Composition Improvement Plan

29 Jul 2026

Composition Improvement Plan

Working document for the next song composition. Based on: Composition Analysis Report + verification against `guide-track-v3-suno-instrumental.json` Key: A minor, 70 BPM, dark folk / neofolk

---

1. Pressure-Testing the Report

Section 1: Harmonic Analysis

MAJOR ERROR - v3 JSON harmonic rhythm is completely inverted.

The report (§1.3) claims:

The actual JSON data shows the exact opposite:

Impact on the diagnosis: The report's central criticism - "the chorus moves faster than the verse, which is inverted" - is wrong. The real problem is different: the verse moves harmonically (Am→G→Am) and the chorus doesn't move at all (pure Am pedal). The chorus IS on the tonic (contrary to the report's claim it "never touches the tonic"), but it provides no harmonic contrast with the verse because the verse also spends most of its time on Am. The issue is not inverted harmonic rhythm - it's insufficient harmonic differentiation between sections. Both verse and chorus are anchored to Am; the verse briefly detours to G, and the chorus never leaves Am at all.

The report's recommendation to "halve the chorus harmonic rhythm" is moot - you can't halve zero. The real recommendation should be: add harmonic movement to the chorus (e.g., Am → Fmaj7, 2 bars each) so the chorus has its own harmonic identity that contrasts with the verse's Am→G→Am pattern.

What the report got RIGHT in §1:

What the report MISSED:

Section 2: Arrangement and Orchestration

ERROR - fingerpicking density understated.

The report (§4.2, §6.1) claims verse fingerpicking density is "0.51-0.57 notes/beat." The actual JSON data:

The density is 53% higher than the report states. This makes the problem significantly worse: the fingerpicking is nearly twice as dense as the report claims, and it's the same density in verse and chorus. The recommendation to cut density by 50% is correct in direction but the starting point is worse than stated.

What the report got RIGHT:

What the report MISSED:

Section 3: Dynamic Arc and Section Contrast

All claims verified as CORRECT.

Velocity arc from JSON:

The arc is structurally correct: build → retreat → rebuild → peak. The V3 pull-back (40% velocity reduction) is real. The final chorus is the loudest section by every measure. The report's assessment here is accurate.

What the report MISSED:

Section 4: Melodic Content in the Backing

Note listing is accurate. The report's transcription of the Verse 1 fingerpicking pattern matches the JSON exactly (A3 at beat 0, E3 at beat 1, A3 at beat 2, C4 at beat 3, E4 at beat 4, etc.).

Assessment is correct:

What the report MISSED:

Section 5: Bass Line Quality

Assessment is accurate. The verse walking bass is correctly identified as functional but melodically uninteresting. The bridge bass is correctly identified as the best in the arrangement. The chorus pedal is correctly identified as needing harmonic context.

Minor error: Report says Verse 1 bass velocities are "30-46." Actual range is 36-44. The range is narrower and lower than stated, but the assessment (functional, not melodic) still holds.

Section 6: Guitar Writing

Assessment is correct and well-supported. The fingerpicked arpeggios are pattern-filling. The strummed guitar adds density, not new frequency content. The recommendations for better guitar writing (reduce density, create repeating melodic fragments, add non-chord tones, vary by section) are all sound.

Additional observation from JSON: The guitar fingerpicking in choruses uses the exact same pattern as verses (A3→E3→A3→C4→E4→C4→A3), just at higher velocities. The chorus guitar adds nothing new melodically - it's the same arpeggio, louder. This is a specific manifestation of the "chorus = verse but louder" problem.

Section 7: Structural Integrity

Assessment is mostly correct. Song form, bridge placement, V3 pull-back, and outro proportioning are all accurately described.

Error in section lengths table: The table is correct (verified against JSON: Intro 16 beats, V1 24 beats, C1 16 beats, etc.). The total is 164 beats = 41 bars at 4 beats/bar. Confirmed.

What the report MISSED:

Section 8: Genre Authenticity

Assessment is correct. The five points (no space, no grit, no high-frequency content, too uniform, no vocal-led arrangement) are all accurate and supported by the JSON data. The reference artist comparisons are appropriate.

Additional observation: The report doesn't mention that the MIDI instrument choices themselves work against genre authenticity. `acoustic_guitar_nylon` for the strummed guitar produces a classical/soft sound - dark folk typically uses steel-string acoustics with a harder attack. The `halo_pad` is too ethereal for dark folk; a darker pad or drone would be more appropriate.

