Quran Recitation AI
An open-source system for following Quran recitation word by word, aligning recognized speech with Quranic text, and identifying skipped verses or page mismatches.
Problem and purpose
Following a recitation against Quranic text requires matching spoken words to their expected position and noticing when a verse is skipped or the reader is on a different page. The project is designed to provide word-level tracking and feedback during practice.
How the system is described
The public repository documents a pipeline that captures audio, transcribes it with a configurable speech-recognition backend, normalizes the recognized text, aligns segments to Quranic text using fuzzy matching, and analyzes sequence order.
- Word-by-word recitation tracking and alignment.
- Skipped-verse and page-mismatch detection.
- Speech recognition options documented in the repository: Groq Whisper API or local NVIDIA NeMo.
- Levenshtein-based fuzzy matching for text alignment.
My role and engineering
Yahya Mahroof created and maintains the project. Its engineering combines a Python backend, Flask-SocketIO session communication, browser audio capture, FFmpeg audio conversion, Arabic speech recognition, and sequence-aware text alignment.