Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Arabic Subtitle Fixer: SRT/VTT Punctuation & Timing

    1

    The player guessed the line's direction from its first strong character, and an English word or a number at the start makes it guess wrong.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Arabic Subtitle Fixer: SRT/VTT Punctuation & Timing

    Arabic Subtitle Fixer: SRT/VTT Punctuation & Timing

    Example session with this skill installed

    Here's the Arabic .srt for my kids' science video on photosynthesis (10 cues, some English terms). Fix the reversed punctuation and tell me what else needs editing before I upload. Use the children's reading speed.

    • Read your context and instructions
    • Compiled the arabic subtitle fixer

    Summary: 18 line fixes were written to photosynthesis.ar.fixed.srt. 17 items still need a human edit, and the overlap on cue 5 comes first.

    1. Fixer output

    python3 scripts/ar_subtitle_fix.py photosynthesis.ar.srt --profile kids

    ──────────────────────────────────────────────────────────────────────────────
     ARABIC SUBTITLE FIXER   photosynthesis.ar.srt   cues: 10 (Arabic: 10)   profile: kids ≤17 cps
    ──────────────────────────────────────────────────────────────────────────────
     FIXED IN OUTPUT FILE
       F1 direction mark (RLM) added ...... 12 line(s)
       F2 Latin ? , ; -> ؟ ، ؛ ................... 2 line(s)
       F3 Persian letters/digits -> Arabic ........ 1 line(s)
       F4 kashida / double spaces removed ........ 1 line(s)
       F5 repeated ? ! ؟ reduced to one mark ..... 1 line(s)
       F6 space before ، ؟ ! removed .............. 1 line(s)
    ──────────────────────────────────────────────────────────────────────────────
     BEFORE -> AFTER (first changed lines; ‹RLM› ‹RLE› ‹PDF› = invisible marks)
       cue 2: Photosynthesis هي العملية التي تصنع بها النباتات غذاءها
           -> ‹RLM›Photosynthesis هي العملية التي تصنع بها النباتات غذاءها
       cue 3: هل تعرفون من أين يأتي الأكسجين الذي نتنفسه?
           -> ‹RLM›هل تعرفون من أين يأتي الأكسجين الذي نتنفسه؟
       cue 4: تمتص الأوراق ثاني أكسيد الكربون CO2 من الهواء, وتمتص الجذور الماء
           -> ‹RLM›تمتص الأوراق ثاني أكسيد الكربون CO2 من الهواء، وتمتص الجذور الماء
    ──────────────────────────────────────────────────────────────────────────────
     NEEDS A HUMAN EDIT (17)   VERDICT: BROKEN TIMING — FIX FIRST
       T1  cue 9: 3 lines (max 2)
       T2  cue 2: line is 55 characters (max 42)
       T2  cue 3: line is 43 characters (max 42)
       T2  cue 4: line is 65 characters (max 42)
       T2  cue 5: line is 58 characters (max 42)
       T2  cue 7: line is 43 characters (max 42)
       T2  cue 8: line is 57 characters (max 42)
       T2  cue 9: line is 44 characters (max 42)
       T3  cue 2: 19.6 chars/sec over the kids limit 17 (55 chars in 2.80s; needs 3.24s or 47 chars)
       T3  cue 3: 23.9 chars/sec over the kids limit 17 (43 chars in 1.80s; needs 2.53s or 30 chars)
       T3  cue 4: 17.1 chars/sec over the kids limit 17 (65 chars in 3.80s; needs 3.82s or 64 chars)
       T3  cue 5: 21.5 chars/sec over the kids limit 17 (58 chars in 2.70s; needs 3.41s or 45 chars)
       T3  cue 8: 27.1 chars/sec over the kids limit 17 (57 chars in 2.10s; needs 3.35s or 35 chars)
       T4  cue 6: on screen 0.60s (min 0.83s)
       T5  cue 9: on screen 8.50s (max 7s)
       T6  cue 5: starts before cue 4 ends (overlap 0.10s)
       T8  cue -: mixed digit styles in the file
            Arabic-Indic 1, Western 11
    ──────────────────────────────────────────────────────────────────────────────
    

