Tintinnabulator

A tintinnabulator based on Professor Milton Mermikides' work. On Youtube Professor Milton presents a wonderful introduction to Arvo Part's tintinnabuli, and I thought it'd be fun to vibe-code a tintinnabulator.

For those interested in such things: Vibe coding the basic functionality took about 10 minutes using github copilot, adding things like T-voice volume and other bits-and-bobs i thought would be nice took about 30 minutes, and it took about 1.5 hours coding the examples! Evidently, AI is not as good at music as it is at coding.

A demo

Install and run with uv

For a new project from scratch, the equivalent setup is:

	uv init
	uv add numpy pygame

In this repository, dependencies are already declared in pyproject.toml. Use:

	uv sync
	uv run tintinnabulator

Run tests and a syntax pass:

	uv run pytest
	uv run python -m compileall src tests

Features

• Clickable multi-octave piano keyboard.
• M-voice (user melody) with piano-like synthesized tone.
• T-voice generated deterministically from tonic triad and tintinnabuli position.
• Major/minor mode triads.
• Position modes:
  • None
  • 2nd Position Superior
  • 1st Position Superior
  • 1st Position Inferior
  • 2nd Position Inferior
• Duration buttons plus Free milliseconds field and slider.
• Dark UI with visual highlights for melody note, T-note, and triad tones.

Project layout

• src/tintinnabulator/music.py: note mapping, triad generation, duration logic, T-voice rules.
• src/tintinnabulator/audio_engine.py: additive synthesis and polyphonic playback.
• src/tintinnabulator/gui.py: tkinter interface and piano interaction.
• src/tintinnabulator/cli.py: command-line entry point.

Tintinnabuli algorithm (implemented)

For each clicked melody note (M-voice), triad candidates are generated from the selected key/mode across MIDI 21-108. The T-voice note is selected by rule:

• 1st Position Superior: nearest triad note >= melody.
• 2nd Position Superior: second nearest triad note >= melody.
• 1st Position Inferior: nearest triad note <= melody.
• 2nd Position Inferior: second nearest triad note <= melody.
• None: no T-voice.

When the requested superior/inferior candidate would fall outside available range, nearest valid in-range triad note is used.

Synthesis summary

• M-voice piano-like tone:

  • harmonic partial stack,
  • slight detuning,
  • short attack and shorter decay,
  • subtle transient for percussive character.

• T-voice bell-like tone:

  • inharmonic partials,
  • softer/slower attack,
  • longer decay tail,
  • mild shimmer modulation.

No external audio samples are required.