Part 28: Translating a Book You Have Not Finished: The Simulcast and the Spec the Pirates Wrote

Here is a working condition that would be considered insane in any other translation job.

You are given episode 3 of a twelve-episode series. You have somewhere between a few hours and a couple of days. You have not seen episode 12. Nobody has — it does not exist yet; it is being animated, possibly being written. You do not know how this ends, what any of it is for, or which of the lines in front of you is the one that matters.

Translate it. It goes live within hours of Japanese broadcast, worldwide, to an audience that will screenshot your mistakes.

That is the simulcast, and it is now how essentially all anime crosses into English. Everything else in this series — Schodt with his forty years, Rachel Thorn's decades of advocacy, the letterer at two in the morning — is a description of a world that no longer exists for the vast majority of what gets translated. This is the world.

The book you have not finished

A novel's translator reads the novel first. This is so obvious it sounds stupid to say, and it is the single largest structural advantage in the craft, and the simulcast translator does not have it.

Reading the whole thing first is not a courtesy. It is how you know what to do. It tells you which repeated phrase is a motif and which is a coincidence. It tells you that the throwaway line in chapter 2 is the thing the ending turns on, so you had better plant it. It tells you the character who sounds stiff is being stiff on purpose. It tells you what the book is about, which is the only thing that lets you decide any of the questions this series has spent twenty-seven parts cataloguing.

Now take it away. Every one of those decisions still has to be made — on a deadline, from inside episode 3, blind.

Consider what that does to the specific problems already on the table. Part 14's pronoun switch: a character says boku in episode 3. In episode 9 they will switch to ore, and it will be the turn of the whole series. You cannot foreshadow what you do not know. You cannot set up the English contrast in advance, because you don't know there's going to be a contrast, and by the time you find out, episode 3 shipped six weeks ago and is sitting on the platform forever. Part 20's name pun: is this character's name load-bearing? Will the show cash it in? Guess. Part 5's honorific policy: you are setting a precedent in episode 3 for a relationship you have not seen develop.

A novel translator makes these decisions with the ending in hand. The simulcast translator makes them as bets, and then has to keep the bets consistent for three months, and cannot revise the early ones.

Most "bad" simulcast subtitles are not carelessness and not incompetence. They are a person guessing correctly about four things and wrong about the fifth, in public, permanently, at speed.

Worse than blind

It gets more specific, and the details matter because they're not widely known outside the trade.

Simulcast translators frequently work from scripts rather than finished episodes — sometimes because the finished episode does not exist yet in a form anyone can send. A script is text. It does not contain the framing, the pause, the way the performance lands, the face on the cut. Everything Part 26 identified as the actual content of a performance is absent from the document being translated.

And Japanese is a language in which enormous amounts of information are carried by omission. Subjects are dropped. Who is speaking to whom, and about whom, is routinely recoverable only from context — from register, from honorific, from who is on screen. Hand a translator a bare script with no picture and you have removed the context that disambiguates the sentence. There are lines that are genuinely, formally ambiguous on paper and completely clear on screen, and the translator is working from paper.

Then the schedule slips — production is late, the episode arrives at the platform hours before broadcast, and the window compresses from days to one evening.

Speed is the product

So why does it work this way? Because speed is not a side effect of the business. Speed is the business.

Part 11 told the first half of this story: scanlation and fansub, the pirates who built a global distribution network for a product nobody was selling them. The argument there was that the pirates wrote the spec — that they established what the audience expected before any legal entity was in a position to offer it.

Look at what the spec said. Fast — same week, ideally same day. Subtitled, not dubbed. Streamed, not shipped. Free at the point of use, which became cheap. Honorifics kept. Translation notes at the top of the screen when something didn't cross.

Now look at the legal industry that replaced it. Fast: adopted, completely, to the point of structural absurdity. Subtitled: adopted. Streamed: adopted. Cheap: adopted. Honorifics: partially, inconsistently, per-house.

The translation notes: gone.

That's the trade, and it's worth staring at. The fansub's defining move — the thing it did that no professional operation had ever done — was to admit in front of the audience that something had not crossed. A line of text at the top of the frame saying: this is a pun, here is why it's funny, we couldn't do it. It was often clumsy. It was sometimes insufferable. It was frequently wrong. And it was the only mechanism in the entire history of this medium by which an audience was routinely told, in the moment, that they were reading a translation.

