# Validated Chabad Trop Motif Table — R' Michoel Slavin recordings

**Source audio:** R' Michoel Slavin (baal korei of 770), Torah readings distributed by Chabad.org, obtained as direct MP3s from TorahDownloads (torahcdn.net). 8 parshiyot measured: Bereishit, Vayeira, Chayei Sarah, Toldot, Vayeshev, Tzav, Shemini, Masei. No purchases, no sign-ins.

**Source text:** Sefaria API v3, "Tanach with Ta'amei Hamikra" (public domain, WLC-based). Parsed with a trope parser built for this validation (Unicode codepoints verified against the actual text; pashta/qadma and mahpach/yetiv disambiguated by position relative to the stressed vowel; paseq attaches to the preceding word as legarmeh; the zarqa/tsinnorit Unicode swap confirmed — zarqa is U+05AE ZINOR in this text).

**Working model being validated:** the 23-motif table in `research-chabad-trop.md` §2.5, transcribed from Stephen Schmidt's 5-page Eastern Ashkenazi sheet-music chart (Agudath Achim, Altoona PA — Conservative, not Chabad-labelled), in C major.

**Method:** pitch contours extracted with a harmonic-product fundamental estimator (octave-robust, verified against raw spectra at multiple points) and cross-checked with CREPE (small model, Viterbi) and librosa pyin/yin. Verse/phrase locations found with faster-whisper (small/base models, Hebrew, word timestamps) plus RMS pause segmentation. Motifs compared by melodic SHAPE (relative interval pattern), not by exact equal-tempered pitch — see §0 for why.

---

## 0. The two findings that condition everything else

### 0a. Slavin's intonation is approximate, not equal-tempered
Phrase-final pitches across the first 190 s of Bereishit do **not** cluster on discrete semitones. They form a continuum from ~165–187 Hz (MIDI 52–54), centred on ~174 Hz (MIDI 52.6), with a standard deviation of ~2.9 semitones across phrase finals. The same is true of the reciting tone: it lives in a band (E3–F#3), not on one pitch. Slavin is a fast reader chanting speech-like, not a singer hitting exact scale degrees. The chart's exact degree sequences (e.g. "2 2 3 1") are therefore an **idealisation**. What can be validated is the melodic SHAPE — the pattern of ups, downs, leaps and cadences — not the exact semitone of every note. Every "CONFIRMED" below means "the shape matches the chart", not "every note is exactly on the chart's pitch".

### 0b. The tonic is not pinned down by the audio alone
Two hypotheses both fit part of the data and neither fits all of it:

| Hypothesis | Evidence for | Evidence against |
|---|---|---|
| **Tonic = D3 (147 Hz)** | Reciting band (52–54) = degrees 2–3, matching the chart's reciting tones (1–2) and phrase endings (etnachta→2, tipcha→3, zakef katan→2). Low dips (47, 49) = degrees 6–7 below, matching revia/tevir. High notes (57, 59, 61, 62) = degrees 5, 6, 7, 1-octave. | Verse-final (sof pasuk) of Gen 1:1 is A2 (110 Hz) = degree 5 **below** the tonic, not degree 1. Parsha-final (sof k'ria) ends on F#2 (42) = degree 3 below, not degree 6 below. |
| **Tonic = A2 (110 Hz)** | Verse-final (sof pasuk) of Gen 1:1 is A2 = degree 1, matching the chart. Parsha-final (sof k'ria) ends on F#2 (42) = degree 6 below, matching the chart's sof k'ria ending on degree 6 below. | Reciting band (52–54) = degrees 5–6, not the chart's reciting tones (1–2). Phrase finals cluster on 5–6, not on 1–2. |

**Resolution adopted for this table:** degrees are expressed relative to **D3 (MIDI 50, 147 Hz)** as the operational tonic, because it best explains the bulk of the melodic content (the reciting band and the phrase endings, which are most of what a learner hears). The low verse/parsha-final cadences (A2, F#2) are reported as a Chabad-specific cadential pattern that departs from the chart. The A2-tonic reading (recite on the dominant, cadence on the tonic) remains a live alternative; the audio does not settle it. A cantor's confirmation is the only thing that would.

---

## 1. Validated motif table (degrees relative to D3 = 147 Hz, MIDI 50)

"Chart" = the working model from `research-chabad-trop.md` §2.5 (degrees in C major, transposed here to D major for comparison: 1=D3 50, 2=E3 52, 3=F#3 54, 4=G3 55, 5=A3 57, 6=B3 59, 7=C#4 61; below tonic: 7=C#3 49, 6=B2 47, 5=A2 45, 4=G2 43, 3=F#2 42).

