@hecto/synth

Audio graphs as plain values. Compose oscillators, filters, envelopes, modulators, and effects with pipe(), then hand the graph to a renderer.

pnpm add @hecto/synth

import { pipe } from '@hecto/fp'
import {
  effects,
  envelope,
  filter,
  instrument,
  oscillator,
  render,
  sampler,
  toWav,
} from '@hecto/synth'

const bass = pipe(
  oscillator('saw', 110),
  filter.lowpass(800, 0.7),
  envelope({ attack: 0.01, decay: 0.3, sustain: 0.4, release: 0.5 }),
  effects.distortion(0.3),
  effects.ensembleChorus({ rate: 0.4, depth: 4.44, mix: 0.35 }),
  effects.tapeDelay({ timeMs: 180, feedback: 0.42, mix: 0.22, tone: 0.72 }),
  effects.plateReverb({ decay: 0.52, damping: 0.4, mix: 0.18 }),
  effects.springReverb({ decay: 0.58, damping: 0.34, tension: 0.5, drip: 0.24, mix: 0.14 }),
  effects.nonlinearReverb({ timeMs: 140, decay: 0.66, drive: 0.22, mix: 0.12 }),
  effects.spaceEcho({ timeMs: 140, feedback: 0.55, mix: 0.28, reverbMix: 0.08 }),
)

const samples = render(bass, { duration: 1, sampleRate: 48_000 })
const wav = toWav(samples, { sampleRate: 48_000 })

const sampled = sampler.instrument({
  zones: [
    {
      samples: new Float32Array([0, 0.7, 0.2, -0.2, 0]),
      sampleRate: 48_000,
      rootMidi: 60,
      keyLow: 48,
      keyHigh: 72,
      velocityLow: 0,
      velocityHigh: 1,
      loop: true,
      loopStart: 1,
      loopEnd: 4,
    },
  ],
})

const lofi = instrument.lofiSampler({
  zones: [
    {
      samples: new Float32Array([0, 0.9, 0.4, -0.15, 0]),
      sampleRate: 48_000,
      rootMidi: 60,
      keyLow: 48,
      keyHigh: 72,
    },
  ],
  bits: 12,
  downsample: 2,
  tone: 0.62,
  drive: 0.18,
})

const acid = instrument.acidBass({
  wave: 'saw',
  cutoff: 720,
  resonance: 0.7,
  envMod: 0.8,
  decay: 0.12,
  accent: 0.4,
  drive: 0.25,
})

const snare = instrument.drumVoice({
  kind: 'snare',
  decay: 0.32,
  snap: 0.7,
  noise: 0.85,
})

const strings = instrument.stringMachine({
  attack: 0.2,
  release: 1.1,
  tone: 0.78,
  depth: 0.8,
  modulation: 0.5,
})

const poly = instrument.polySynth({
  pulseWidth: 0.42,
  sub: 0.4,
  cutoff: 1400,
  resonance: 0.35,
  envMod: 0.45,
  chorus: 0.55,
})

Graphs are serializable node trees. The offline renderer is deterministic, noise() is seeded, typed-array buffers stay caller-owned, and modulation is represented as explicit graph edges. Public render() uses the embedded Rust/WASM DSP engine, and live playback routes through the WASM AudioWorklet runtime.

Live trigger() notes use atSec on the AudioContext clock. gateMs, velocity, and note frequency are encoded into the same WASM event request as offline rendering, so suspending the context also suspends scheduled playback. Finite envelope and effect tails retain their derived duration. For gated graphs without a finite tail, play(ctx, graph, { maxTailMs }) bounds child ownership after the gate; the default is 30 seconds. Trigger voices without an explicit gate use a one-second gate. stop() remains the immediate terminal operation, while release() applies the handle's output fade at the current audio-clock time. Runtime-gated releases scheduled with atSec take effect at the first render-quantum boundary at or after that audio-clock time; the derived release duration begins at that boundary.