Quranic Phonemizer
المحوّل الصوتي القرآني
Quran text to phonemes with tajweed, waqf, and riwayah awareness. The phonetic ground truth the other projects build on.
- stage
- Stable
- openness
- Open source
The problem, solved once
Quranic pronunciation depends on the riwayah, the tajweed rules in force, and where the reciter stops. General Arabic tools do not model any of these.
The Phonemizer represents those rules once, as a grapheme-to-phoneme layer with waqf and ibtidaa transformations and a tajweed breakdown per sound.
It is published as a Python package and a peer-reviewed paper, and it is the phonetic layer the Aligner and Universal Audio build on.
What you get
- Hafs68 base phonemes plus 5 optional tokens
- Warsh71 base phonemes plus 5 optional tokens
- Waqf and ibtidaa transformations, tajweed breakdown per sound
- Python package on PyPI, paper on OpenReview
Where we are, where we are heading
- nextMore riwayat
- nowWarsh71 base phonemes; the layer the Aligner and Universal Audio build on.
- 2026Peer-reviewed and on PyPI
- 2025Hafs grapheme-to-phoneme68 base phonemes, waqf and ibtidaa transformations, tajweed breakdown per sound.
How to help here
- Improve a pronunciation rule or add a regression test.
- Report a text example the output gets wrong.