    2. Suggested human edits (SUGGESTED — approve before applying)

    CueFindingSuggested edit
    343 chars, 23.9 cpsهل تعرفون مصدر الأكسجين؟ (24 chars, within the 30 allowed)
    4 + 5T6 overlap, 65 charsEnd cue 4 at 00:00:11,800, then split it: تمتص الأوراق ثاني أكسيد الكربون من الهواء، / وتمتص الجذور الماء
    60.60 s on screenExtend to 0.85 s and move the start of cue 7 to 15,350
    857 chars, 27.1 cpsتنتج الورقة نحو 5 مل أكسجين بالساعة (34 chars; the Western digit also settles T8)
    93 lines, 8.5 sSplit into two cues at «رئة الأرض»

    3. Next step

    Apply the approved edits, re-run the fixer on the result, then test the file in YouTube's player before publishing.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The problem

    You upload Arabic subtitles and the question mark appears on the wrong end: ?هل تعرف. The text is fine. The player guessed the line's direction from its first strong character, and an English word or a number at the start makes it guess wrong.

    The same files usually carry:

    • Latin , and ? instead of ، and ؟
    • Persian ی and ک
    • stretched kashida
    • legacy Windows-1256 encoding
    • cues too fast to read

    What it does

    Fixes automatically, into a new file:

    • direction marks
    • ؟ ، ؛
    • Persian letters and digits
    • kashida
    • repeated ?!
    • spaces before punctuation
    • Windows-1256 → UTF-8

    Measures every cue against the Netflix Arabic Timed Text Style Guide:

    LimitValue
    Characters per line42 (diacritics not counted)
    Lines2
    Reading speed20 cps, or 17 for children's content
    Duration5/6 s to 7 s
    Overlapsnone

    Suggests the human edits: shorter wording or new timing per cue, marked SUGGESTED for your approval.

    SRT and WebVTT. Never overwrites your original. Running it twice changes nothing.

    Why trust it

    • Standard-library Python, no network.
    • An independent QA pass tried to break it (Windows-1256 files, symlinks, tags with commas, VTT without hours, re-runs); every bug it found was fixed before release.

    What it will not tell you

    Translation accuracy, grammar, sync to speech or shot changes. The direction fix depends on the player, so test one file in yours first.

    SRT and WebVTT only (no ASS/SSA). UTF-8 and Windows-1256 input. Does not translate, judge grammar or sync to audio, and does not check shot changes. The direction fix depends on the player: test one file in your target player first.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 9 days ago

    Requires

    Python 3.6+
    Terminal / shell
    Python

    Permissions

    File Scopes

    the
    .srt/.vtt
    file
    you
    name
    (read)
    *.fixed.srt
    /
    *.fixed.vtt
    (write)

    Writes a new file next to the input, never the input itself. No network.

    Listed9 days ago

    What's inside

    Frequently Asked Questions

    Players pick a line's direction from its first strong character. A line that starts with an English word or a number is treated as left-to-right, so the final ؟ moves to the left. The fixer adds an invisible right-to-left mark at the start of each Arabic line.

    The Netflix Arabic Timed Text Style Guide: 42 characters per line, 2 lines, 20 characters per second for adults and 17 for children. Minimum 5/6 s and maximum 7 s come from Netflix's general requirements.

    No. It writes <name>.fixed.srt and refuses to overwrite the input. Running it again on its own output changes nothing.

    Yes for SRT and VTT uploads. YouTube's own Arabic limits were not checked; the Netflix numbers are used as a professional default.

    No. It repairs script and layout and measures timing against limits. Translation and sync to speech are out of scope.

    New in Social Media