"Measured" = the shape extracted from Slavin's audio. "Verdict" is CONFIRMED (shape matches), CORRECTED (shape differs systematically), PARTIAL (some evidence, not conclusive), or UNCERTAIN (no clean measurement).

### Disjunctives (phrase-ending taamim)

| Ta'am | Chart (degrees) | Measured (Slavin) | Verdict |
|---|---|---|---|
| **Sof pasuk** (siluk) | 2 2 3 1 (ends on tonic) | Gen 1:1 final note is a sustained **A2 (110 Hz, degree 5 below)**, 0.9 s, decaying into a 4 s pause. Other verse finals vary (see §3). The 2-2-3 rise is present in the reciting band, but the cadence lands **low**, not on the tonic. | **CORRECTED** — cadence ends on degree 5 below, not degree 1. (Single clean verse; other verse finals are less clear — see §3.) |
| **Etnachta** | 1 1 2 (ends on 2) | Phrase finals in etnachta positions measure ~165–185 Hz (degrees 2–3). Gen 1:4 etnachta (~33 s) ends ~185 Hz (degree 3); Gen 1:5 etnachta (~44 s) ends ~165 Hz (degree 2). The 1-1-2 shape (two level notes then a step up) is present in the reciting band. | **CONFIRMED** (shape; ends on 2, occasionally 3). |
| **Zakef katan** | 5 5 3 2 (ends on 2) | Gen 1:2, 1:4, 1:6 zakef-katan phrases show a high note (57, degree 5) followed by a descent through 54 to 52 (degrees 3→2). The 5-5-3-2 shape is present. | **CONFIRMED** |
| **Zakef gadol** | 1 1 3 5 3 2 | No clean isolated example measured (18× in Bereishit, but always run together with neighbouring phrases in the audio segmentation). | **UNCERTAIN** |
| **Tipcha** | 2–3 (before etnachta/sof pasuk); 2–5–3 (end of reading) | The most common motif (270× in Bereishit). Tipcha phrases sit in the reciting band (52–54) with a 2→3 upward step. The 2–5–3 end-of-reading variant was not isolated. | **CONFIRMED** (2–3); end-of-reading variant UNCERTAIN |
| **Revia** | 3 3 2 1 7 6 (ends on 6 below) | Gen 1:2 revia (וְהָאָרֶץ, 9.0–11.2 s): notes 57, 53, 55, 54, then 54, 55, 53, 52, **47** — the phrase descends to degree 6 below (B2, 47) at the end. The low dip to 6-below is the revia's signature and it is present. | **CONFIRMED** |
| **Tevir** | 2 1 7 1 2 (darga lead-in 5 5 4 3 2, or merkha 2 2) | Gen 1:4 and Gen 1:8 tevir phrases show the low dip to degree 7 below (49, C#3) and a return to the reciting band. The darga lead-in descent (5→2) is present in the Gen 1:4 tevir phrase. | **CONFIRMED** (moderate confidence) |
| **Pashta** | 1–5 (mahpach lead-in 5 5 1) | Gen 1:2 pashta (הָיְתָה תֹהוּ) and Gen 1:5 pashta (לָאוֹר) show the 1→5 leap (low note ~50 then high ~57). The mahpach lead-in (high 5-5-1) is present before pashta phrases. | **CONFIRMED** |
| **Yetiv** | 5 5 (then munach 3 3 2 + katon 5 5 3 2) | Present in Gen 1:31 and Gen 2:3, but no clean isolated measurement (run together with neighbouring phrases). | **UNCERTAIN** |
| **Segol** | 1 2 1 (with a B♭ melisma) | Gen 1:7 segol (וַיַּעַשׂ אֱלֹהִים אֶת הָרָקִיעַ, 60–62 s): notes span 48–64 with a high note (64, E4). A melisma is present, but the B♭ (♭7) in the chart's melisma was **not** confirmed — no stable B♭ was measured. | **UNCERTAIN** (melisma present; B♭ not confirmed) |
| **Zarka** | 1 1 + descending melisma (♭7…4) | 0 occurrences in Bereishit (the zarqa codepoint U+05AE appears only in the target verses). Not measured. | **UNCERTAIN** (not in sample) |
| **Telisha gedola** | recite ×8, close 3 2 3 | 3× in Bereishit. No clean isolated measurement. | **UNCERTAIN** |
| **Telisha ketana** | recite ×8, close 3 2 1 | 15× in Bereishit. No clean isolated measurement. | **UNCERTAIN** |
| **Geresh** | 5 5 2 | Gen 1:9 geresh (יִקָּווּ הַמַּיִם): high notes (57, degree 5) present, falling to the reciting band. | **CONFIRMED** (moderate confidence) |
| **Gershayim** | 1 2 3 2 1 2 3 | 7× in Bereishit. No clean isolated measurement. | **UNCERTAIN** |
| **Kadma-v'azla** | kadma 1 1; azla 1 3 5 3 6 5 | No clean isolated measurement (the azla's rise to degree 6 was not isolated). | **UNCERTAIN** |
| **Pazer** | 1 1 2 2 3 6 5 3 2 | 2× in Bereishit. No clean isolated measurement of a standalone pazer. (Its shape is relevant to the shalshelet — see below.) | **UNCERTAIN** |
| **Sof k'ria** (end-of-reading coda) | merkha 2 2 / tipcha 2 5 3 / merkha 3 1 / 2 3 2 6(below) | Gen 6:8 (last verse of Bereishit, 1730–1747 s): the coda recites in the 51–54 band, rises to 57–59, and ends with a sustained **descent to F#2 (42)** over the last ~2.5 s. The final low note matches the chart's ending on degree 6 below **if the tonic is A2**; relative to D3 it is degree 3 below. The broad shape (recite → rise → low close) matches the chart. | **CONFIRMED** (broad shape; final-note degree depends on the tonic interpretation — see §0b) |

