Translate Documentary Testimony for a Pitch Without Changing What Was Claimed
Translate Documentary Testimony for a Pitch Without Changing What Was Claimed
Between a recording and the sentence that lands in a pitch deck, a claim passes through four hands: the recording, the transcript, the translation, and whoever writes the pitch. Every hand is trying to make the material more usable, and every hand can quietly upgrade it.
The upgrade is rarely dramatic. A possibility becomes a fact. Something that was heard becomes something seen. A statement stops belonging to the person who made it and starts belonging to the film. By the time anyone notices, the line reads beautifully and says something the speaker never said.
So the working rule for translated testimony is this: keep recording, transcript, translation and your own paraphrase as four distinguishable things. Ask a competent language collaborator which words carry uncertainty, reported speech, emphasis, idiom or culturally specific meaning. Then decide, deliberately, whether this pitch needs a short checked quotation, an attributed paraphrase, or a plainly stated open question. Fluent English is what a good translation produces. It is not evidence that one happened.
Four records, not one
The recording is what was said, in the room, in the source language. The transcript is a written record of that, also in the source language — already an interpretation, because somebody decided where the sentences end and which clause a "he said" attaches to. The working translation is an English rendering, made by a particular person on a particular date, with a version number. The pitch excerpt is the sentence you are proposing to print.
These are four documents, and they can disagree with each other in ways that matter. If the translator revises version three into version four, the pitch sentence has to be read again. If the transcript misheard a name, everything downstream is wrong in a way that no amount of good English will fix. So keep a locator that walks backwards through the chain: recording file, timecode, transcript line, translation version, collaborator. That locator does not have to appear in the pitch. A sensitive recording does not need to be handed to a funder in order for your team's source chain to be responsible. It does need to exist somewhere that isn't anyone's memory.
One habit prevents a lot of confusion: when you show a colleague "the quote," say which layer you mean. Most disagreements about a translated line happen because one person is looking at the source-language transcript and the other is looking at an English sentence that was never spoken.
What a sentence carries besides its subject
A translated line can keep every noun and still lose the claim. The things that go missing while the topic stays put:
Source of knowledge. Saw it, heard it, inferred it, assumes it. These are different claims wearing the same words.
Certainty. Might, probably, definitely, or whatever the source uses instead of a modal. Languages differ in how they mark information source and degree of commitment — some with particles, some with verb forms, some not in the grammar at all but in the choice of words. How your particular source does it is a question for someone who knows that language variety.
Attribution. Whose claim is this, and is the speaker passing along someone else's claim inside their own? Nested reports are where pitch sentences go wrong most often.
Time. "Already," "by then," "the next morning." A paraphrase that drops already has changed the order of events even though nothing sounds different.
Extent. The door or a door. The water or some of the water. Everyone, or everyone in the house.
Emphasis and idiom. Repetition, word order, a proverb, a term of address, a kinship word more precise than "cousin" — or broader than it, which the word "cousin" can hide.
Not every loss is the same size. Losing an emphatic "myself" trims emphasis. Losing might changes what the speaker is willing to assert. Learn to tell those apart and you can polish freely in one direction and stop dead in the other.
An English fixture, and what gets lost
Everything in this section is invented, written in English, and has been through no translator at all. It is a logic fixture — a way to practise the attribution and certainty checks before you have a real passage. It is not translated testimony, and no language review has touched it.
"My cousin said the door might already have been open. I did not see it myself."
What it actually contains: a speaker; an attribution, since the possibility belongs to the cousin; a claim nested inside a claim; the modal might, not was; the temporal marker already; and a statement about the speaker's own basis for knowing, which is that they did not see it. It does not say the cousin saw the door open. The fixture is silent on that, and you should stay silent too — the temptation to assume a witness behind every report is exactly what turns a possibility into an observation.
Now the reductions. Each one reads better than the fixture and costs something:
- "The door was already open." Possibility becomes fact, the cousin vanishes, and a secondhand account now reads as somebody's direct observation. Three losses in five words.
- "My cousin opened the door." Keeps the cousin, invents an action.
- "I saw the door was already open." Inverts the one thing the speaker was explicit about.
- "The door might have been open." Keeps the possibility, loses whose it was. The doubt now appears to belong to the production rather than to a person.
- "My cousin said the door was already open." Keeps the attribution, drops the possibility entirely. This is the most common loss in a pitch, because the modal is the first thing to go when a writer wants a declarative sentence.
- "My cousin said the door might have been open, and I never saw it." Careful: never adds extent the source didn't claim.
Here is a paraphrase that holds:
According to the interviewee, their cousin had said the door might already have been open. The interviewee had not seen it directly.
Check it honestly. The attribution is intact. The possibility is intact. The non-observation is intact. Already survives. Nothing new has been added. It is longer, more clotted and less quotable than "The door was already open" — and that is the price, which in a pitch is usually worth paying. It still needs one more label, because the English on the page is a sentence no one actually spoke.
Fluent is not the same as faithful
Some awkwardness is load-bearing. A translation that repeats a name where English would reach for a pronoun may be marking a shift of subject or a point of emphasis. Nesting that sounds clumsy in English — he said that he said — may be the shape of reported speech in the source. "My father's younger brother" may sound over-explained and be more precise than the word you want to replace it with. A sentence that leaves a term in the source language may be leaving it there because the nearest English word is a different concept.
So the question you put to a collaborator is not "make this read well." It is: which of these oddities, if I smooth them, would change the claim? Once you have that answer you can smooth everything else without worry.
