Sound toolkit #

The first Aesel Sound lane composes short instrumental phrases and loops using Pop's own sinepower voice, wobble/bitcrush effects, and Bjorklund rhythms. No samples, account, network, external executable, or personal voice assets are required. Source modules in pop/ are verbatim copies; provenance.json records the exact original paths and SHA-256 hashes. The Aesel repository license applies.

score.json is the editable recipe:

{
  "bpm": 100,
  "beats": 8,
  "instrument": "sinepower",
  "preset": "stab",
  "loop": false,
  "notes": [
    { "at": 0, "midi": 60, "duration": 1, "gain": 0.3 },
    { "at": 2, "midi": 67, "duration": 1, "gain": 0.3 }
  ],
  "effects": [{ "type": "wobble", "rate": 2, "depth": 0.4 }]
}

Times are beats (quarter notes). Presets are lead, pad, and stab.

Keep note times on beats or simple subdivisions and loops a whole number of beats. The retired Electron sound deck is no longer included; native Sound support is tracked in apple/aesel/ROADMAP.md. set_score validates and renders a full recipe. render reads the saved recipe. rhythm replaces notes using { "hits": 3, "pulses": 8, "midi": 60 } and enables looping. analyze measures the saved WAV without changing its audio.

Outputs are 24 kHz mono 16-bit PCM sound.wav, analysis.json (duration, actual PCM RMS/peak, 256-bin min/max waveform), and render provenance recording recipe and WAV hashes plus gain attenuation. Versions must preserve all four outputs. The shared project layer owns revision history and rollback.

Rendering is deterministic. Clips are capped at 30 seconds, 128 notes, four effects, and 180 seconds of accumulated voice windows. Peak attenuation limits the mix to 0.95; quiet work is not automatically amplified. Non-loop phrases get 10 ms edge fades. Loop release tails wrap into the start; the loop length is exactly the declared beat duration. Modulation rates may still cause a loop seam. Playback must remain user-controlled. No auto-play occurs in this adapter.

This slice has no sample import, recording, multitrack editing, loudness mastering, or hosted sound generation. RMS/peak are sample measurements, not LUFS/true peak.

To refresh bundled source, copy each provenance.json source to its recorded file, regenerate its SHA-256, review upstream changes, then run:

node --test easel/test/sound.test.mjs