Section 9: Recurring Composition Failures

Eight failures identified. Accuracy assessment: 1. Chorus harmonic rhythm = verse harmonic rhythm - Partially wrong for v3 JSON. The v3 JSON chorus has zero changes; the verse changes every 2 bars. The problem is not that they're the same, but that the chorus is too static (zero movement) while the verse has modest movement. The broader point about insufficient harmonic differentiation is correct. 2. Chord vocabulary pollution - Correct for A/B analyses. The v3 JSON already narrows to ~5 chords, which the report acknowledges. 3. Melody contracts in chorus - Cannot verify from JSON (vocal melody not encoded). From A/B analyses, likely correct. 4. No high-frequency content - Correct, confirmed. No note above E5 (659 Hz) in the JSON. 5. Hook has no fixed apex - Cannot verify from JSON (vocal melody not encoded). From A/B analyses, correct. 6. Vocal buried under band - Correct by implication. The arrangement fills 55-660 Hz continuously; no gap for vocal. 7. Weight and brightness inversely coupled - Correct by implication. Every louder section adds density in the same low-mid band. 8. No bridge dropout (Rinse Light A) - Correct. Pass of the Shuttle v3 JSON fixes this.

Section 10: What Dakota Is Getting Right

All claims verified as correct:

What the report MISSED:

Section 11: Specific Rewrite Recommendations

24 recommendations made. Assessment of key ones:

1. "Halve the chorus harmonic rhythm" - Moot for v3 JSON. The chorus has zero changes. The real recommendation: add 1-2 chord changes to the chorus (e.g., Am → Fmaj7, 2 bars each). 2. "Cut chord vocabulary to 7" - Good, already partly done in v3 JSON. The v3 uses ~5 chords. Keep this discipline. 3. "Add high-frequency element" - Essential, confirmed. No content above 659 Hz fundamental anywhere. 4. "Write a hook with fixed apex" - Correct but applies to vocal, not backing. Note for future composition. 5. "Reduce verse fingerpicking density by 50%" - Correct, but starting point is 0.875, not 0.51. Target should be 0.44 or below (3-4 notes per 8 beats, as stated). 6. "Write the verse small" - Good advice. Verse 1 bass at velocity 36-44 is already moderate. Could go lower (30-38). 7. "Add a rest before each chorus entry" - Good advice. No rests encoded before choruses in JSON. 8. "Voice strummed guitar higher" - Essential. Currently A3-E5 (220-660 Hz); should be A4-E6 (440-1320 Hz) or higher. 9. "Leave frequency gap for vocal" - Essential. Current arrangement fills 55-660 Hz continuously. 10. "Push Verse 3 quieter" - Good. Current Verse 3 bass at 24-30 could go to 18-24.

Section 12: Composition Learning Path

Six exercises proposed. Assessment:

Missing exercise: The report doesn't include an exercise for chorus harmonic identity - writing a chorus that has its own chord progression distinct from the verse. This is the actual failure the v3 JSON reveals.

---

2. Top 5 Compositional Failures

Failure #1: Chorus has no harmonic identity

What's wrong: Every chorus (C1, C2, Final C) is a pure Am tonic pedal. Zero chord changes. The bass plays only A1, the pad holds only A2, the strummed guitar voices only Am triads (A3/C4/E4/A4). The chorus is harmonically identical to the tonic portions of the verse - just louder and with more layers.

Evidence:

Why it hurts: The chorus cannot register as an arrival because it's harmonically the same as the verse's home base. Loudness and layer count are the only differentiation. Without harmonic contrast, the chorus feels like a louder verse, not a destination.

What the fix looks like: Give the chorus its own 2-chord progression: Am for 2 bars → Fmaj7 for 2 bars (i → VI), held at one chord per 2 bars. This gives the chorus a harmonic identity that contrasts with the verse's Am→G→Am. The Fmaj7 provides the submediant colour that dark folk uses (reference: Chelsea Wolfe's modal vamps, Low's drone-based choruses). Keep the bass as a pedal on the root of whichever chord is current (A1 for Am, F1 for Fmaj7).

What the fix does NOT look like: Do not make the chorus change chords every bar (change-rate 1.0). That was the failure of both A versions. Do not use a progression that's the same as the verse. Do not use dim chords or secondary dominants - the chorus should use diatonic chords only (Am, Fmaj7, Cmaj7, Dm).

Failure #2: Guitar fingerpicking is chord-outline arpeggio, not melody

What's wrong: The fingerpicked guitar cycles through chord tones in a predictable up-down pattern (A3→E3→A3→C4→E4→C4→A3) at a density of 0.875 notes/beat - nearly one note per beat. The pattern is identical in verse and chorus (same notes, same rhythm, same order), just at different velocities. The guitar provides harmonic information but zero melodic direction.