### Conjunctives (lead-in taamim)

| Ta'am | Chart | Measured | Verdict |
|---|---|---|---|
| **Munach** | 3 1 (before etnachta); 3 3 2 (before katon/yetiv); 1 1 (before zarka/telisha/segol) — varies | Munach notes sit in the reciting band (52–54) in all measured positions, consistent with the chart's figures (all of which are in the 1–3 range). The contextual variation was not separately resolved. | **CONFIRMED** (in the reciting band) |
| **Kadma** | 1 1 (inside kadma-v'azla) | Not isolated. | **UNCERTAIN** |
| **Mahpach** | 5 5 1 (before pashta) | High notes (57, degree 5) present immediately before pashta phrases, falling to the reciting band — consistent with 5-5-1. | **CONFIRMED** (with pashta) |
| **Merkha** | 7–2–7 (before tipcha); 2–2 (before tevir/sof k'ria); 3–1 (before sof pasuk) — varies | Merkha notes sit in the reciting band (52–54) with the expected small up/down figures. The three contextual variants were not separately resolved. | **CONFIRMED** (in the reciting band) |
| **Darga** | 5 5 4 3 2 (before tevir) | The descent 5→2 is present before tevir phrases (Gen 1:4). | **CONFIRMED** (with tevir, moderate confidence) |

