utils/music_theory

music_theory

Music theory from first principles, as typed Python you can hear. Follows Luke Haas's "Music Theory for Programmers" section by section: a sound is a number changing over time, pitch is multiplicative, the tuning everyone wants cannot exist (3ⁿ ≠ 2ᵐ), so equal temperament trades purity for a system that works in every key. Scales, chords and progressions are then plain integer arrays on top of that, and a small numpy synth turns them into WAV files.

Section-by-section notes, with what to listen for, live in DOCS-article-walkthrough.md.

Install

pip install "music_theory @ git+https://github.com/kuhyx/utils@music-theory-v0.1.0#subdirectory=music_theory"

Only dependency: numpy. Playback is left to aplay (or anything that opens a 16-bit mono WAV); the package never touches an audio device itself.

Hear the article

python -m music_theory list                       # every demo and its section
python -m music_theory explain comma              # just the numbers
python -m music_theory demo all --out /tmp/mt     # WAV per section + facts
aplay -q /tmp/mt/06_comma.wav

demo	article section	what you hear
tone	1 sound	A4, 440 Hz; the FFT peak is printed
envelope	2 envelopes	same sine with a 10 ms then a 500 ms attack
waveforms	2 timbre	sine, triangle, square, saw at 220 Hz
octaves	3 octaves	A1 to A5, each double the last
consonance	3 consonance	2:1, 3:2, 4:3, 5:4, 16:15, then √2
comma	4 the comma	A4 against A4 × 1.0136, beating 6 times a second
temperament	5 equal steps	twelve steps of 2^(1/12)
divisions	5 why twelve	pure fifth vs the 5-, 7-, 12-, 19-, 24-TET fifth
scale	6 scales	C major, then A minor
modes	6 modes	seven rotations of the same array
chords	7 qualities	major, minor, dim, aug, and three sevenths on C
diatonic	8 keys	I ii iii IV V vi vii° of C
transpose	8 transposition	I V vi IV in C, then +4 in E
resolution	9 tension	V7 then I; the tritone collapses
progressions	9 loops	pop, jazz, axis, fifties
piece	11 the piece	bass + arpeggio + seeded melody, A minor
bracket	(beyond)	the same piece clean, then degraded

Render a recipe

A composition is data. render turns a recipe JSON into a loop-length WAV and records its provenance (recipe, package version, SHA-256) in a manifest:

{"key": "A3", "mode": "major", "tempo_bpm": 112, "progression": [1, 6, 3, 7],
 "bars": 16, "seed": 20260911, "waveform": "saw", "degrade": 0.0}

python -m music_theory render --recipe city.json --out city.wav --manifest MANIFEST.json
python -m music_theory render --recipe city.json --out ruin.wav --degrade 1.0 --manifest MANIFEST.json

degrade (0..1) is the same piece, ruined: past 0.5 the mode flips to its darker relative, the tempo drags, melody notes drop out and drift off pitch, and the signal is bit-crushed, hissed and low-passed. degrade: 0 is byte- identical to the original, so "the ending is the opening, destroyed" is a one-flag transform on the same seed.

Render a score

python -m music_theory render --score examples/waltz.txt --out waltz.wav

Named voices, noise drums, meter, A/B sections. Format: score.py docstring; examples: examples/waltz.txt (3/4), examples/jig.txt (6/8).

Render sound effects

sfx renders spellcode's spell and move sounds -- sfxr-style parameter sets built from this package's own primitives, no AI audio and no sfxr port -- as 16-bit mono WAVs plus a MANIFEST.json (every parameter, the seed, the package version and each file's SHA-256):

python -m music_theory sfx --out sfx/                         # all 22, 44.1 kHz
python -m music_theory sfx --out sfx/ --sample-rate 22050     # match the music beds
python -m music_theory sfx --out sfx/ --only fire_hit --only dodge

Presets: <element>_cast, <element>_hit and <element>_impact for arcane, fire, water, lightning, earth and wind, plus dodge, parry_normal, parry_perfect and fizzle. Each is seeded from its name, so a re-render is byte-identical. A sound is a few sfx.Layers (transient, body, tail, sub) mixed and normalised; define a new one with sfx_layers.make(name, *layers) and render it with sfx.render.

Modules

module	article	contents
pitch	1, 5	MIDI ↔ Hz (440·2^((n−69)/12)), cents, note names
tuning	3, 4, 5	harmonic series, just ratios, the comma, best_divisions
scales	6	step patterns, modes as rotations, pitch_at_degree
chords	7, 9	qualities, stacked thirds, Roman numerals, V7 → I
progressions	8, 9	named loops, degree → chord, transpose
synth	1, 2	waveforms, ADSR, low-pass, mix, normalise
sequencer	10	beat grid, NoteEvent, bar-exact render
compose	11	Recipe → bass, arpeggio, melody
degrade	—	the ruin transform
analysis	—	FFT peak, spectral centroid, band energy, octave-band slope
noise	—	seeded white and pink noise, Poisson grain gates
sweep	—	pitch curves (glide, vibrato, arpeggio, jumps, retrigger), phase-accumulating oscillator, ring mod
envelopes	—	exponential decay, attack-hold-punch-decay Shape
biquad	—	resonant low/high/band-pass with a moving cutoff
sfx, sfx_layers	—	sfxr-style layered parameter set and its renderer
sfx_presets*, sfx_render	—	spellcode's 22 presets, WAVs + manifest
wav, manifest	—	16-bit mono PCM out, SHA-256 provenance

Develop

pip install -r requirements.txt
python -m pytest --cov=music_theory   # 100 % branch coverage is enforced
pre-commit run --config music_theory/.pre-commit-config.yaml --all-files   # from ~/src/utils

Consumed by ~/src/roadside-assistance (tools/render_music_beds.sh) for its three world-state music beds.