Evidence:

Why it hurts: The guitar competes with the vocal for space (0.875 notes/beat is close to vocal density), provides no melodic interest of its own, and makes the arrangement sound like a MIDI demo rather than a composed piece. The lack of variation between verse and chorus means the guitar can't contribute to sectional contrast.

What the fix looks like: Write a 2-bar repeating melodic fragment (4-5 notes) with a recognisable shape. Example over Am: A3 (3 beats) → C4 (2 beats) → E4 (3 beats) → [rest, 2 beats]. Transpose this fragment to follow the harmony. Cut density to 3-4 notes per 8 beats (0.375-0.5 notes/beat). Add one non-chord tone per 4 beats (B3 as passing tone between A3 and C4, G3 as neighbour tone to A3). Vary the fragment by section: Verse 1 = sparse and low, Verse 2 = add a high extension, Verse 3 = harmonics or single sustained notes.

What the fix does NOT look like: Do not replace arpeggios with a completely different texture (e.g., strumming). The fingerpicking concept is correct for dark folk. Do not add more notes to make it "more interesting" - fewer notes with more intention is the goal. Do not use random non-chord tones - each one should serve a melodic function (passing, neighbour, anticipation).

Failure #3: No content above 660 Hz - the arrangement lives in one frequency band

What's wrong: Every note in the arrangement falls between A1 (55 Hz) and E5 (659 Hz). No fundamental frequency above 660 Hz is encoded anywhere. The brightness parameters are all low (pad 28, GFP 54, strum 48 out of 127). Every "bigger" section adds more content in the same 55-660 Hz band, which increases density without increasing perceived size.

Evidence:

Why it hurts: When the chorus adds the strummed guitar, it adds content in 220-660 Hz - the same band the fingerpicking already fills. The ear perceives this as "more of the same sound," not "a bigger sound." A real chorus lift requires opening a new frequency band. This is why every "bigger" section also sounds duller - spectral centroid drops as density increases because all the new content is in the low-mids.

What the fix looks like: Add a dedicated high-frequency element that enters at Chorus 2 and stays through the final chorus. Options: (a) a sustained high guitar harmonic at A5 (880 Hz) or E6 (1320 Hz), (b) a tambourine or shaker sample (2-5 kHz), (c) revoice the strummed guitar one octave higher (A4/C5/E5/A5, 440-880 Hz). Additionally, revoice the pad to A4 (440 Hz) instead of A2 (110 Hz) - this moves it out of the bass's octave and into a register nothing else occupies, making it audible for the first time.

What the fix does NOT look like: Do not add a cymbal or hi-hat - that's a mixing decision, not a compositional one, and it doesn't fit the dark folk aesthetic. Do not just turn up the brightness parameter on existing instruments - the problem is the absence of content in a frequency band, not the timbre of existing content. Do not add a second guitar layer in the same octave - that compounds the problem.

Failure #4: No rests, no gaps, no breathing room

What's wrong: Every section except the bridge caesura has continuous content on every beat. The bass plays every 2 beats in verses, every 4 beats in choruses. The guitar fires every 1-1.5 beats for the entire verse. The pad sustains continuously. There is not a single beat of silence in the entire 164-beat arrangement outside the bridge caesura. No rest before any chorus entry. No gap between verse phrases.

Evidence:

Why it hurts: Dark folk is defined by space as much as by content. Low, Sun Kil Moon, and Mark Lanegan all use dramatic gaps - 4-beat rests between phrases, sections with only one instrument, moments of near-silence. Without space, the arrangement sounds mechanistic and synthetic. Without a rest before the chorus, the chorus can't feel like an arrival - it just continues the stream.

What the fix looks like: Add a 1.5-beat rest before each chorus entry (all instruments silent). Add a 2-beat rest at the midpoint of each verse (after the first 4-bar phrase). Drop the bass entirely for the first 8 beats of Verse 3. End the outro on a clean stop (hold final Am for 4 beats, then cut to silence) instead of a fade.

What the fix does NOT look like: Do not add rests randomly - place them at phrase boundaries (end of 4-bar phrases) and before section entries. Do not use rests as a substitute for dynamic contrast - they complement velocity changes, they don't replace them. Do not make the rests too long (4+ beats) outside the bridge - that becomes avant-garde, not dark folk.

Failure #5: Verse and chorus use identical guitar patterns

What's wrong: The guitar fingerpicking pattern in the chorus is the same pattern as the verse - same notes (A3, E3, C4, E4), same order, same rhythm, same density (0.875 notes/beat). The only differences are velocity (louder) and the addition of A4/C5 in Chorus 2 and Final Chorus. The guitar contributes nothing to sectional contrast.

