Recognize the reciter
An ECAPA-TDNN model creates a voice embedding. ONNX Runtime runs inference, then cosine similarity ranks matches against profiles for 20 reciters.
AUDIO · AI · WEB
Quranic Shazam.
Discover who is reciting and which verse you’re hearing. An audio recognition app for Qur’an recitation, with live listening and file uploads.
A LOOK INSIDE
Two ways to explore a recitation.
BEHIND THE EXPERIENCE
From sound to a meaningful match.
Find the verse
A Quran-tuned Whisper model transcribes the Arabic audio. Normalized text is matched against 6,236 verses to find likely surah and ayah matches.
Listen for more context
When a verse match is ambiguous, progressive listening gathers more audio context. Uncertain results stay marked as uncertain.
BUILT WITH
The technology behind it.
- React 19
- TypeScript
- Vite
- Tailwind CSS 4
- Python
- FastAPI
- ONNX Runtime
- faster-whisper
- Docker
- Render
Frontend: React 19 and TypeScript, built with Vite and Tailwind CSS 4. Browser microphone capture and file uploads connect to the API.
Backend: Python and FastAPI, with WebSockets for live reciter recognition. ONNX Runtime handles speaker embeddings; faster-whisper and CTranslate2 handle verse transcription.
Delivery: Docker containers on Render, with Nginx serving the frontend. GitHub Actions checks the backend and frontend before deploying affected services.