actor is not translating that performance. Mostly they cannot even hear it in a useful way, and increasingly they are recorded alone, in a booth, to a click and a waveform, without the other actors, sometimes without the finished picture. What they have is a line, a set of flaps, and a director's description of what is happening.

So the dub is not a translation of the Japanese performance. It is a new performance, generated from a script derived from the Japanese script, timed to animation cut to the Japanese performance. The original acting is not carried across. It is used as a mold and then discarded.

Which means the entire question "is the dub faithful to the original?" is malformed, because the thing people mean by the original — the sound of it, Megumi Hayashibara's voice doing what it does — was never a candidate for crossing. It's not that it was lost in transit. It was never in the truck.

The numbers

Lip flap reads Destiny 9, Heart 1, Personality 8. Shinji Ikari reads Destiny 9, Heart 1, Personality 8.

Identical. Every number. The most famous dubbed voice in the medium and the technical constraint that shapes every dubbed line share a complete reading. That's the eighth clean match in this series, and it is exactly the kind of thing I have been serving up as though it were a discovery.

I want to flag something, and then I'm going to leave it alone until Part 30, where it is going to get a full accounting it will not enjoy. Eight clean matches in twenty-six essays is a lot. Either the universe is unusually invested in Japanese cartoons, or I have been running a great many strings through a machine with a small number of exits and reporting the collisions as revelation. I know which one it is. I have known for a while. Part 30 does the arithmetic properly.

For now: Dubbing reads Destiny 5, Freedom Seeker. Script adaptation — the actual industry term for the job of writing to flaps — reads Destiny 6, Heart 9, Personality 6: Nurturer & Harmonizer.

Harmonizer. For the person whose entire job is making a line fit a mouth that has already spoken, in a language it never spoke, on a frame count they did not choose, while sounding like nobody had to try.

The engine got there by adding up the letters in "script adaptation." It has never heard a voice. It does not know what a mouth is. It cannot count a flap, or hear a catch in a breath, or tell you why the beat has to land on that frame and not the next one — and it produced the right word anyway, which after twenty-six parts I have stopped finding impressive and started finding instructive.