Evidence:

Why it hurts: If the guitar, bass, and pad all do the same thing in verse and chorus, the only differentiation is loudness. This is the arrangement equivalent of a song that's all one section. The listener's ear needs textural or pattern change to perceive a new section, not just a volume change.

What the fix looks like: Give the chorus a different guitar treatment: (a) switch from fingerpicking to a sustained held chord (one strum per 2 bars, let it ring), (b) move the fingerpicking to a different register (A4-C5 instead of A3-C4), (c) change the pattern - instead of ascending arpeggios, use a descending line or a repeated 2-note figure. The chorus guitar must sound different from the verse guitar in pattern, register, or texture.

What the fix does NOT look like: Do not just make the chorus guitar louder - that's the current failure. Do not add more notes to the chorus guitar - that increases density without creating contrast. Do not change the instrument - the acoustic guitar is correct for dark folk; change what it plays, not what it is.

---

3. Next Arrangement Spec

Song parameters: A minor, 70 BPM, 4/4, dark folk / neofolk Form: Intro (2 bars) → V1 (4 bars) → C1 (6 bars) → V2 (4 bars) → C2 (6 bars) → Bridge (4 bars) → V3 (4 bars) → Final C (6 bars) → Outro (2 bars) = 38 bars Chord vocabulary: Am, Am7, Fmaj7, Cmaj7, Dm, E (bridge only). 6 chords total. Verse progression: Am (1 bar) → Fmaj7 (1 bar) → Cmaj7 (1 bar) → Am (1 bar). Change-rate = 1.0/bar. Chorus progression: Am (2 bars) → Fmaj7 (2 bars) → Cmaj7 (2 bars). Change-rate = 0.5/bar. Bridge progression: Dm (2 bars) → E (2 bars). Change-rate = 0.5/bar.

Intro - 2 bars (8 beats)

Rest before entry: none (song start)

Harmony: Am

| Layer | Notes | Start | Duration | Velocity | Notes | |-------|-------|-------|----------|----------|-------| | pad | A4 | 0 | 7.5 | 24 | Sustained drone, high register (440 Hz) - NOT A2 | | guitar_fingerpick | A3 | 0 | 3.0 | 28 | Single note, let ring | | guitar_fingerpick | E4 | 4 | 3.0 | 26 | Single note, let ring | | bass | - | - | - | - | No bass in intro | | guitar_strum | - | - | - | - | No strum in intro | | high_element | - | - | - | - | No high element in intro |

Dynamic target: Quietest section. Establish the tonic and the mood. Two notes of guitar over a high pad drone. Empty, atmospheric.

New element - `high_element`: Instrument: `guitar_harmonics` (or `acoustic_guitar_steel` with brightness 70). Role: high-frequency content. Volume: 40. Pan: 50. Reverb: 20. This instrument will carry content above 880 Hz starting from Chorus 2.

Verse 1 - 4 bars (16 beats)

Rest before entry: 0 (guitar continues from intro)

Harmony: Am → Fmaj7 → Cmaj7 → Am (1 bar each)

| Layer | Notes | Start | Duration | Velocity | Pattern | |-------|-------|-------|----------|----------|---------| | bass | A1 | 0 | 3.0 | 38 | Root, held 3 beats | | bass | F1 | 4 | 3.0 | 36 | Root of Fmaj7 | | bass | C2 | 8 | 3.0 | 36 | Root of Cmaj7 | | bass | A1 | 12 | 3.0 | 38 | Return to tonic | | guitar_fingerpick | A3 | 0 | 2.5 | 34 | Melodic fragment start | | guitar_fingerpick | C4 | 3 | 2.0 | 32 | | | guitar_fingerpick | E4 | 5 | 2.5 | 36 | | | guitar_fingerpick | [rest] | 8 | 2.0 | - | 2-beat gap | | guitar_fingerpick | C4 | 10 | 2.0 | 32 | Fragment over Fmaj7 (transposed) | | guitar_fingerpick | E4 | 12 | 2.0 | 34 | | | guitar_fingerpick | G4 | 14 | 2.0 | 36 | Non-chord tone (9th of Fmaj7, adds colour) | | pad | A4 | 0 | 3.5 | 22 | Follows chord roots, high register | | pad | F4 | 4 | 3.5 | 20 | | | pad | C5 | 8 | 3.5 | 20 | | | pad | A4 | 12 | 3.5 | 22 | | | guitar_strum | - | - | - | - | No strum in verse | | high_element | - | - | - | - | No high element in V1 |

Dynamic target: Quiet. Bass at 36-38, guitar at 32-36, pad at 20-22. This is the floor - the chorus needs to feel louder than this.

