translit

A small Python library for name-aware transliteration. Eight language pairs. github.com/T-Fizz/translit

Try it

→

given-first order (Japanese / Korean)

…

loading Python in browser…

Runs the actual Python library locally in your browser via Pyodide + WebAssembly. First load is ~30 MB (cached after); subsequent calls are instant. Thai is the only pair unavailable here — its library has native code that Pyodide can't run. Use the package directly for that.

Examples

from translit_core import transliterate

transliterate("カナちゃん", "en")                              # → "Kana-chan"
transliterate("山田太郎", "en", source_lang="ja")              # → "Yamada Taro"
transliterate("山田太郎", "en", source_lang="ja",
              name_order="given-first")                      # → "Taro Yamada"
transliterate("김정은", "en", source_lang="ko")                # → "Kim Jeong-eun"
transliterate("Иван Сергеевич Тургенев", "en")                # → "Ivan Sergeevich Turgenev"
transliterate("नरेंद्र मोदी", "en", source_lang="hi")           # → "Narendra Modi"
transliterate("عبد الرحمن", "en", source_lang="ar")            # → "Abdul-Rahman"
transliterate("ทักษิณ", "en", source_lang="th")               # → "Thaksin"
transliterate("John Smith", "ja")                              # → "ジョン・スミス"

Returns None when the engine can't deterministically romanize an input. Fail-soft: callers see a clear signal, never a half-translated guess.

Install

pip install -e git+https://github.com/T-Fizz/translit.git#egg=translit-core

Supported pairs

Source	Target	Method
Japanese (kana + kanji)	Latin	pykakasi passport-style + 22 honorifics + reverse-katakana round-trip
Chinese (Hanzi)	Latin	pypinyin (no tones)
Korean (Hangul)	Latin	RR + 36-name surname overlay + 3 honorifics
Russian (Cyrillic)	Latin	BGN/PCGN press-style
Hindi (Devanagari)	Latin	IAST + diacritic strip + schwa deletion
Arabic	Latin	Curated 50-name overlay
Thai	Latin	RTGS via pythainlp
English (Latin)	Japanese (katakana)	alkana + acronym fallback

What it handles

Japanese honorifics including compounds (ちゃん, 先生, 兄ちゃん, にゃん)
CJK source disambiguation (田中 → Tanaka if ja, Tian Zhong if zh)
Multi-kanji morpheme splitting (山田太郎 → "Yamada Taro")
Round-trip katakana → Western name (ヴィクター → "Victor")
Korean traditional surname spellings (Kim/Lee/Park, not Gim/I/Bak)
Korean 2-syllable family names (남궁/황보/사공/제갈/선우/독고)
Russian press-style (Mikhail not Mihail, Fyodor not Fedor, ъе/ье → ye)
Hindi schwa deletion with cluster + aspirated-digraph awareness
Arabic input normalization: tashkeel, alif variants, tatweel, Allah ligature
Half-width katakana / full-width Latin (NFKC)
Punctuation stripping, silent-character dropout detection
Family-first vs given-first name ordering (name_order=)

INTERNALS.md — every edge case, why each library was picked or discarded
DESIGN.md — architecture and API contract for the optional FastAPI service
TENETS.md — the principles behind the deterministic-and-fail-soft choices