9:41

Brahui Dictionaryبراہوئی لغت

Published parallel sentences. Connected speech is where a word list stops helping — intonation and vowel reduction only show up in a whole clause.
Every sound of Brahui with the Urdu letter that carries it. Drill repeats it three times; Once says it in a frame word. ڷ /ɬ/ exists in no TTS voice on earth — the substitution you hear is the voice giving up, not a spelling error.
Latin → Brahui script

Brahui orthography — store this

 
/ /

Vocalized for TTS — send this to the voice

 

Fallbacks

Diacritic-stripping 
Devanagari (Hindi voice) 
English respelling 
ASCII accepted: sh=ş · zh=ź · LH=ļ · GH=ģ · T=ŧ · D=đ · R=ŕ · N=ń · aa=á · ii=í · uu=ú.

Not shortcuts, because they destroy real contrasts: tt dd rr are genuine geminates (baŧŧa, izzat) and kh gh ch are aspirated loan consonants (khuriyá, bhallo). For /x/ and /ɣ/ type x and ģ.

A capital at the start of a word is read as ordinary capitalisation, so is and not ńá. Mid-word the shortcuts still hold, so baTTa is baŧŧa. If a word really begins with ŧ ń ş ź ģ ŕ đ ļ, type that letter.
Brahui script → Latin
Latin 
Say as 
IPA 
Where every entry comes from
The Latin is the source of truth. Script, IPA, Devanagari and the respelling are regenerated from it on every read, so a fix to the engine propagates to every entry at once.
Books used
Grierson, G. A. — Linguistic Survey of India, Vol. IV: Muṇḍā and Dravidian Languages (Calcutta, 1906). Brahui occupies pp. 619–681: sketch grammar, three specimens, and the Standard List of Words and Phrases in the Brahui of Kalat (items 1–241), which is where the LSI entries and most of the sentences in Phrases were read off.
PDF (66 MB) · archive.org item
Bray, Denys de S. — The Brahui Language, Part I: Introduction and Grammar (Calcutta, 1909). Used to settle readings the 1906 scan had damaged — khākhar “fire”, tuh “moon”, kātum “head”, bīsh “donkey”, “slave” — and for the BRAY entries.
archive.org item
Universal Declaration of Human Rights, Article 1 — Brahui parallel text, Brahui Language Board / University of Balochistan. The engine in this app is validated against it: 11 of 11 words render exactly.
Ali, Liaquat & Masato Kobayashi — Brahui Texts (ILCAA, Tokyo University of Foreign Studies, 2024). Story titles are glossed by the authors, so those pairings are published; everything else drawn from the corpus is marked CORP and still needs a speaker.
Data
Bulk add
Why the diacritics matter
Urdu text normally omits short-vowel marks, and a TTS engine guesses them from its own Urdu lexicon — which contains no Brahui words. Every output here is fully vocalized (zabar / zer / pesh / sukun) so the engine reads exactly what is written and the long–short vowel contrast survives. Feed the engine the vocalized string, never the bare consonant skeleton.