The industry kept every part of the pirates' spec except the honesty. The speed made money. The honesty made the seam visible, and Part 21 already established the rule: the seam is not allowed to be visible, because the craft's success condition is invisibility.

So we have arrived at a delivery system that is fast, legal, cheap, comprehensive, better than anything the pirates managed — and that has quietly deleted the one feature that ever told the viewer the truth about what they were watching.

Nobody is allowed to talk to anybody

One more structural absurdity, and it's the one that produces the errors fans find most damning.

A big property is not one translation. It's several, running in parallel, in different media, at different points in the story, done by different people at different companies under different contracts.

The manga is being translated by one house, and it is years ahead — the anime is adapting volume 8 while the manga's English edition is on volume 24. The anime is being simulcast by a platform, weekly, blind. There may be a light novel with a third translator. There is merchandise with a fourth. A game with a fifth. Each of them is inventing English for the same invented vocabulary — the technique names, the ranks, the fictional institutions, all the coined nouns that Part 20 showed are load-bearing.

They frequently do not match. The same attack has two English names. A rank is a "captain" here and a "commander" there. A character's family name is romanised two ways depending on which product you bought. Fans catalogue these in enormous documents and treat them as evidence of contempt.

They are evidence of a schedule. The manga's translator, three years downstream, cannot phone the simulcast translator to agree a term, because the simulcast translator is working tonight on an episode that airs in the morning and the manga translator settled that word in 2021 in a different company under a different licence. And even where a glossary exists — and good houses maintain them — the anime translator is often blind to what the manga's English did, because they are not translating the manga, they are translating a script that arrived four hours ago.

The result is that English-language fans experience a franchise as a set of slightly incompatible dialects of itself, and reasonably conclude somebody is being careless. Nobody's careless. There is simply no mechanism, anywhere in the pipeline, by which the people making these decisions are permitted to be in the same room.

The chain this completes

And now the part that made me want to write this essay, because three separate parts of this series turn out to be one story.

Part 11: the pirates establish that the audience expects it now, this week, for nothing.

Part 28 — here: the legal industry, to compete, adopts that expectation as a hard constraint. Speed becomes non-negotiable. The working conditions of every anime translator alive are now downstream of what a group of unpaid enthusiasts proved was possible in 2003.

Part 24: a machine arrives that is very fast and very cheap, into an industry that has spent twenty years making speed and cheapness the only axes that matter.

The machine is not an invasion. It is the logical terminus of a demand curve that started with fansubs. Everyone in this chain got what they asked for. The audience asked for it now, and the industry delivered it now, and having established that now is the product, it turns out there is something that can do now better than a person can, forever, without sleeping. Part 24 asked whether the machine widens the sieve. It does. It also arrived at exactly the moment the industry had finished optimising itself into a shape the machine fits perfectly.

Nobody did anything wrong. That's the uncomfortable bit. Fansubbers loved the medium. The audience wanted the show. Crunchyroll built something genuinely good and genuinely legal that pays Japanese rights holders. Every step was reasonable and the destination is a job that a person is structurally worse at than a machine, because the job has been redefined until the only remaining virtue is speed.

The numbers

Simulcast reads Destiny 9, Heart 4, Personality 5.

Fansub reads Destiny 9, Heart 4, Personality 5.

Identical. All three.

I ran "Fansub" back in Part 11 and got Destiny 9, Heart 4, Personality 5, and built an essay called The Pirates Wrote the Spec around a shared Destiny 9. I ran "Simulcast" this week without thinking about it and the engine handed back the same three numbers, in the same order, for the legal industry that was built specifically to destroy the thing Part 11 was about.

The pirates and the platform that killed them are, according to this machine, the same entity.

That is the best coincidence this series has produced. It is the ninth clean match. It says exactly the thing I spent two thousand words arguing, which is that simulcast is fansub — licensed, funded, deadlined, and with the honesty removed — and the arithmetic said it in six characters and eight characters without knowing what either word refers to.

And it is worth nothing, and I am now two parts away from proving that with the same engine, over eight thousand names, and I would ask you to hold onto how good this one felt. Remember that you almost believed it. That feeling is the entire subject of Part 30.

Speed reads Destiny 22. Master Builder — the second-highest number the system has, handed to the one virtue this industry has left.

That one I'd almost let stand.