### Rare taamim (not in the Schmidt chart; C-confidence in the report)

| Ta'am | Report's claim | Measured (Slavin) | Verdict |
|---|---|---|---|
| **Shalshelet** | Forum: "basically 2 Pazers". 4× in Torah. | Measured all 4: Gen 19:16 (וַיִּתְמַהְמָהּ, Vayeira 657.7–660.2 s, 2.5 s), Gen 24:12 (וַיֹּאמַר, Chayei Sarah 437.0–442.6 s, 5.5 s), Gen 39:8 (וַיְמָאֵן, Vayeshev 922.2–927.7 s, 5.5 s), Lev 8:23 (וַיִּשְׁחָט, Tzav 1139.7–1141.8 s, 2.1 s). Shape in all four: recite on ~52 (degree 2), then a **rising scale/glide** to a high peak (degree 5–6, up to degree 1-octave in Gen 39:8), then a sharp drop to degree 7 below (49), then back to the reciting tone. The two longer examples (Gen 24:12, Gen 39:8) clearly show **two rising units** (first peak ~57–59, dip, second peak ~59–62, fall) — consistent with the "2 Pazers" idea in broad shape. But the note-level detail does **not** match the chart's pazer (1-1-2-2-3-6-5-3-2): the rises are gradual scales, not the pazer's stepped pattern with a leap to 6, and the second unit peaks higher (degree 1-octave) than the pazer's 6. | **CORRECTED in detail, CONFIRMED in broad shape** — two melismatic rising units, not two literal pazers. |
| **Karnei parah** | Forum: "combination of Mercha Kefulah and Telisha Gedola". 1× in Torah (Num 35:5). | Num 35:5 (Masei ~749.5–755 s, on בָּאַמָּה, preceded by yerach ben yomo on אַלְפַּיִם): an extended melisma with **two high plateaus** (~55, then ~58–60) separated by low dips to ~47 (degree 6 below). Consistent with a compound motif (two parts). Exact note values uncertain (octave artifacts in the measurement). | **PARTIAL** — compound melisma confirmed; exact notes uncertain |
| **Mercha kefula** | No motif given. 5× in Torah. | Lev 10:1 (לֹ֦א, Shemini 323.0–327.0 s): recite ~52, dip to ~47, rise to a high plateau ~57–59 (peak 59, degree 6), fall back to ~52, dip to ~46. A clear high melisma (~2.5 s on a one-syllable word). Gen 27:25 (ל֦וֹ, Toldot ~1326–1328 s): noisy; shows a low dip (~43–45) rather than a high rise — **inconsistent** with the Lev 10:1 shape, so the Gen 27:25 measurement is treated as unreliable (the window may have caught the wrong word, or the method failed there). | **PARTIAL** — Lev 10:1 shows a high melisma to degree 6; Gen 27:25 unclear |
| **Yerach ben yomo** | No motif given. 1× in Torah (Num 35:5). | Num 35:5 (on אַלְפַּיִם, immediately before the karnei parah): part of the extended melisma described under karnei parah (the first high plateau, ~55). Not separately resolvable from the karnei parah. | **PARTIAL** — present as the first part of the compound melisma |
| **Munach legarmeh** | Contextual munach figure. | Present in the text (Gen 1:5 mahpach+paseq, Gen 1:29–30 munach+paseq, Num 35:5 munach+paseq). No clean isolated pitch measurement. | **UNCERTAIN** |

---

## 2. What changed vs the working model

1. **Sof pasuk — CORRECTED.** The chart ends the verse on the tonic (degree 1). Slavin's Gen 1:1 ends on a sustained A2 (110 Hz, degree 5 below the D3 operational tonic). The 2-2-3 rise is present, but the cadence lands low. (Whether this is "degree 5 below D3" or "degree 1 of an A2 tonic" is the §0b ambiguity.)
2. **Shalshelet — CORRECTED in detail.** The forum's "basically 2 Pazers" is right in broad shape (two rising melismatic units) but wrong at note level: the rises are gradual scales, not the pazer's stepped 1-1-2-2-3-6 pattern, and the second unit peaks higher (degree 1-octave) than the pazer's degree 6.
3. **Karnei parah / yerach ben yomo / mercha kefula — first measurements.** The report had no motif for these (C confidence, forum only). Now measured: karnei parah + yerach ben yomo form a compound melisma with two high plateaus (~55, ~58–60) separated by low dips (~47); mercha kefula (Lev 10:1) is a high melisma peaking at degree 6 (59). These are the first measured values for these taamim in the Chabad tradition (that this validation found).
4. **Sof k'ria — CONFIRMED in shape, with a tonic caveat.** The coda's shape (recite → rise → low close) matches the chart. Its final note (F#2, 42) matches the chart's "degree 6 below" only under the A2-tonic reading; under the D3 tonic it is degree 3 below.
5. **No other motif changed.** The 12 motifs marked CONFIRMED match the chart's shape. None of the common motifs was found to have a systematically different shape from the chart — the Chabad-vs-Lithuanian difference, at least in Slavin's reading, is **not** in the common motifs' shapes. It is in the cadences (sof pasuk ending low) and in the rare melismas (shalshelet detail).