Density: Guitar FP = 6 notes / 16 beats = 0.375 notes/beat (down from 0.875). Bass = 4 notes / 16 beats = 0.25. Total instrumental density = 0.625 notes/beat (down from 1.625).

Key change from v3: Guitar plays a melodic fragment (A3→C4→E4) over Am, then transposes it (C4→E4→G4) over Fmaj7. The fragment has shape and direction. A 2-beat rest at beat 8 creates breathing room. Bass holds roots for 3 beats instead of walking every 2 beats - simpler, more sustained, more dark folk.

Chorus 1 - 6 bars (24 beats)

Rest before entry: 1.5 beats of silence (all instruments off at beats 15.5-16 of verse)

Harmony: Am (2 bars) → Fmaj7 (2 bars) → Cmaj7 (2 bars)

| Layer | Notes | Start | Duration | Velocity | Pattern | |-------|-------|-------|----------|----------|---------| | bass | A1 | 0 | 7.0 | 48 | Root, sustained 7 beats | | bass | F1 | 8 | 7.0 | 46 | Root of Fmaj7 | | bass | C2 | 16 | 7.0 | 46 | Root of Cmaj7 | | pad | A4 | 0 | 7.5 | 28 | High register, follows chord | | pad | F4 | 8 | 7.5 | 26 | | | pad | C5 | 16 | 7.5 | 26 | | | guitar_fingerpick | A3 | 0 | 3.0 | 40 | Sustained notes, not arpeggios | | guitar_fingerpick | E4 | 4 | 3.0 | 42 | | | guitar_fingerpick | F3 | 8 | 3.0 | 40 | Over Fmaj7 | | guitar_fingerpick | C4 | 12 | 3.0 | 42 | | | guitar_fingerpick | C4 | 16 | 3.0 | 40 | Over Cmaj7 | | guitar_fingerpick | G4 | 20 | 3.0 | 44 | | | guitar_strum | [A4, C5, E5, A5] | 0 | 5.0 | 40 | High voicing (440-880 Hz) | | guitar_strum | [A4, C5, E5, A5] | 8 | 5.0 | 40 | Same chord - Am pedal above | | guitar_strum | [F4, A4, C5, F5] | 16 | 5.0 | 40 | Fmaj7 voicing (349-698 Hz) | | high_element | - | - | - | - | No high element until C2 |

Dynamic target: Moderate lift from verse. Bass at 46-48 (vs. 36-38 in verse). Guitar at 40-44 (vs. 32-36). Strum enters at 40. The lift comes from: (1) new layer (strum), (2) higher voicings, (3) longer sustain, (4) harmonic movement (Am→Fmaj7→Cmaj7 vs. verse's faster Am→Fmaj7→Cmaj7→Am).

Key changes from v3:

Verse 2 - 4 bars (16 beats)

Rest before entry: 0 (bass continues from C1 final chord)

Harmony: Am → Fmaj7 → Cmaj7 → Am (same as V1)

| Layer | Notes | Start | Duration | Velocity | Change from V1 | |-------|-------|-------|----------|----------|----------------| | bass | A1 | 0 | 3.0 | 40 | Slightly louder (+2) | | bass | C2 | 2 | 1.5 | 36 | Passing tone added | | bass | F1 | 4 | 3.0 | 38 | | | bass | A1 | 6 | 1.5 | 34 | Passing tone | | bass | C2 | 8 | 3.0 | 38 | | | bass | E2 | 10 | 1.5 | 34 | Passing tone | | bass | A1 | 12 | 3.0 | 40 | | | guitar_fingerpick | A3 | 0 | 2.5 | 36 | Same fragment, +2 velocity | | guitar_fingerpick | C4 | 3 | 2.0 | 34 | | | guitar_fingerpick | E4 | 5 | 2.5 | 38 | | | guitar_fingerpick | A4 | 7 | 2.0 | 40 | NEW: high extension (440 Hz) | | guitar_fingerpick | C4 | 10 | 2.0 | 34 | Transposed fragment | | guitar_fingerpick | E4 | 12 | 2.0 | 36 | | | guitar_fingerpick | G4 | 14 | 2.0 | 38 | | | pad | A4 | 0 | 3.5 | 24 | Same as V1, +2 | | pad | F4 | 4 | 3.5 | 22 | | | pad | C5 | 8 | 3.5 | 22 | | | pad | A4 | 12 | 3.5 | 24 | | | guitar_strum | - | - | - | - | No strum | | high_element | - | - | - | - | No high element |

Dynamic target: Slightly louder than V1 (bass +2, guitar +2). The progression is the same, but the bass adds passing tones (C2, A1, E2) and the guitar adds A4 as a high extension. This is the "escalation" verse - same structure, brighter, slightly more urgent.

