Quran ASR Models Comparison

Upload or record a short Quranic recitation and compare transcriptions side-by-side.

Models:

IJyad/whisper-large-v3-Tarteel (large-v3, high accuracy)
deepdml/whisper-medium-ar-quran-mix-norm (medium)
naazimsnh02/whisper-large-v3-turbo-ar-quran (turbo, fast & accurate)
Habib-HF/tarbiyah-ai-whisper-medium-merged (medium, merged general + Quran)

Record from mic or upload audio file (WAV/MP3, preferably Quran recitation)

IJyad/whisper-large-v3-Tarteel

Transcription

deepdml/whisper-medium-ar-quran-mix-norm

Transcription

naazimsnh02/whisper-large-v3-turbo-ar-quran

Transcription

Habib-HF/tarbiyah-ai-whisper-medium-merged

Transcription

Notes:

Best for short Quran recitations (mic recordings are usually <30s).
Transcriptions are plain Arabic text (no tashkeel/diacritics in most cases).
GPU highly recommended — CPU will be slow.
These models are Quran-specialized; general Arabic speech may not work well.