Part 17: The Compression Nobody Notices: What Subtitles Delete

The sub-versus-dub argument is the oldest fight in this fandom, and it is conducted almost entirely by people who have not noticed what a subtitle is.

The standard position: dubs are a compromise for people who cannot be bothered to read, subs are the original with a helpful line of text underneath, and a serious person watches subs. The first clause is snobbery, the second is false, and the third follows from the second, so it is worth dismantling the second.

Because subtitles are not a transparent window onto the Japanese. They are the most aggressively compressed translation in the entire trade, performed under constraints so severe that professional guidelines are written in characters per second, and they routinely throw away a third of what was said.

The Arithmetic of a Subtitle

A subtitle has to be read. That single requirement generates the whole problem.

Reading is slower than listening, and a viewer is also trying to watch a film. So the industry works to reading-speed limits — the widely used figures land around seventeen characters per second for adult programming, with a maximum of two lines and something like forty-two characters per line. Those are not stylistic preferences. Below those speeds the line is comfortable; above them, viewers start missing text, or fixate on it and stop seeing the image.

Now put a Japanese line into that box. Japanese can be extremely dense — a compact clause can carry the propositional content, the speaker's gender and class through the pronoun of Part 14, their relationship to the addressee through the honorific of Part 7, and their emotional register through a sentence-final particle, all inside a couple of seconds of speech. Rendering all of that explicitly in English takes far more than forty-two characters. It does not fit. It is not close to fitting.

So the subtitler cuts. Not occasionally — as the fundamental operation of the craft. Subtitling literature has long put the typical reduction somewhere in the range of a fifth to two-fifths of the source. Call it a third, on average, gone.

And it is not random third. You cut in a priority order, and the order is forced: propositional content survives, because without it the plot stops. What gets cut is everything else — hedges, repetitions, the second half of a redundant pair, discourse particles, the hesitations, the softeners. Which is to say: subtitling systematically deletes exactly the material that carries character and leaves exactly the material that carries plot.

A hesitant, over-polite man and a blunt one, saying the same thing, converge on the same subtitle. The information that they were different people was in the padding, and the padding is what a subtitler is trained to cut, because the padding is where the extra characters are.

That is not a failure. It is the craft working correctly. Every good subtitler knows precisely what they are throwing away and throws it away anyway, because the alternative is a viewer who reads instead of watching, which loses more.

And It Sits On the Picture

The second cost is the one nobody mentions, and this site spent twenty-four essays on it without connecting them.

The Grammar of the Screen series is about composition: what a frame does with its space, where the eye is sent, what the bottom third of the image is doing. Anime composition is deliberate to an almost obsessive degree — layouts are drawn, approved, and fought over, and the placement of a figure in a frame is a decision somebody defended.

A subtitle puts a two-line white box over the bottom of that frame, for most of the runtime.

Not on a black bar. On the image. And the viewer's eye — this is the part that hurts — is down there, reading, for a large fraction of every shot. You are looking at the words when the cut happens. You are reading a line when the thing the shot was built for occurs, and you catch it in peripheral vision, and you have the plot and you did not see the picture.

So the purist watching subs is receiving: a text with a third of it deleted, weighted to preserve plot over character, printed over the bottom of the composition, while their eyes are off the image for most of the runtime. That person is not experiencing the original. They are experiencing a specific, heavy, invisible set of compromises — and feeling superior to a person experiencing a different set.

The Honest Comparison

Set the two side by side and the argument dissolves.

A dub keeps the words and loses the voice. Every line, complete, in full — no character limit, no reading speed, all the hedges and softeners intact, because speech runs at speech speed. You get the whole script. What you lose is the original performance: the actual actor, their timing, the sound of the character. And you lose lip-sync fidelity, which forces its own rewrites — lines get padded or clipped to fit mouth flaps, which is its own compression.

A sub keeps the voice and loses the words. The original performance is right there, untouched — the actual actor, their breath, the pronoun, the honorific, all audible. And you get about two-thirds of the text, plot-weighted, with your eyes off the picture.

Neither is the original. Both are large, structural losses. They are simply losses in different places, and — this is the whole point — a viewer who understands no Japanese cannot detect either one. The dub-watcher cannot hear the missing performance. The sub-watcher cannot see the missing third of the script, because it is missing. Each is confidently unaware of their own hole.

