cpfair/quran-align

Performs automatic speech recognition on Qur'anic audio using CMU Sphinx with speaker-specific acoustic models and Qur'an-specific language models, then aligns recognized words to reference text via edit-distance matching and MFCC-based refinement to syllable boundaries. Outputs JSON timing data compatible with EveryAyah recordings, with pre-generated datasets available for multiple qaris that achieve 98.5-99.9% word segmentation accuracy and <73ms average timing deviation validated against independent implementations.

241 stars. No commits in the last 6 months.