about

shazamzam listens to a few seconds of Quran recitation and tells you which surah and ayah you are hearing. It works for recordings and for live recitation, even imperfect recitation.

Under the hood it is two steps: a speech model tuned for Quranic recitation transcribes what it hears, and the (often imperfect) transcript is then located in the fixed text of the Quran (Hafs reading). Because the text is fixed and known, a rough few-second sample is usually enough to find the exact ayah.

Your audio is used only to answer your request: it is processed in memory and never stored. See privacy.

standing on open shoulders

shazamzam is only possible because these projects are open:

• Tarteel AI's whisper-base-ar-quran model, a Quran-tuned fine-tune of OpenAI's Whisper
• The Tanzil project: verified Quran text, the Saheeh International translation, and the transliteration shown with every result
• faster-whisper and CTranslate2, which make the speech model fast on a small CPU server
• EveryAyah, whose per-ayah recitation audio was used to test the pipeline
• The Amiri Quran typeface by Khaled Hosny, and Inter by Rasmus Andersson
• FastAPI and Zola

Quran text, translation, and transliteration come from the Tanzil project (tanzil.net) and are used with attribution, as their license asks.