The only defensible version of the purist position is Part 7's: if you have learned enough Japanese to hear the honorifics and the pronouns, the subtitle becomes a safety net over an original you are mostly receiving directly, and then it genuinely is better. That is a real argument. It is also an argument for studying Japanese, not for feeling smug at people watching a dub.

Keikaku Means Plan

The fansub scene, which Part 11 credited with breaking the localization machine, hit this wall from the other side, and produced the funniest artifact in the history of this fandom.

Fansubs had no reading-speed discipline and no institutional restraint, and they were made by people who loved the material and wanted you to have all of it. So they gave you all of it. Translator's notes on screen, mid-scene. Coloured fonts per character. Karaoke subtitles for the opening, romanized and translated and bouncing in time. Sign-subs in the corner while dialogue-subs ran along the bottom. Occasionally three simultaneous text objects over a composed frame, and a note explaining a cultural reference that had already gone past.

Which produced the immortal line — from a Death Note fansub, and now a permanent piece of internet furniture:

"Just according to keikaku." With a note across the top of the screen: TL note: keikaku means plan.

It is perfect. It is perfect because the correct translation was "just according to plan" — four words, no note, no cognitive interruption, done. The subtitler had the English word. They knew the English word; they wrote it in the note. And they left the Japanese in the line and explained it above, because retention had stopped being a strategy and become an identity — the mark of the authentic edition, the thing that distinguished a real fansub from the licensed garbage of Part 5.

This is the leave-it instinct of Parts 6, 7, 13, and 16 — the one that has been right four times — driven straight off the cliff. Part 7 derived the rule: leave it when the system is small, closed, and high-frequency; translate it otherwise. Keikaku is an ordinary noun in an open class of tens of thousands. It fails the test completely. There is no reason on earth to leave it, and it got left, because the scene had learned that leaving things in was what respect looked like and stopped asking what it was for.

Every strategy in this series becomes a mannerism the moment it stops being a decision.

The Numbers, Which Have One Good Line Left

Subtitles reduces to Destiny 1, Heart's Desire 8, Personality 11. The Leader and Pioneer; the Achiever's money and authority on the heart; the Visionary's master 11 on the face it shows the world.

Which is nothing. It is a word. It is not a work, or a person, or a title anyone chose — it is an English common noun I typed into a function to see what would happen, and the function did arithmetic on it, as it would on refrigerator.

And that is worth one paragraph, because it is the cleanest way to see what this method is. There is no sense in which "subtitles" has a destiny. There is no sense in which the practice of subtitling has a Personality of 11. The arithmetic will produce a full reading, complete with an exalted master number and a paragraph of vibration, for any string at all. It has never once declined. It has never returned an error, never reported insufficient data, never said this is not the kind of thing that has a number.

A method that answers every question with equal confidence is not answering questions. That is the whole of the case, and it took three hundred and eighteen essays to say it the first time, and here it is again, free, sitting on a common noun.

Fansub comes out a Destiny 9 — endings and the closing of cycles — which rhymes with Part 11, where Scanlation's 9 and Shueisha's triple 9 shared a destiny, and the cycle did close, and it means nothing. Ring, door, look.

The Close

What survives the crossing? In subtitles: the voice, the performance, the sound of the thing — and about two-thirds of the words, chosen so that you always know what happened and often do not know who it happened to.

That is not a scandal. Part 6 ended by saying the border is not a wall someone built out of carelessness but a river that was always there, and the subtitle is the most conscientious bridge anyone has managed: built by people who know exactly what falls off it, who have measured the drop in characters per second, and who cut the hedges rather than the plot because a viewer who cannot follow the story has lost everything rather than a third.

The scandal is only the smugness. Every English-language viewer of anime is watching a compromised object — always, without exception, in both formats, forever. The sub-watcher's compromise is invisible to them, which makes it feel like fidelity. It is not fidelity. It is a different third of the work, missing, in a place they cannot see, over a picture they are not looking at because they are reading.

Just according to keikaku. The word was plan. They knew it was plan. They wrote it down, in yellow, across the top of a composed frame, while the scene went past underneath.