Chorus 2 - 6 bars (24 beats)

Rest before entry: 1.5 beats of silence

Harmony: Am (2 bars) → Fmaj7 (2 bars) → Cmaj7 (2 bars) - same as C1

| Layer | Notes | Start | Duration | Velocity | Change from C1 | |-------|-------|-------|----------|----------|----------------| | bass | A1 | 0 | 7.0 | 50 | Louder (+2) | | bass | F1 | 8 | 7.0 | 48 | | | bass | C2 | 16 | 7.0 | 48 | | | pad | A4 | 0 | 7.5 | 30 | Louder (+2) | | pad | F4 | 8 | 7.5 | 28 | | | pad | C5 | 16 | 7.5 | 28 | | | guitar_fingerpick | A3 | 0 | 3.0 | 42 | Same pattern, +2 | | guitar_fingerpick | E4 | 4 | 3.0 | 44 | | | guitar_fingerpick | F3 | 8 | 3.0 | 42 | | | guitar_fingerpick | C4 | 12 | 3.0 | 44 | | | guitar_fingerpick | C4 | 16 | 3.0 | 42 | | | guitar_fingerpick | G4 | 20 | 3.0 | 46 | | | guitar_strum | [A4, C5, E5, A5] | 0 | 5.0 | 42 | Louder (+2) | | guitar_strum | [A4, C5, E5, A5] | 8 | 5.0 | 42 | | | guitar_strum | [F4, A4, C5, F5] | 16 | 5.0 | 42 | | | high_element | A5 | 0 | 7.5 | 34 | NEW: enters here. Sustained A5 (880 Hz) | | high_element | A5 | 8 | 7.5 | 34 | Continues over Fmaj7 | | high_element | C6 | 16 | 7.5 | 36 | Moves to C6 (1047 Hz) over Cmaj7 |

Dynamic target: Louder than C1. The high_element enters here for the first time, carrying content at 880-1047 Hz. This is the moment the arrangement "opens up" - the listener hears a frequency band that hasn't existed before.

Key change: The `high_element` (guitar harmonics) enters at Chorus 2 and provides the 880+ Hz content that no previous arrangement has had. It sustains one note per chord, following the harmony.

Bridge - 4 bars (16 beats)

Rest before entry: 0 (direct from C2)

Harmony: Dm (2 bars) → E (2 bars)

| Layer | Notes | Start | Duration | Velocity | Notes | |-------|-------|-------|----------|----------|-------| | bass | D1 | 0 | 7.0 | 30 | Same as v3, keep | | bass | E1 | 8 | 7.0 | 32 | Same as v3, keep | | pad | D4 | 0 | 7.5 | 20 | Moved up from D2 to D4 (293 Hz) | | pad | E4 | 8 | 7.5 | 22 | Moved up from E2 to E4 (330 Hz) | | guitar_fingerpick | D3 | 0 | 2.0 | 24 | Same as v3 | | guitar_fingerpick | A3 | 2 | 1.5 | 22 | | | guitar_fingerpick | D4 | 4 | 2.0 | 26 | | | guitar_fingerpick | F4 | 6 | 1.5 | 24 | | | guitar_fingerpick | E3 | 8 | 2.0 | 26 | | | guitar_fingerpick | B3 | 10 | 1.5 | 24 | | | guitar_fingerpick | E4 | 12 | 2.0 | 28 | | | guitar_strum | [D3, A3, D4] | 0 | 3.5 | 22 | Same as v3, keep | | guitar_strum | [E3, B3, E4] | 8 | 3.5 | 24 | Same as v3, keep | | high_element | - | - | - | - | Drops out for bridge | | caesura | - | 12 | 4 | - | Full stop, all instruments silent, 4 beats |

Dynamic target: Sharp drop from C2. Bass from 50 to 30 (40% reduction). All layers drop. The caesura at beat 12 provides 4 beats of total silence. This is the tension moment.

Key change from v3: Pad moved from D2/E2 (73-82 Hz) to D4/E4 (293-330 Hz). Gets it out of the bass register and into a range where it's actually audible and contributes colour. The bridge guitar and bass remain as v3 - they were correct.

