Subtitle Bilingual Localization

What It Solves

Bilingual subtitle localization often fails when a model simply translates cues: it may drop blocks, break SRT numbering, lose the WEBVTT header, or shorten lines until they are no longer watchable. This skill targets .srt, .vtt, and pasted subtitle text, with the goal of producing complete, structurally stable, view-friendly bilingual subtitles.

How It Works

• Default output: English on the first line, Chinese on the second, plus a plain-text preview.
• Structure preservation: keep SRT indices, timecodes, block order, and VTT headers; do not merge, split, or retime cues unless asked.
• Script-assisted workflow: use parse_subtitles.py to create a JSON intermediate, rebuild_subtitles.py to regenerate files, validate_subtitles.py to check missing blocks, broken indices, and missing bilingual lines, split_merge_batches.py for long content, and length_check.py for readability risk.
• Long-subtitle handling: parse first, split into batches when needed, then merge, rebuild, and validate before delivery.

Boundaries

• Best for translating existing subtitle files or near-SRT/VTT text, not for editing video timing or producing media.
• If the input is badly encoded or severely damaged, surface the issue instead of guessing.
• Length checks are risk signals, not replacements for contextual translation; keep Chinese natural and concise without deleting substantive information.