Hebrew & Arabic Subtitle Fixer (RTL punctuation, encoding, timing)

Fix the classic problems of Hebrew and Arabic subtitles in one run:

• Punctuation on the wrong side – .שלום → שלום., ?ماذا → ماذا؟-style issues, dialogue dashes at the end of the line
• Gibberish characters (שלו×, ÷ìåí) – auto-detects Windows-1255 (Hebrew), Windows-1256 (Arabic), UTF-16 and UTF-8 BOM and outputs clean UTF-8
• RTL display – adds invisible RLM marks (or RLE/PDF embedding for old TVs) so players such as VLC, Plex, Kodi, YouTube, browsers and smart TVs show the line right-to-left
• Timing – shift by milliseconds, convert 23.976 ↔ 25 fps and other frame rates
• Convert SRT ↔ WebVTT, remove <i>/{\an8} tags and hearing-impaired notes ([music]), re-wrap long lines
• Quality report – overlapping cues, cues that are too short or too fast to read (characters per second)

Pure file processing: no scraping, no external APIs. Works on files you upload or link to.

Input example

{
  "subtitleUrls": [{ "url": "https://my-bucket.s3.amazonaws.com/episode-01.he.srt" }],
  "outputFormat": "srt",
  "punctuationSwap": "auto",
  "shiftMs": -1500,
  "removeHearingImpaired": true
}

You can also paste the subtitles into subtitleText.

Before / after

Input (Windows-1255, typed in a left-to-right editor):

1
00:00:01,000 --> 00:00:03,000
.שלום, מה שלומך

2
00:00:03,500 --> 00:00:05,000
?מה קורה
אני בסדר -

Output (UTF-8, logical punctuation + RLM marks, displays correctly everywhere):

1
00:00:01,000 --> 00:00:03,000
שלום, מה שלומך.

2
00:00:03,500 --> 00:00:05,000
מה קורה?
- אני בסדר

Dataset item:

{
  "fileName": "broken_he_cp1255.srt",
  "status": "ok",
  "detectedEncoding": "cp1255",
  "outputKey": "fixed-0001-broken_he_cp1255.srt",
  "report": { "cues": 4, "rtlLines": 5, "punctuationSwapped": true, "overlaps": 0, "tooFast": 0, "tooShort": 0, "emptyRemoved": 1, "warnings": [] }
}

The fixed file is saved in the run's key-value store under outputKey (download it from the Storage tab or via API).

How "auto" punctuation detection works

A file is treated as "typed backwards" only if at least 3 right-to-left lines start with sentence punctuation and such lines outnumber correctly punctuated lines 2:1. Correct files are never modified, so it is safe to run on mixed batches. Force the behaviour with "punctuationSwap": "always" or "never".

Pricing

$0.01 per subtitle file fixed. Failed downloads and files that are not subtitles are free.

Integrations

Call it from the Apify API, Make, Zapier or n8n after your transcription or translation step, or chain it with Hebrew & Arabic Audio Transcription + RTL Subtitles.

Privacy

Files are processed only in your run. Logs contain no file names, URLs or subtitle text. Only process subtitles you have the rights to use.