Automatic word alignment

Word timings are automatic and can be early, late or incomplete. No one has listened to every word. When timing is missing, does not fit the words, overlaps ambiguously, exceeds the actual recording duration, or has a reviewed audio problem, Arabic letters for the whole ayah turn green instead. No word timing is guessed.

Audio checks: 11 November 2026

Independent waveform screening has attempted at least one ayah from every surah for every reciter. 1,259 of 62,360 reciter/ayah recordings (2.02% coverage) have been screened against actual audio. 1,134 of 1,140 reciter/surah sample combinations completed strict decoding; six had MP3 frame errors. These percentages describe coverage, not timing accuracy. Checking continues.

Energy and silence checks can expose major timing problems but cannot identify the spoken word. A separate free speech-recognition check is also being used on selected clips without supplying the expected words as a prompt. It can mishear words, invent extra words or place their boundaries badly. Its output is never used to change the Quran text or blindly replace timings.

251 original structural fallbacks plus nine reviewed audio exceptions give 260 listed whole-ayah fallbacks. Runtime checks may add a fallback for an individual playback.

New reviewed audio exceptions

• ajmi 17:56 - Repeated phrase: independent whole-clip and cropped ASR detect earlier occurrence while supplied timestamps follow later occurrence; whole-ayah fallback avoids misleading first occurrence.
• ajmi 112:1 - Large disagreement between supplied first-three-word intervals and independent whole-clip plus cropped ASR. Intro/short-crop recognition is uncertain; whole-ayah fallback until reliable alignment review.
• dosari 53:32 - Repeated phrase: independent whole-clip ASR and separate crops identify wa-idh antum ajinnatun at17.6-21.36s and21.4-24.64s. Source word19/20timings follow later occurrence; conservative whole-ayah fallback avoids misleading earlier occurrence.
• ajmi 5:61 - MP3 frame error detected. Tolerant decoding continues, but word timing is withheld until this recording can be checked reliably.
• ajmi 21:57 - MP3 frame error detected. Tolerant decoding continues, but word timing is withheld until this recording can be checked reliably.
• ajmi 22:40 - MP3 frame error detected. Tolerant decoding continues, but word timing is withheld until this recording can be checked reliably.
• ajmi 32:16 - MP3 frame error detected. Tolerant decoding continues, but word timing is withheld until this recording can be checked reliably.
• ajmi 34:28 - MP3 frame error detected. Tolerant decoding continues, but word timing is withheld until this recording can be checked reliably.
• ajmi 46:18 - MP3 frame error detected. Tolerant decoding continues, but word timing is withheld until this recording can be checked reliably.

Repeated-phrase cases were checked with separate audio crops. The uncertain Ajmi 112:1 case uses a conservative fallback, not a claim that speech recognition is correct. For frame-error cases, audio may still play in tolerant browsers; whole-ayah fallback does not repair the recording.

Timing: QuranLab and ©2016 Collin Fair / cpfair/quran-align, CC BY 4.0. Retained timing values are unchanged; nine unreliable timing entries have been disabled. Fields are reformatted and split by surah. Source · Licences. Recordings are streamed only from each table's exact EveryAyah source; no audio is hosted. Recording copyrights remain with their reciters/producers. No endorsement implied.