RTL Subtitle Fixer

Fixes right-to-left subtitle files (Arabic, Hebrew, Persian, Urdu) whose punctuation, numbers or English words show up on the wrong side in some players and editors. Give it SRT or WebVTT files by URL or pasted text. You get a corrected SRT, a corrected VTT and a short report.

What it fixes

• Line direction. Each right-to-left line is wrapped in Unicode embedding marks (RLE at the start, PDF at the end), so a player that assumes left-to-right text still lays the line out right-to-left. Full stops, question marks and commas then land at the left end of the line, where a reader expects them. Numbers, times such as 10:30, and English words inside the line keep their own reading order.
• Multi-line cues. Every line of a cue is wrapped on its own.
• Reversed punctuation (optional). With repairReversed on, a closing mark at the start of a line (.مرحبا) moves to the end, and an opening bracket at the end of a line moves to the start. This is for files an earlier tool damaged. Leave it off otherwise.
• Encodings. Reads UTF-8 with or without a BOM, UTF-16 with a BOM, Windows-1256 (Arabic) and Windows-1255 (Hebrew). Output is always UTF-8.
• Timing lines. Copied unchanged. When the output format differs from the input (VTT in, SRT out, or the reverse), only the millisecond separator changes (, or .), a missing hour is added for SRT, and VTT cue settings such as align:end are dropped from the SRT.
• Other lines. Lines with no right-to-left letters, or with more Latin than right-to-left letters, are left as they are. <i> tags and {\an8} positioning stay outside the wrap.
• Running it twice changes nothing. You can feed the output back in.

What it does not do

• It does not burn subtitles into video, and it does not shape or restyle letters.
• It does not translate, retime, or fix spelling.
• It cannot rescue text that was fully reversed letter by letter.
• Embedding marks are invisible characters. A player or editor that strips them, or that draws its own direction logic on top, may still show the line wrongly.
• The repairReversed rule is a heuristic. It skips lines that already end with punctuation and leading ellipses (...), so it can miss some damage and should be checked on a sample.
• Persian and Urdu use the same rules as Arabic. Only Arabic and Hebrew were used in testing.
• Windows-1255 and Windows-1256 are told apart by counting right-to-left letters, so a very short file may be guessed wrongly. Check the inputEncoding field in the report.

How it was tested

Results depend on the player. This was not tested on Plex, Premiere Pro, Emby, Infuse, VLC or any other real player or editor. What was done:

• Unit tests with Arabic and Hebrew lines (punctuation, numbers, mixed English, tags, VTT and SRT, all supported encodings, running twice).
• Each fixed line was run through the Unicode bidirectional algorithm with a left-to-right base direction, which imitates a player that forces left-to-right. The result matches how the plain line looks in a correct right-to-left player, and the unfixed line does not.
• Before and after lines were drawn to an image and checked by eye, using the same simulation.

If a player ignores direction marks, try the repair option or tell us in an issue which player and what you see.

Input

Field	Meaning
subtitleUrls	Public URLs of .srt or .vtt files
subtitleText	SRT or WebVTT text pasted in, handled as one more file
repairReversed	Move reversed punctuation back (default off)

Output

One dataset row per file: source, srt, vtt, report, error. The report has inputFormat, inputEncoding, cues, linesFixed, linesLeftAlone and repairReversed. A file that cannot be read gets a row with error and does not stop the others. Files up to 20 MB are accepted.

Pricing

Pay per event: one charge for each subtitle file converted, nothing for files that fail.