← nerv-analyticLoom

polyglot subtitle intelligence

Loom

Drop in any video, pick a native and a target track, and get back a stitched subtitle file with phonetic readings layered above the foreign script — all in your browser, no upload required.

writing systems

Romanization + per-token annotation across CJK, Brahmic, Cyrillic, and RTL scripts. Long-vowel modes for Japanese (macrons / doubled / unmarked). Phonetic-system overrides for Thai (Paiboon+ / RTGS / IPA), Arabic, Persian, Urdu.

  • Japanese · 日本語
  • Mandarin · 普通话
  • Cantonese · 廣東話
  • Korean · 한국어
  • Thai · ภาษาไทย
  • Cyrillic · Кирилица
  • Hindi · हिन्दी
  • Bengali · বাংলা
  • Tamil · தமிழ்
  • Telugu · తెలుగు
  • Hebrew · עברית
  • Arabic · العربية
  • Persian · فارسی
  • Urdu · اُردُو

In-browser

ffmpeg.wasm probes and demuxes the container. html2canvas rasterizes layered subtitles for PGS encoding. Nothing larger than ~100 KB ever leaves the page.

Two outputs

A 4-layer .ass file (native, target, romanized, annotation) and a PGS .sup bitmap track. Both at 1080p, ready to mux back into the source MKV.

Server-light

The hosted API only handles romanization + furigana — the parts that need MeCab, jieba, pythainlp, aksharamukha. Everything else is yours.