Translate SRT files without losing sync

Free SRT translators and their real limits

The 2026 SERP is full of "translate SRT free" tools that differ in one thing that matters: what stops you at the paywall. The honest comparison is not the language count — most claim 60 to 200+ — it is the free quota, the file-size ceiling, and whether the download is gated.

The one universal property: tools that parse the format natively keep every sequence number, timecode, and blank line byte-identical and translate only the text lines. Tools where you paste the raw file into a generic text translator are the ones that break — timestamps get mangled, cue blocks merge, and the file stops loading. If a tool asks you to paste the whole SRT as text, it is the wrong tool.

The reading-speed trap (why translated subtitles feel too fast)

Translation keeps the timestamps and changes the text — but the new text is rarely the same length. German runs roughly 25% longer than English; so do Portuguese and Dutch. A three-second English cue that holds ~45 characters comfortably (15–17 characters per second) can demand 20+ CPS in German, which is past the comfortable ceiling and off-screen before the viewer finishes reading.

Broadcast caps to know: up to 42 characters per line, maximum two lines per cue, reading speed around 17 CPS for general audiences (Netflix allows 20 for adults, ~15 for accessibility). The fix is condensing, not re-timing: instruct the translator (human or AI) that meaning beats word-for-word fidelity, cut filler phrases, and split an overloaded cue into two when the timing window allows.

After any machine pass, run a re-fit check in a subtitle editor: Subtitle Edit (free, Windows) flags overlong lines and reading-speed violations per cue (Tools > Reading speed), and Aegisub (free, all platforms) shows the waveform so you can spot cues that overflow their window. DeepL preserves timestamps but does not condense — the re-fit pass is where German translations get fixed.

Encoding is the other silent breaker: save the translated file as UTF-8 without BOM. Target scripts add their own quirks — Spanish adds inverted punctuation, French swaps quotation marks for guillemets, Japanese needs full-width character counting, and RTL scripts (Arabic, Hebrew) render differently across players. None of these are translation errors; they are formatting issues that a correct translation introduces.

The LLM route (best quality, manual batching)

For idioms, technical terms, and tone, an LLM with a subtitle-aware prompt outperforms statistical machine translation — the trade-off is batch size: most models process 50–100 cues per request, so a feature-length file means sequential chunks. The prompt that works:

Cost is negligible at this scale: a feature-film file is a few thousand tokens per chunk. Format preservation lands around 90% in testing — below native SRT editors, above generic translators — so keep the same post-pass: check sequence numbers, blank-line separators, and the 42-character cap before shipping.

Translated subtitles on YouTube

Two distinct jobs. As a viewer, YouTube's auto-translate overlays machine-translated captions on any video that already has a caption track (CC → Settings → Subtitles → Auto-translate) — 100+ languages, no download possible, accuracy roughly 70–90% because errors in the source captions get amplified by translation.

As a creator, the durable path is uploading the translated file as a proper track: YouTube Studio > Content > your video > Subtitles > ADD LANGUAGE > Upload file > With timing. A translated SRT you control beats the viewer-side overlay: it ranks in local-language search, works offline, and carries your corrections. SRT, VTT, SBV, and TTML are accepted (10 MB, 10,000 cues, plain UTF-8).

If you do not have an SRT yet, create one locally first: VocalFuse transcribes audio and video on your Windows machine with a Whisper-class engine — nothing uploads — and exports SRT, VTT, or TXT from the same pass. The free subtitle generator guide covers the generator field; this page assumes the subtitle file already exists and needs to cross a language boundary.

Bilingual subtitles (both languages at once)

A bilingual SRT shows the original line and the translation together — the standard format for language learning and for international releases where part of the audience knows the source language. Tools like Pancake's subtitle translator and VexaScribe offer a bilingual export toggle; the manual equivalent is interleaving the translated lines under the originals in a text editor, one cue block at a time, keeping the timestamps untouched.

Watch reading speed twice over on bilingual files: two lines now carry two languages, so condense the translation harder or accept a faster read. If you are translating your own captions for a course or channel, translate from the corrected transcript, not the auto-caption draft — source errors propagate into every language you ship.

Frequently asked questions