Tutorials Video

Translating Subtitles Without Embarrassing Yourself

Intermediate · ~25 min

Overview

Translated subtitles are the cheapest real localization there is, and the easiest to get publicly wrong, because errors sit on screen in writing, in front of exactly the audience that can read them. The workflow that holds up is machine translation as a free first draft, a native speaker as the quality gate, and an editorial pass for the places translation physically doesn't fit the timing. This builds on the site's guides to captions and subtitles and caption quality.

What You Need

A corrected source-language subtitle file (garbage in, translated garbage out)
A machine translation step, and a subtitle editor. The site's SRT Editor runs in the browser
A native-speaker reviewer per target language, even an informal one

Steps

Start from a corrected source subtitle file

Every error in the source multiplies across languages. A misheard name wrong in one file becomes wrong in ten. Do the source correction pass (names, numbers, jargon, speaker labels) before translating anything. It's the highest-leverage minutes in the whole workflow.

Use machine translation as the draft, never the deliverable

Modern machine translation produces a genuinely useful first pass for major language pairs, fast, free, mostly right on conversational content. Publishing it unreviewed is where channels embarrass themselves: the errors cluster on idioms, tone, and names, exactly the lines viewers screenshot.

Put a native reviewer at the QC gate

A native speaker skimming the translated file catches in minutes what no automated check can: unnatural phrasing, wrong register (formal/informal address matters enormously in many languages), and idiom failures. No native reviewer available for a language? That's a signal to rely on the platform's auto-translation instead of publishing your own unverified file.

Handle timing and length shifts across languages

Translations run longer or shorter than the source, sometimes dramatically. Where a translated line can't be read in its allotted time, condense the meaning rather than cramming the text: subtitle translation is editing, not transcription. Keep the source timing. Adapt the words to fit it.

Deal with idioms, wordplay, and on-screen text

Idioms get replaced with the target language's equivalent meaning, not translated literally. Wordplay usually dies in translation. A brief functional line beats a broken joke. On-screen text visible in the picture needs a subtitle rendering it ("[Sign: Closed on Mondays]") since the pixels don't translate.

Manage per-language files with a naming convention

One file per language, named with standard language codes (video-title.es.srt, video-title.de.srt), stored with the project. Ad-hoc naming is how the German file gets uploaded to the French slot. An error viewers notice immediately and you don't.

Pro Tips

Build a per-language glossary of your recurring terms and names as reviewers correct them. The second video in each language gets cheaper than the first.
Translate your title and description along with the subtitles. A translated video that surfaces with an English title halves its own click-through in that market.
Start with one or two languages your analytics actually support rather than ten speculative ones, review capacity is the bottleneck, so spend it where the audience is.

Knowledge Base

Subtitle Translation Is Editing Under Constraint

What separates subtitle translation from document translation is the timing box: every line must be readable in the seconds the source allotted, in a language that may need more words to say the same thing. That makes it an editorial craft, deciding what meaning survives condensation, rather than a lookup task, and it's why the "cram the full translation in" approach produces subtitles nobody can finish reading.

The Native Reviewer Is the Difference Between Reach and Ridicule

Machine translation errors aren't random. They're systematic in ways native speakers spot instantly and non-speakers can't see at all: wrong formality register, literal idioms, tone-deaf word choices. A published subtitle file is a public claim of caring about that audience. An unreviewed one that gets the basics wrong communicates the opposite of what the localization was for.

Where This Fits

This guide covers one specific part of captions and access. The wider picture, caption formats, reading speed, speaker identification, what automatic captioning still gets wrong, and a practical QA pass, is in Beyond Auto-Captions: Caption Quality, Styling, and Readability, which frames the discipline as a whole and links out to the detailed guides underneath it, including this one. If you are starting from scratch rather than solving a specific problem, read that first and come back here.

FAQ

Q: Is machine translation good enough for subtitles now?
A: As a first draft, genuinely yes for major language pairs, modern machine translation gets conversational content largely right. As a published deliverable without review, no: it reliably fumbles idioms, tone, names, and anything culture-bound, and subtitle errors are public and screenshot-able. The draft is free. The native review pass is what makes it publishable.

Q: Why do translated subtitles overrun the timing?
A: Languages differ in length for the same meaning, translations can run considerably longer or shorter than the source text. A subtitle timed for a short English line may need more characters in another language than the duration allows at a readable pace, which means the fix is editorial (condense the line) rather than mechanical (squeeze more text into the same seconds).