Word timing data by Collin Fair (cpfair), quran-align, release 2016-11-24. https://github.com/cpfair/quran-align/releases/tag/release-2016-11-24 Creative Commons Attribution 4.0 International (CC BY 4.0) https://creativecommons.org/licenses/by/4.0/ Recordings: EveryAyah Alafasy_128kbps and Husary_64kbps. Adaptations: split data by surah; ordered by ayah; omitted zero or negative intervals; mapped exceptional word boundaries to Tanzil Uthmani v1.1; preserved shared timings as ranges. Timing revision 2: restored the 41 previously unrepresented word occurrences (25 Alafasy and 16 Husary) using the exact EveryAyah recordings, retaining the recording-specific clock. Measured the word boundaries with a Quran-trained wav2vec2 CTC acoustic recognizer, confirmed them in a second acoustic window, and repositioned intersecting neighboring cues. For Husary 6:165 the acoustic model emits فيما as one unit; the distinct م onset marks the beginning of the printed second word ما. Unrelated source intervals that cover multiple written words retain their shared timings. These are machine-assisted timing estimates. Recognition model: rabah2026/wav2vec2-large-xlsr-53-arabic-quran-v_final (Apache 2.0); ONNX conversion by mostafanewir70, revision f90650aac5d9dcff15230a19998c0027f45b33f1. https://huggingface.co/mostafanewir70/quran-recitation-ctc-onnx Quran recitations are streamed from Al Quran Cloud / Islamic Network under the provider's terms for downloading and streaming for personal and educational use. Attribution belongs to each named reciter. Quran recordings are not licensed by the quran-align data license. Al Quran Cloud states that it does not own the recordings' copyrights and may request removal from a commercial product. https://alquran.cloud/cdn https://alquran.cloud/terms-and-conditions Word timing data © 2016 Collin Fair (cpfair), quran-align, release 2016-11-24. Creative Commons Attribution 4.0 International. https://github.com/cpfair/quran-align/releases/tag/release-2016-11-24 https://creativecommons.org/licenses/by/4.0/ Four additional recording-specific editions: Abdul_Basit_Murattal_64kbps, Abdurrahmaan_As-Sudais_192kbps, Abu_Bakr_Ash-Shaatree_128kbps, and Hani_Rifai_192kbps. Verse data were ordered and divided by surah. The three known Uthmani token splits and transparent disjoint-letter offsets were normalized to the displayed Tanzil Uthmani text. Nonpositive intervals and ambiguous overlap time were omitted. Missing source cues were not replaced with proportional word estimates. The timing source uses forced alignment and its timestamps are approximate. The other eleven recitation editions have no matched timing data in this collection. Timing revision for the additional four recitations. Segments were rechecked against the actual Islamic Network verse recordings and the displayed Tanzil Uthmani word sequence. Restored 164 previously untimed printed word positions in 130 verse recordings: 64 Abdul Basit Murattal, 30 Sudais, 31 Abu Bakr Ash-Shaatree, and 39 Hani Rifai. 161 restored positions have an individual word cue. Three positions retain a shared measured phrase when the internal syllables in the recording could not be distinguished confidently (Sudais 5:95, Rifai 38:86 and 41:29). Repeated recitation is represented in order, including Abdul Basit 6:70 and Ash-Shaatree 41:40. Word boundaries are machine-assisted acoustic estimates, not proportional timing divisions. Quran-specific recognition model: rabah2026/wav2vec2-large-xlsr-53-arabic-quran-v_final, ONNX conversion mostafanewir70/quran-recitation-ctc-onnx (f90650aac5d9dcff15230a19998c0027f45b33f1, Apache 2.0). Original quran-align data credits and license apply as described above. https://huggingface.co/mostafanewir70/quran-recitation-ctc-onnx