AI Skill Report Card

Translating SRT Subtitles

A-83·Jul 31, 2026·Source: Web
14 / 15

Given an SRT file, translate only the dialogue text into Arabic. Keep index numbers and timestamps untouched. Preserve line breaks per subtitle block.

Input block:

42
00:03:14,200 --> 00:03:17,800
You should have left this town a long time ago.

Output block:

42
00:03:14,200 --> 00:03:17,800
كان يجب أن تغادر هذه المدينة منذ زمن طويل.
Recommendation
Add a second example showing a trickier case (idiom/slang translation or multi-line block split) to demonstrate range beyond simple dialogue.
14 / 15

Progress:

  • Step 1: Parse the SRT into blocks (index, timestamp, text lines).
  • Step 2: Detect tone/register of the film (formal drama, slang-heavy comedy, action, etc.) to calibrate Arabic style.
  • Step 3: Translate each block's text, respecting original line breaks (2 lines in → 2 lines out when possible).
  • Step 4: Handle idioms, cultural references, and profanity with natural Arabic equivalents, not literal translations.
  • Step 5: Keep untranslatable elements as-is (names, on-screen text markers like [MUSIC], ).
  • Step 6: Reassemble full SRT with original index/timestamp structure intact.
  • Step 7: Do a consistency pass — same character names, recurring terms, and tone stay uniform across the whole file.
  • Step 8: Output the complete translated SRT file, not a summary.

For long files (100+ blocks), process in chunks of ~50 blocks, maintaining a running glossary of names/terms for consistency across chunks.

Recommendation
Include a bad-output example (e.g., overly literal translation or broken timestamp) to contrast with the good example.
  • Use Modern Standard Arabic (MSA) by default for broad readability, unless the user requests Egyptian, Gulf, or Levantine dialect.
  • Match subtitle reading speed: keep translated lines roughly similar length in reading time (Arabic script is often more compact — avoid over-lengthening).
  • Preserve the exact SRT format: index number, --> timestamp line, text, blank line separator.
  • Keep character names transliterated consistently (e.g., "Sarah" → "سارة" every time, not alternating spellings).
  • Translate meaning and tone, not word-for-word — jokes, sarcasm, and slang need natural equivalents.
  • Keep formatting tags (<i>, <b>) wrapped around the correct translated text.
  • Numbers, times, and proper nouns (brand names, places) stay accurate; adapt units only if explicitly asked.
  • Build and reuse a glossary for recurring terms (character names, nicknames, running jokes) across the whole file.
  • Don't renumber or re-time subtitle blocks — timing must exactly match the source video.
  • Don't merge or split subtitle blocks even if a sentence spans two blocks; translate per-block matching original segmentation.
  • Don't translate [SOUND EFFECTS], ♪ lyrics ♪, or speaker labels unless asked — flag them but keep bracket conventions.
  • Don't produce overly literal/word-for-word Arabic that sounds unnatural — prioritize fluent, screen-appropriate phrasing.
  • Don't switch between MSA and dialect mid-file — pick one register and stay consistent.
  • Don't summarize or skip blocks for "efficiency" — full files must be translated completely, block by block.
  • Don't forget right-to-left punctuation nuances (Arabic question mark ؟, comma ،) instead of Latin punctuation.
0
Grade A-AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
14/15
Workflow
14/15
Examples
14/20
Completeness
17/20
Format
15/15
Conciseness
13/15