How to Translate SRT Subtitles and Keep the Timing

Translating an SRT file is two problems wearing one filename. The first — translating the text — machines now handle well. The second — keeping the result readable at the original timing — is where naive approaches fail, because translated text is usually longer than the source, and the timestamps don't know that. Here's the workflow that handles both, and the fixes for the lines that overflow.

Why you can't just paste an SRT into a translator

An SRT file is structure plus text: cue numbers, timestamps, line breaks — and the dialogue woven between them. Feed the whole file to a generic translator and three things break:

• Structure gets mangled. Timestamps "translated", cue numbers merged, arrows reformatted — the file no longer parses.

• Context gets lost. Subtitle cues split sentences across entries; translating cue-by-cue without seeing neighbors produces disjointed grammar, especially in languages that reorder sentences.

• Length explodes silently. The translation lands, the timing doesn't move, and lines that were comfortable at 15 characters per second are now sprinting at 25.

The right workflow

1. Start from a clean source file. Fix transcription errors before translating — every source error becomes a weirder target error. If you're generating subtitles from scratch, review names and jargon first (our subtitling guide covers that pass).

2. Use a subtitle-aware translator. A dedicated subtitle translator parses the SRT, translates the dialogue with sentence context across cues, and writes a valid SRT back — same cue count, same timestamps, translated text.

3. Review for length and idiom. Scan for crowded lines (below), and give culture-bound phrases a human glance — idioms translate literally into confusion.

4. Validate and ship. Check the file loads in a player, spot-check timing at three or four points, and export to VTT as well if a web player needs it (SRT vs VTT explains when).

The length problem, and what to do about it

Translation changes text length — English into German or French typically grows noticeably, into Finnish it may shrink, and CJK languages pack more meaning per character but read at their own pace. The subtitle consequence: a cue timed for the original may now hold more text than a viewer can read before it vanishes. Industry style guides (the BBC's subtitle guidelines are a good public reference) treat reading speed and line length as the core constraints of the craft.

When a translated line is too long for its slot, you have four levers, in order of preference:

1. Condense the translation. Subtitling is not court reporting — trimming filler while keeping meaning is standard professional practice. "I think that perhaps we should…" becomes "We should…".

2. Re-break the lines. Two balanced lines read faster than one long one; break at phrase boundaries, not mid-idea.

3. Merge or re-split cues. If two short cues carry one sentence, merging them buys reading time without touching the video.

4. Extend timing slightly — only where a gap follows the cue; stealing time from the next line just moves the problem.

Why translated lines overflow — the character-count physics per language — is a topic of its own, and we go deeper in our piece on text expansion and timing.

Translating into many languages at once

The economics compound nicely: the timing work exists once, in the source file. Each new language is a translation pass plus a length review — which is how one subtitle master becomes ten market-ready files in an afternoon. Keep the source SRT as the canonical file; regenerate translations from it when the source changes, rather than patching each language by hand. This is the same one-transcript-many-outputs logic as our video translation workflow.

Common mistakes

• Translating before correcting. Ten source typos become ten translation mysteries across every language you ship.

• Trusting cue-by-cue translation. Sentences that span cues need context-aware translation or they read as word salad.

• Skipping the target-language review. Five minutes from a native reader catches the idiom failures that make screenshots.

• Forgetting encoding. UTF-8, always — accented characters and non-Latin scripts break on anything else.

• Patching translations by hand in each file. Fix the source, retranslate; hand-patched language files drift apart permanently.

Frequently asked questions

Will the timestamps change when I translate an SRT?

With a subtitle-aware tool, no — cues and timing stay identical, only the text changes. That's the whole point of translating the file rather than the video.

Can I translate an SRT with a generic AI chatbot?

For a short file, sometimes — with careful prompting and luck on structure. At any volume, a subtitle-aware tool is faster and doesn't corrupt timestamps.

How do I check reading speed?

Rough rule: adult viewers read comfortably around 15–17 characters per second in Latin scripts. If a line looks dense, it is — condense or re-break.

Should I translate the SRT or dub the video?

Subtitles first: near-free, editable, and the original voices survive. Dubbing earns its cost only where reading is a real barrier — the trade-offs are in our subtitles vs dubbing guide.

The bottom line

Correct the source, translate with a tool that respects the file's structure and sentence context, then spend your human minutes where machines still lose: crowded lines and idioms. The timestamps never had to move — the craft is fitting the new language into the time the old one left behind.