Verse 3 - 4 bars (16 beats)

Rest before entry: 0 (direct from bridge caesura - silence flows into V3)

Harmony: Am → Fmaj7 → Cmaj7 → Am (same as V1/V2, but see note)

| Layer | Notes | Start | Duration | Velocity | Change from V1 | |-------|-------|-------|----------|----------|----------------| | bass | - | 0 | 8.0 | - | NO BASS for first 8 beats | | bass | A1 | 8 | 3.0 | 24 | Bass enters at beat 8 | | bass | F1 | 12 | 3.0 | 22 | | | guitar_fingerpick | A3 | 0 | 4.0 | 22 | Very quiet, long durations | | guitar_fingerpick | E4 | 4 | 4.0 | 24 | | | guitar_fingerpick | F3 | 8 | 4.0 | 22 | Bass enters here | | guitar_fingerpick | C4 | 12 | 4.0 | 24 | | | pad | A4 | 0 | 7.5 | 16 | Barely audible | | pad | F4 | 8 | 7.5 | 14 | | | guitar_strum | - | - | - | - | No strum | | high_element | - | - | - | - | No high element |

Dynamic target: Near-silence. Bass absent for first half. Guitar at 22-24 (near-zero). Pad at 14-16 (barely a whisper). This is the most extreme pull-back - the quietest moment in the song. The silence after the bridge caesura flows directly into this section, so the listener experiences 4 beats of silence (bridge) + 8 beats of near-silence (V3 first half) = 12 beats of almost nothing before the bass enters.

Key change from v3: Bass drops out entirely for the first 8 beats (v3 kept it at velocity 24-30). Guitar uses 4-beat sustained notes (v3 used 1.3-1.5 beat notes). Only 4 guitar notes in the entire verse (vs. 12 in v3). This is genuine sparseness, not just "quieter version of verse."

Final Chorus - 6 bars (24 beats)

Rest before entry: 1.5 beats of silence

Harmony: Am (2 bars) → Fmaj7 (2 bars) → Cmaj7 (2 bars) - same as C1/C2

| Layer | Notes | Start | Duration | Velocity | Change from C2 | |-------|-------|-------|----------|----------|----------------| | bass | A1 | 0 | 7.0 | 56 | Loudest bass in song | | bass | F1 | 8 | 7.0 | 54 | | | bass | C2 | 16 | 7.0 | 54 | | | pad | A4 | 0 | 7.5 | 32 | Loudest pad | | pad | F4 | 8 | 7.5 | 30 | | | pad | C5 | 16 | 7.5 | 30 | | | guitar_fingerpick | A3 | 0 | 3.0 | 48 | Loudest GFP | | guitar_fingerpick | E4 | 4 | 3.0 | 50 | | | guitar_fingerpick | F3 | 8 | 3.0 | 48 | | | guitar_fingerpick | C4 | 12 | 3.0 | 50 | | | guitar_fingerpick | C4 | 16 | 3.0 | 48 | | | guitar_fingerpick | G4 | 20 | 3.0 | 52 | | | guitar_strum | [A4, C5, E5, A5, C6] | 0 | 5.0 | 48 | Widest voicing (440-1047 Hz) | | guitar_strum | [A4, C5, E5, A5, C6] | 8 | 5.0 | 48 | | | guitar_strum | [F4, A4, C5, F5, A5] | 16 | 5.0 | 48 | Fmaj7 voicing to F5-A5 (698-880 Hz) | | high_element | A5 | 0 | 7.5 | 42 | Loudest high element | | high_element | A5 | 8 | 7.5 | 42 | | | high_element | E6 | 16 | 7.5 | 44 | Highest note in song (1320 Hz) |

Dynamic target: Peak. Every layer at its loudest. The high_element reaches E6 (1320 Hz) - the highest note in any arrangement so far. The strummed guitar adds C6 (1047 Hz). The frequency spectrum now spans 55 Hz (bass A1) to 1320 Hz (high element E6), with content in every octave. This is what "opening up" sounds like.

Key change from v3: Final chorus is 6 bars (not 4). High element reaches E6 (not just A5). Strummed guitar voicings include C6 (not just E5). The harmonic progression is Am→Fmaj7→Cmaj7 (not pure Am pedal). Every layer is at peak velocity. The arrival is encoded in level AND brightness AND harmonic movement - three dimensions of "bigger," not just one.

Outro - 2 bars (8 beats)

Rest before entry: 0

Harmony: Am

| Layer | Notes | Start | Duration | Velocity | Notes | |-------|-------|-------|----------|----------|-------| | bass | A1 | 0 | 4.0 | 40 | Sustained | | bass | A1 | 4 | 4.0 | 30 | Decrescendo | | guitar_fingerpick | A3 | 0 | 3.0 | 36 | | | guitar_fingerpick | E4 | 4 | 3.0 | 28 | | | pad | A4 | 0 | 7.5 | 28 | | | guitar_strum | [A4, C5, E5] | 0 | 3.0 | 36 | One strum, let ring | | high_element | A5 | 0 | 4.0 | 34 | Sustained, then cut |

Ending: At beat 6 (2 beats before the end), all instruments cut to silence. Clean stop. No fade. The final sound is a guitar harmonic (A5) ringing into silence for 2 beats, then nothing. Dark folk endings are abrupt - let the silence do the work.