---

## 3. What remains uncertain or unverified

- **The tonic (§0b).** The single biggest open question. D3 explains the reciting band and phrase endings; A2 explains the verse/parsha-final cadences. The motif table's degrees are expressed relative to D3 by convention, but if the true tonic is A2, every degree in §1 shifts (the reciting band becomes degrees 5–6, and the "correction" to sof pasuk disappears). A cantor's confirmation is needed.
- **Verse-final (sof pasuk) variation.** Only Gen 1:1's final note (A2) was measured cleanly. Other verse finals measured (Gen 1:2 ~56, Gen 1:3 ~61, Gen 1:4 ~50, Gen 1:5 ~53, Gen 1:6 ~47, Gen 1:7 ~54, Gen 1:8 ~49, Gen 1:9 ~57) are scattered — this is likely a verse-boundary alignment error (the "last phrase" sampled was not always the sof-pasuk phrase), not a real variation. A clean measurement of ~10 verse finals is owed.
- **10 of the 23 chart motifs have no clean measurement:** zakef gadol, yetiv, segol (B♭ unconfirmed), zarka (not in the Bereishit sample), kadma, telisha gedola, telisha ketana, gershayim, kadma-v'azla, pazer (standalone). They occur in the audio but were always run together with neighbouring phrases (Slavin pauses at only ~70% of phrase boundaries: 773 audio phrases vs 1113 text phrases in Bereishit), so they could not be isolated by pause segmentation.
- **Mercha kefula (Gen 27:25).** The measurement contradicts the Lev 10:1 shape (low dip vs high rise). One of the two measurements is wrong (likely Gen 27:25, where the window may have caught the wrong word). Needs re-measurement.
- **Segol's B♭.** The chart's segol melisma includes a B♭ (♭7). No stable B♭ was measured in the Gen 1:7 segol. The melisma is present but the flattened note is unconfirmed.
- **Zarka, telisha gedola/ketana, gershayim, kadma-v'azla, pazer, yetiv, zakef gadol, kadma, munach legarmeh.** No clean isolated pitch measurement. These need either a slower/deliberate recording or manual (human) verification.
- **Etnachta's reported Slavin variation.** The report flagged (from a Chabad.org forum comment) that Slavin flattens the etnachta to near-monotone, vs the chart's 1-1-2. The measured etnachta phrases do show a small rise (2→3) rather than a flat monotone, but the rise is small (~1–2 semitones) and within the intonation noise — so the "flattened etnachta" claim is **neither confirmed nor refuted** by this measurement. It remains an open question.
- **Alignment method.** Whisper's word timestamps on chanted Hebrew are unreliable (it hallucinates/fills in words and its timings shift by seconds). RMS pause segmentation under-counts phrases (Slavin doesn't pause at every disjunctive). Both were used together with manual cross-checking, but a forced-alignment tool trained on chant (or manual annotation) would be more reliable.

---

## 4. Summary counts (23 chart motifs)

| Verdict | Count | Motifs |
|---|---|---|
| **Validated as-is (CONFIRMED)** | **12** | Etnachta, zakef katan, tipcha, revia, tevir, pashta, munach, mahpach, merkha, darga, geresh, sof k'ria |
| **Corrected** | **1** | Sof pasuk (cadence ends low, not on the tonic) |
| **Uncertain** | **10** | Zakef gadol, yetiv, segol, zarka, kadma, telisha gedola, telisha ketana, gershayim, kadma-v'azla, pazer |

Plus the 5 rare taamim not in the chart: shalshelet (corrected in detail), karnei parah (partial), mercha kefula (partial), yerach ben yomo (partial), munach legarmeh (uncertain).

**Single biggest blocker to a trustworthy Chabad motif table:** the tonic is not pinned down, and Slavin's intonation is approximate (speech-like, not equal-tempered). Together these mean the table's exact degree sequences cannot be validated note-by-note from this audio alone — only the melodic shapes can. The way past it is human validation: a cantor (ideally one who knows Slavin's tradition) confirming the intended tonic and the intended degrees for each motif, ideally against a slower, deliberate teaching recording where each motif is sung in isolation. The audio analysis in this file establishes the shapes and the cadential pattern; it does not, by itself, settle the degrees.

