Measure what you hear.

This pilot tests an annotation workflow. Text highlights and ayah navigation are candidates awaiting expert confirmation. No automatically estimated interval is presented as a human measurement.

Before marking

1. Listen at normal speed and verify the ayah, target word, and rule applicability. The clip can contain neighboring ayah audio. A navigation warning requires listening, not automatic acceptance.
2. Record whether the reciter actually stops at the target. This is separate from the word’s textual position at the end of an ayah.
3. Mark the target by acoustic evidence. Do not select boundaries to fit a remembered number of haraka.

Operational boundaries

Family	Interval to mark
Qalqala	Release onset to the end of the post-release event; exclude the preceding closure and the following segment. Write down ambiguity or an alternative boundary interpretation.
Geminate ghunna	Onset to end of sustained nasal resonance in the geminate نّ or مّ. Describe uncertain transitions.
Madd wajib	Sustained vowel nucleus before the hamza within the same word, excluding hamza closure/release. Record the actual stopping behavior.
Madd lazim	Sustained vowel nucleus before necessary sukun, excluding the following consonant. Identify kalimi or harfi.

These are provisional acoustic definitions. The study owner and expert panel should agree on difficult examples during training and freeze the definitions before the main study. Qalqala release duration and total stop-consonant duration are different measurements.

Local reference sounds

Independently mark at least five short vowel nuclei and five natural madd intervals in nearby speech at the same local pace. Identify the word and vowel (a, i, or u). Exclude phrase-final lengthening, secondary madd, pauses, ambiguous transitions, and the target itself. References must be distinct and cannot overlap the target or one another.

The analysis uses the median short vowel duration and half the median natural madd duration as two separate estimates of a haraka. Counts and raw milliseconds are visible here; expected counts and normalized target results are withheld from this worklist.

If the clip lacks enough references, explain the shortage in calibration notes. Submit the raw measurement with that limitation. The estimate for a reference family with fewer than five eligible examples remains unavailable. The owner must provide more context where necessary before normalized analysis.

Waveform controls

• Click the waveform to seek. Use Space to play or pause when focus is outside a form field.
• Choose “Edit target” or a reference’s edit button. Press I for its start and O for its end at the cursor. Drag interval edges or enter times directly for fine adjustment.
• Use zoom and the spectrogram to inspect boundaries. “Play selection” and “Loop” replay the selected interval.
• Slower playback is for listening only. All saved coordinates use the original native sample rate.

Saving and independent review

“Save draft” preserves incomplete work. “Submit measurement” checks required judgments, valid boundaries, and reference eligibility. Non-realized events have no target interval; they are never stored as a zero duration.

Unsaved changes have a recovery copy on this device. Only an explicit successful save stores them centrally. Each save has a revision and an audit history. A stale edit is rejected instead of overwriting another save. If a revision conflict occurs, export your local draft, reload to obtain the newest saved revision, then use “Restore last saved” to discard the recovered stale draft before editing.

Your invitation is a private bearer credential, valid for 30 days unless revoked. Sessions last seven days. You can view only your own assigned events and annotations. Do not share invitation links. The owner can revoke access and export study data; no invitation email is sent by this app.

Recordings use study codes to reduce expectation bias, although familiar voices may still be recognized. Independent reviewers must measure overlapping assignments without sharing boundaries. The study owner should preassign at least 20% of events to two experts and resolve disagreements after independent annotation.

Study scope

The planned corpus has murattal recordings of Al-Maida (5), Fatir (35), and Qaf (50) by ten reciters. The initial pilot contains 32 excerpts from four reciters and 148 unreviewed target candidates. The pilot does not yet represent the full corpus.

Differences between these three surahs cannot isolate surah length from surah identity. Calibration, recording provenance, stopping behavior, and recitation route must be considered before interpreting theoretical agreement.