Key change from v3: Outro is 2 bars (not 3). Ends on a clean stop, not a fade. The high_element rings out last, leaving the impression of brightness - the opposite of v3, where the fade left the impression of diminishing low-end rumble.

---

4. Practice Exercise

Exercise: Chorus Harmonic Identity Etude

What to write: Compose two 8-bar sections in A minor at 70 BPM for solo acoustic guitar (fingerpicked, no other instruments):

Play Section A, then Section B, then Section A again. Listen without looking at the notation.

Pass criterion: Section B must sound like an arrival - a moment where the music settles and opens up - without any change in volume, instrument, or velocity. The only difference between A and B is the harmonic rhythm (1 chord/bar vs. 1 chord/2 bars) and the guitar pattern (arpeggiated vs. sustained). If you can hear the arrival, you've learned to use harmonic rhythm as a structural tool. If you can't, keep refining the melodic fragments until the contrast is audible.

How long it should take: 1-2 composition sessions (30-60 minutes each).

What a failed attempt looks like:

---

5. What to Keep - The 3 Things Dakota Is Getting Right

1. The dynamic arc (build → retreat → rebuild → peak)

Where it appears: The velocity structure across all sections. Verse 1 bass at 36-44 builds to Chorus 1 at 44-48, drops to Verse 2 at similar level, builds to Chorus 2 at 46-50, drops sharply at the bridge (26-32), stays low in Verse 3 (22-30), peaks at Final Chorus (52-56). This is a textbook dark folk dynamic arc.

Why it works: The pull-back before the final peak is the single most powerful structural tool in slow tempo music. At 70 BPM, the listener needs a reason to keep listening for 4+ minutes. The arc creates tension and release at the sectional level - the bridge dropout and Verse 3 near-silence make the final chorus feel like an arrival, not just another section. The A/B analysis confirms: V3→Final Chorus transition is +5.47 dB, the best in either file.

Do not change: Keep the arc structure. Keep the 40% velocity reduction at the bridge. Keep Verse 3 as the quietest section. Keep the final chorus as the loudest. The arc is the skeleton - everything else is flesh.

2. The bridge (Dm → E major with caesura)

Where it appears: Bridge section, start_beat 96, 16 beats. Bass: D1 (8 beats) → E1 (8 beats). Guitar: Dm arpeggios (D3/A3/D4/F4) → E arpeggios (E3/B3/E4). Strum: Dm triad → E triad. Caesura at beat 12: 4 beats of total silence.

Why it works: The Dm → E major progression is the only genuine harmonic tension in the arrangement. Dm (iv) moves to E (V) - a secondary dominant that creates expectation of resolution back to Am. The G♯3 in the melody (confirmed in the guide-track diagnostic) is the only raised 7th in the entire piece. The velocity drop (bass from 50 to 30, 40% reduction) creates textural contrast. The caesura provides absolute silence - the most dramatic moment in the song. This is the model for what a dark folk bridge should be.

Do not change: Keep the Dm → E progression. Keep the caesura. Keep the velocity drop. Keep the bass at D1/E1 (lowest register = darkest moment). The bridge is the one section that needs no improvement.

3. The tempo map with subtle deceleration

Where it appears: The tempo_map field in the JSON: 70 → 69.8 → 70 → 69.5 → 70 → 69.8 → 68.5 BPM. The final chorus drops to 68.5 BPM - a 1.5 BPM deceleration from the peak.

Why it works: At 70 BPM, each beat is 0.857 seconds. At 68.5 BPM, each beat is 0.876 seconds - 19 milliseconds longer. Over a 24-beat final chorus, that's 456 ms of additional time - nearly half a second of "stretching out." The listener doesn't consciously notice the tempo change, but they feel the final chorus as slightly wider, slightly heavier, slightly more significant. This is the kind of detail that separates a sequenced arrangement from a performed one.

Do not change: Keep the deceleration toward the final chorus. Keep the micro-fluctuations (69.8, 69.5) in earlier sections - they prevent the arrangement from sounding metronomic. The tempo map is a sophisticated touch that most MIDI arrangers never consider.

---

*This plan is a working document. Use it as the starting point for the next composition. The arrangement spec in §3 can be directly encoded into the JSON format. The exercise in §4 should be completed before writing the next full song - it develops the single most important missing skill.*

Back to studies