How to Translate an SRT File with Google Translate
Google Translate can't open an .srt file: its document upload takes formats like .docx, .pdf, .pptx and .xlsx. You can paste the subtitles in as text, but the timecodes and cue numbers get translated along with the dialogue and often come back damaged. The reliable way to use it is to translate only the text lines and paste them back. A subtitle translator does the same job without the copy-pasting.
Below: what actually happens when an SRT goes through Google Translate, a step-by-step method that keeps the file intact, and how it compares with ChatGPT and a dedicated subtitle tool.
Can Google Translate translate an SRT file?
Not directly. Choose Documents in Google Translate and the file picker lists document formats (.docx, .pdf, .pptx, .xlsx). An .srt is a plain-text file, but it isn't on the list, so people usually try one of two workarounds:
- Paste the file into the text box. Open the SRT in Notepad or TextEdit, copy everything, and paste it into Google Translate.
- Change the file format. Rename it to .txt or paste it into a Word document, upload that, then rename the result back to .srt.
Both get the words translated. Neither tells Google Translate that the numbers and timecodes are structure that must not change, and that is where the trouble starts.
What goes wrong when an SRT goes through Google Translate
A single SRT cue looks like this:
12 00:01:12,400 --> 00:01:15,000 I didn't think you'd actually 13 00:01:15,100 --> 00:01:17,300 come back here.
- Timecodes get changed. Google Translate can add spaces (
00: 01: 12,400), swap the comma for a period, or turn-->into a different arrow, especially when translating into a language that writes numbers or punctuation differently. One broken timecode can make a player skip that cue or refuse to load the file. - Split sentences come out as fragments. In the example, one sentence runs across cues 12 and 13. Translated line by line, “I didn't think you'd actually” and “come back here” are handled as separate scraps, so word order and grammar can come out wrong, especially between languages with different word order such as English and Japanese or German.
- Lines get merged or moved. The translator may join a two-line cue into one, or pull text across the blank line that separates cues. If text ends up in the wrong block, every later cue is out by one.
- Character limits. The text box takes about 5,000 characters per request, which is only a few minutes of dialogue. A feature film means dozens of paste-translate-copy rounds, and names or terms can be translated differently from one chunk to the next.
- Line lengths change. Translations often run longer than the original. A line that fit on screen in English can become too long to read in the time it's shown.
The safe manual method: translate only the text
If you want to use Google Translate anyway, keep the timecodes out of it completely. Only the dialogue goes through the translator, and you rebuild the file yourself.
Extract the text lines
Run the file through SRT to TXT. It removes the numbers and timecodes and puts each subtitle on its own line. Keep the original .srt; you'll need it to rebuild.
Translate in chunks under the limit
Paste up to about 5,000 characters at a time into Google Translate and copy the results into a new text file in order. Check that the line count of each chunk matches before moving on.
Paste each line back into the SRT
Open the original SRT and replace the text under each timecode with the matching translated line. A spreadsheet helps: original lines in column A, translations in column B, so you can spot a missing line before it shifts everything after it.
Save as UTF-8 and test
Save the file as UTF-8 so accented letters and non-Latin scripts display correctly, then load it in VLC and check the start, middle and end.
This keeps the timing safe, but it's tedious: a 90-minute film has well over 1,000 cues, so step 3 alone takes a long evening. It also doesn't fix the fragment problem. Step 1 joins a cue's own lines, but a sentence that spans two cues is still translated as two pieces.
Google Translate vs ChatGPT vs a subtitle translator
| Google Translate (manual) | ChatGPT | Subtitle translator | |
|---|---|---|---|
| Accepts an .srt file | No — paste or convert | Paste or attach | Yes |
| Timecodes and cue numbers | Can be mangled unless you strip them | Can be renumbered or altered | Kept unchanged |
| Sentences split across cues | Translated as fragments | Partly, if it sees enough context | Translated as whole sentences |
| Full-length movie | ~5,000 characters per paste | Long files get cut off or chunked | Whole file in one go |
| Effort for a 90-minute film | Hours of copy-paste | Many chunks plus checking | Upload, pick a language, export |
| Cost | Free | Free tier / subscription | Free preview, then paid |
Translation method comparison table
ChatGPT has its own failure modes with SRT files: dropped cues, renumbering, truncated long files. We covered them in Can ChatGPT translate an SRT file?
Translate an SRT and keep the timing
The SubtitleFlow SRT translator takes the .srt itself, so there's nothing to strip or paste back.
Upload the .srt
Drop the file in. Cue numbers and timecodes are read as structure, not text, and are never sent for translation.
Pick the target language
Sentences that run across several cues are translated as whole sentences, with the surrounding lines as context, then split back across the original cues. That is the “I didn't think you'd actually / come back here” case Google Translate gets wrong.
Check the free preview
The first 30 lines are translated free, so you can judge the quality on your own file before paying for the rest.
Review and export
Fix any line in the editor, then download an .srt with the same number of cues and the same timecodes as the original, or a bilingual file with both languages.
Have a .vtt instead? Use the VTT translator, or convert it first with VTT to SRT.
Frequently asked questions
Can Google Translate translate an SRT file?
Not as a file. The document upload in Google Translate accepts formats such as .docx, .pdf, .pptx and .xlsx, and .srt is not one of them. You can paste the subtitle text into the text box or convert the file to a supported format, but Google Translate treats it as ordinary text, so the numbering and timecodes are not protected.
Why does Google Translate mess up the timecodes?
Because to Google Translate a line like 00:01:12,400 --> 00:01:15,000 is just text. Depending on the language pair it can add spaces, swap the comma for a period, change the arrow, or move lines around. A media player then fails to read that cue, or the whole file.
Is there a character limit when pasting subtitles into Google Translate?
Yes. The text box takes about 5,000 characters at a time, which is only a few minutes of dialogue. A full movie SRT has to be split into many chunks, and each chunk is translated without knowing what came before it.
What is the safest way to use Google Translate for subtitles?
Translate only the words. Extract the text lines (one line per subtitle), translate them, then paste each translated line back into the original file in the same position, so the numbers and timecodes never pass through the translator. It works, but it's slow and easy to get off by one line.
Is Google Translate or ChatGPT better for SRT files?
ChatGPT usually gives more natural dialogue, but both treat the SRT as plain text: ChatGPT can renumber, merge or drop cues, and Google Translate can mangle timecodes and translate split sentences as fragments. For a file you actually need to play, a subtitle translator that keeps the timing is safer.
How do I translate an SRT file and keep the timing?
Upload the .srt to a subtitle translator such as SubtitleFlow's SRT translator. Only the text inside each cue is translated; cue numbers and timecodes come back unchanged, so the file stays in sync. You can preview the first 30 lines free before deciding.
Skip the copy-paste
Upload an SRT, pick a language, and get back a file with every timecode unchanged and sentences translated in context.
The first 30 lines are free to preview.