Tadabur

A large-scale Quran audio dataset featuring ... reciters and word-level alignment — built for Arabic speech research.

---

Overview

Tadabur covers the entire Qur’an with automatic word-level timestamp alignment and structured JSON metadata for each file. It is the largest and most diverse publicly available Qur’anic speech dataset.

🕐 Audio	1400+ hours
🎙️ Reciters	600+
🔗 Alignment	Word-level timestamps

---

Dataset Structure

Each example contains an audio file and a JSON metadata file:

{
  "reciter_id": 88,
  "surah_id": 3,
  "ayah_id": 82,
  "word_alignments": [
    {
      "word": "أفلا",
      "start": 0.089,
      "end": 0.87
    },
    ...
  ],
  "text_ar_simple": "افلا يتدبرون القران ولو كان من عند غير الله لوجدوا فيه اختلافا كثيرا",
  "text_ar_uthmani": "أَفَلَا يَتَدَبَّرُونَ ٱلْقُرْءَانَ ۚ وَلَوْ كَانَ مِنْ عِندِ غَيْرِ ٱللَّهِ لَوَجَدُوا۟ فِيهِ ٱخْتِلَـٰفًا كَثِيرًا",
  "ayah_duration_s": 10.9,
  "audio_filename": "tadabur_spk0088_S3_A82_db1f8e71_000003.wav"
}

---

Models

Whisper models fine-tuned on Tadabur for Qur'anic ASR, available on HuggingFace:

Model	Base	Status	Link
Tadabur-Whisper-Small	Whisper Small	✅ Available	🤗 HuggingFace

---

Citation

@misc{alherran2026tadabur,
  author = {Alherran, Faisal},
  title  = {Tadabur: A Large-Scale Quran Audio Dataset},
  year   = {2026},
  url    = {https://github.com/fherran/tadabur}
}

---

Contact

For questions or collaboration, feel free to reach out on LinkedIn.

_{Released under CC BY-NC 4.0 · For research and educational use only · Please engage with Qur'anic content respectfully.}