---

## 5. Evidence log (measurements)

| Measurement | Location | Key numbers |
|---|---|---|
| Reciting band / phrase-final cluster | Bereishit 0–190 s, 90 phrase finals | Centre ~174 Hz (MIDI 52.6), range 165–187 Hz, std ~2.9 semitones; mode MIDI 53 (F3) |
| Verse 1 final note (sof pasuk) | Bereishit 4.05–4.95 s | Sustained A2, 108–112 Hz (MIDI 45), 0.9 s, then a 4.0 s pause |
| Sof k'ria final note | Bereishit 1744.2–1746.8 s | Descent 59.4 → 41.5–43 Hz (MIDI ~42, F#2), sustained ~2.5 s |
| Shalshelet Gen 19:16 | Vayeira 657.7–660.2 s | 2.5 s; recite ~52, rise to ~58, drop to 49, back to ~52 |
| Shalshelet Gen 24:12 | Chayei Sarah 437.0–442.6 s | 5.5 s; two rises (peak ~57.5, dip to 49.4, peak ~59.4, fall to ~52) |
| Shalshelet Gen 39:8 | Vayeshev 922.2–927.7 s | 5.5 s; two rises (peak ~59, dip to ~52, peak ~62, fall to ~54) |
| Shalshelet Lev 8:23 | Tzav 1139.7–1141.8 s | 2.1 s; noisy; high notes ~60 alternating with low ~47 |
| Mercha kefula Lev 10:1 | Shemini 323.0–327.0 s | Rise to high plateau ~57–59 (peak 59), fall to ~52, dip to ~46 |
| Mercha kefula Gen 27:25 | Toldot ~1326–1328 s | Noisy/ambiguous; low dip ~43–45 — contradicts Lev 10:1; treat as unreliable |
| Karnei parah + yerach ben yomo Num 35:5 | Masei 749.5–755.0 s | Two high plateaus (~55, ~58–60) separated by dips to ~47 |
| Revia Gen 1:2 | Bereishit 9.0–11.2 s | Descent to 47 (degree 6 below) at phrase end |
| Etnachta Gen 1:4 / 1:5 | Bereishit ~33 s / ~44 s | Final ~185 Hz / ~165 Hz (degrees 3 / 2) |
| Audio/text phrase counts (Bereishit) | Whole parsha | 773 audio phrases (RMS pauses) vs 1113 text phrases — Slavin pauses at ~70% of disjunctives |
| Recordings | TorahDownloads (torahcdn.net), "Distributed by Chabad.org" | Bereishit 1747 s, Vayeira 2051 s, Chayei Sarah 1440 s, Toldot 1485 s, Vayeshev 1453 s, Tzav 1351 s, Shemini 1210 s, Masei 1417 s; 32 kbps MP3 |

## 6. Files produced by this validation

- `~/workspace/lein-practice/validate/trope.py` — trope parser (Sefaria text → per-word trope labels, with position disambiguation and legarmeh handling).
- `~/workspace/lein-practice/validate/phrases.py` — phrase builder (verses → phrases by disjunctive) + the working-model degree table.
- `~/workspace/lein-practice/validate/measure.py` — harmonic-product pitch measurer (octave-robust fundamental estimation + stable-note extraction).
- `~/workspace/lein-practice/validate/segment.py` — RMS pause segmentation of a recording.
- `~/workspace/lein-practice/validate/whisper_bereishis_300.json`, `whisper_gap_5_30.json` — whisper word-level alignments (first 5 min of Bereishit; note the reliability caveats in §3).
- `~/workspace/lein-practice/validate/gen1.json`, `target-verses.json`, `bereishis-text.json` — Sefaria texts used.
- `~/workspace/lein-practice/validate/crepe_bereishis_180_*.npy` — CREPE pitch track (small model) for the first 180 s of Bereishit.
- `~/workspace/lein-practice/audio/` — the 8 Slavin MP3s (renamed by parsha).
- `~/workspace/lein-practice/audio/bereishis-segments.json` — RMS pause segmentation of Bereishit.

## 7. Method notes (for reproducibility)

- **Pitch:** the harmonic-product fundamental estimator (geometric mean of the magnitudes at f, 2f, 3f, weighted by requiring energy at f itself) was used for the reported numbers, because librosa pyin/yin and CREPE both showed octave errors on this fast, speech-like chant (pyin reported subharmonics at MIDI 34–45; CREPE locked onto the 2nd harmonic at 440 Hz in quiet tails). The harmonic-product estimates were spot-checked against raw FFT spectra at multiple points (e.g. t=0.9 s: no energy at 165 Hz, dominant at 330 Hz — the voice really is at 330 Hz there; t=4.3 s: dominant at 110 Hz).
- **Alignment:** faster-whisper (small/base, int8, CPU) gives the correct word *sequence* but unreliable *timestamps* on chant (it hallucinates the opening words and its timings shift by seconds). Verse locations were therefore cross-checked against the RMS pause structure and the known text order, and each rare-motif measurement was verified by confirming the surrounding words matched the target verse.
- **Tonic:** degrees in §1 are relative to D3 (MIDI 50) by convention (§0b). If the true tonic is A2, add 5 semitones to every degree offset (i.e. the reciting band becomes degrees 5–6 and the sof-pasuk "correction" disappears).
- **Tools installed for this validation:** CREPE 0.0.16 (+ TensorFlow 2.21), faster-whisper 1.2.1, librosa, aubio, matplotlib — all in `~/workspace/.venv`. Note: `pip install crepe` needs `--no-build-isolation` + `wheel` + `setuptools==69.5.1` (the build env's newer setuptools lacks `pkg_resources`), and `TMPDIR` must point off the 512 MB `/tmp` tmpfs (the CREPE model files exceed it).