And do not polish before the answer arrives. A polished translation is sticky. It is the version everyone remembers, including the people who later have to correct it.
Quotation, paraphrase, or a stated open question
Three options, and a reason to pick each.
A short checked quotation when the wording itself is the point — a phrase the film will echo, a line with a rhythm worth keeping — and the rendering has been checked by someone who knows the variety. Mark it as what it is: translated from the language, by whom. Quotation marks are a promise that a person said those words. When the English is a translator's sentence, the promise carries a qualifier, and the qualifier belongs in the pitch rather than in your head.
An attributed paraphrase when the pitch needs the substance rather than the wording: who knew what, how they knew it, how sure they were. Keep the attribution and the modal. Let the paraphrase carry what a quotation cannot in the space you have.
A stated open question when a phrase is genuinely ambiguous and checking has not resolved it. A sentence like "This could mean the family still holds the land, or that they hold the documents for it; our translator is confirming before we use either" belongs in a pitch. Vague hedging does not, because it hides the specific problem from the one reader who would want to know about it.
Two rules that do not bend. Paraphrase never goes inside quotation marks. And the sentence introducing a quote must not assert more than the quote does — "She describes watching the water reach the road" over a quote that says the water might have come up is not a framing choice, it is a different claim.
Rights and consent are separate territory. A checked translation tells you what words mean. It tells you nothing about whether you may use them.
What a language collaborator can and cannot give you
Can: identify how uncertainty is marked and what it commits the speaker to; separate the speaker's own knowledge from what they were told; flag idiom, emphasis, honorifics and culturally specific meaning; produce two or three alternative renderings and say what each gains and loses; and name plainly what cannot be carried into English.
Cannot: confirm that the door was open. A translation check settles what a sentence means, not what happened. Keep those findings apart in your notes — otherwise a line reading "translation verified" gets read six weeks later as "event confirmed," and the pitch starts leaning on something nobody ever established.
Ask for the right variety, too. A speaker of a different regional variety may misread an idiom, and the misreading will look exactly like fluency.
A machine rendering can tell you roughly what a passage is about. It will not tell you which reading the grammar allows, or how the speaker marked where their knowledge came from, and there is no one to ask why it chose the words it did. Use it to orient yourself. Do not quote it.
What to ask for, concretely: three versions of the same passage — a literal one, an idiomatic one, and the misleading one that reads best. The third is the useful one. It shows you the shape of the trap you were about to walk into.
The check you run on your own revision
Six questions on every draft sentence:
- Who is claiming this, in the sentence as written?
- How do they know?
- How certain are they?
- When, relative to what?
- How much of it?
- Does my sentence assert more than the source did?
Then the trap that catches careful people: returning nouns prove nothing. Door and open appear in every version above. The claim lives in the verbs, the modals, the pronouns, the tense and the attribution, and that is where you look.
If you have to choose between a sentence that reads well and a sentence that says what the source says, the pitch can survive the clunkier one. What it cannot survive is a funder discovering at the edit that the film's central claim was never made.
The excerpt, its locator, and its open question
End with the excerpt as the recipient will read it, labelled; the locator kept where it belongs; and any remaining language question written out exactly rather than absorbed into confident prose. Record the collaborator's decision as well — which rendering they chose, on what date, and what they said it cost. When someone revises the line in three weeks, that note is how you'll know whether the revision was harmless.
Filled with the invented fixture, and with the placeholders you would replace with real information, it looks like this:
Translated from [language and variety] by [name], checked [date].
According to [speaker], their cousin had said the door might already have been open. [Speaker] did not see it directly.
Source locator (internal, not for the public pitch): interview recording, timecode, transcript line, translation version.
Open question: whether the word rendered "said" also carries a suggestion of doubt. Under review — until that is settled, we paraphrase rather than quote.
Frequently asked questions
Why keep the recording, transcript, translation, and pitch excerpt separate?
They are four documents that can disagree. The transcript already interprets where sentences end and what a 'he said' attaches to; a working translation has a version and a translator; the pitch sentence is what you propose to print. Keeping them distinct lets revisions or mishearings be traced instead of silently changing the claim.
What is a language collaborator able to check, and what can they not confirm?
They can identify how uncertainty, attribution, emphasis, idiom, honorifics, and culturally specific meaning are marked; separate the speaker's knowledge from what they were told; offer alternative renderings and say what each gains or loses; and name what cannot be carried into English. They cannot confirm that the event happened. A translation check settles what a sentence means, not what occurred.
When should a pitch use a checked quotation, an attributed paraphrase, or a stated open question?
Use a short checked quotation when the wording itself is the point and the rendering has been checked by someone who knows the language variety, marked as translated and by whom. Use an attributed paraphrase when the pitch needs the substance: who knew what, how they knew it, and how certain they were. Use a stated open question when a phrase is genuinely ambiguous and checking has not resolved it, naming the specific problem rather than hedging vaguely.
What do common reductions do to a claim like 'my cousin said the door might already have been open'?
They can change possibility into fact, remove the cousin, make a secondhand account read as direct observation, invent an action, invert the speaker's basis for knowing, lose whose doubt it was, or drop the possibility entirely. Adding words like 'never' can also add extent the source did not claim. Returning nouns do not prove fidelity; the claim lives in verbs, modals, pronouns, tense, and attribution.
How should a pitch handle a line whose meaning is still being checked?
It should write out the remaining language question exactly rather than absorb it into confident prose, and it can paraphrase rather than quote until the question is settled. The body also advises labelling the excerpt, keeping the source locator internal, and recording the collaborator's decision, date, and what they said the rendering cost.