Bilingual Subtitles to Word

Two shapes of “bilingual subtitles”

Searchers usually have one of two files. The first is a single SRT or VTT where every cue has two lines: the source language, then the translation, sometimes with a blank line or a slash between them. Players show that as one caption. The second is two complete files, episode.en.srt and episode.es.srt, timed to the same video and easy to drift apart if only one was edited.

A third shape shows up in fansub ASS/SSA: one Dialogue event with both languages in the text, or two styles (top and bottom) that a player positions on screen. Word has no playhead and no top/bottom caption lanes. Positioning tags are dropped. You keep the words and, if you ask for them, the times. The ASS-specific flattening rules live on ASS to Word. The reading problem is the same.

This converter does not translate. It will not detect that line 2 is a translation of line 1, and it will not score which language is which. If the words are wrong, fix them in the caption file or in the document after export.

One cue, two lines

A cue like this is the common delivery from bilingual captioners:

12
00:01:04,200 --> 00:01:07,800
We can ship the cut on Friday.
Podemos entregar el corte el viernes.

The parser keeps both lines in that cue’s text. It does not turn them into a two-column table, and the Word paragraph does not become a styled “original / translation” pair on its own. Treat the downloaded block as the raw pair: both languages, one time range, ready to edit.

Turn Merge same speaker off before you download if you need each pair to stay intact. Merge is on by default, and unlabeled cues are treated as the same speaker. Any following cue that starts within about 2.5 seconds is joined with a space, and extra whitespace — including the line break between English and Spanish — collapses. Five tidy bilingual captions become one paragraph of alternating languages. That is the right behavior for a single-language interview. It is the wrong behavior for a line-by-line translation review.

With merge off, each cue is its own block. Timestamps, if shown, sit on that pair so a reviewer can find the frame. After the file is in Word you can still split a block into two paragraphs, apply a character style to the translation line, or convert a clean list into a table. Doing that from one-block-per-cue is mechanical. Doing it after an accidental merge means reconstructing the pairs by hand.

Prefixes change the parse. A cue that starts with EN: or [EN] is read as speaker “EN” plus the remaining text. Only the start of the cue is treated as the label. A second line that begins with ES: stays inside the text; it does not become a second speaker. If your file uses those prefixes on purpose, leave them — they are useful hooks for Find in Word. If they were only a captioning convenience, you can delete them in the document once, instead of inside every cue.

Two files, one document

Separate language tracks are easier to keep honest. Convert each file on its own. Use the same timestamp setting for both so the ranges still match. Then in Word or Google Docs, insert a two-column table and paste one language per column. A reviewer can comment on the Spanish cell without touching the English timing note beside it.

Do not paste both raw SRTs into one text box and hope the clocks interleave. Cue 10 in the English file is not guaranteed to be the same moment as cue 10 in the Spanish file once someone has split or joined a line. Match on the time range, not on the cue index. If the ranges already disagree, you have a sync problem, not a Word problem — see fix SRT timing before you build the table.

When the shared deliverable is a Google Doc rather than a .docx on disk, upload after the table exists, or upload each language and assemble the table in Docs. Comments and suggest mode are the point of that handoff. SRT to Google Docs covers what survives the upload.

Settings that match the review you are actually running

• Translation QC against picture: merge off, timestamps on, one block per cue. Keep the original subtitle file.
• Client reading copy in one language: export or paste only that language, merge on, timestamps off.
• Glossary pass: timestamps on even in the “clean” copy, so a term can be checked against the frame where it was spoken. Hide them only in the final export.
• Right-to-left text: the converter does not set paragraph direction. After download, select the RTL paragraphs in Word and set direction there. Do not expect ASS alignment tags to do it.

ASS karaoke and color tags are not a bilingual feature. They are styling. They flatten to plain text so the words remain. If the translation lived inside an override tag rather than in the dialogue text, it will not appear — that text was never dialogue. Check the live preview before you trust a long download.

What to hand the translator

Hand them the .docx or Doc plus the video, and say which column is source. Building that column — a real Word table, matched on the time range rather than the cue number — is bilingual subtitle columns. This page stops at one block per cue. Also hand them the caption file if they are allowed to change timing. A Word comment that says “line is early” does not move the cue. Someone still has to edit the SRT, VTT, or ASS.

Keep cue numbers out of the document. They feel like a key, and they rot as soon as anyone inserts a line. Time ranges survive splits better. If you turned timestamps off to make the page look like prose, write the filename of the caption file in the header so the next person can recover the clocks without asking you to export again.

Related workflows

• Fix SRT Timing Before Word
• SRT to Google Docs
• ASS to Word
• All subtitle-to-Word guides
• SRT to Word converter — same tool the buttons above open, with these toggles already set.