Parameters

AudioBits API · 0.2.0 · Available on npm.

Impact's intensity is play-only; thruster's throttle is live. Both range from 0 to 1. Defaults are 0.5 and 0.2 respectively.

Install impact, thruster from the source registry under src/sounds/recipes/, then use this host code in src/:

import { createAudio } from "audiobits";
import { impact } from "./sounds/recipes/impact/recipe";
import { thruster } from "./sounds/recipes/thruster/recipe";
const audio = createAudio();
const hit = audio.sound(impact);
const motor = audio.sound(thruster);
export async function playFromGesture() {
  await audio.start();
  hit.play({ parameters: { intensity: 0.8 }, seed: 42 });
  const voice = motor.play({ parameters: { throttle: 0.2 }, seed: 42 });
  voice.set({ throttle: 0.75 });
  voice.stop();
  await voice.ended;
}
export async function dispose() {
  await audio.dispose();
}

Omitted controls use defaults. Up to 16 parameters declare min, max, default, and mode. Live declarations include smoothing of 0.005–1 seconds. Mappings target source frequency, filter frequency, or layer gain, with linear or exponential control-to-target conversion. Live targets ramp linearly from their current value over smoothing time. Layer gain ramps in linear amplitude; the envelope remains independent. Automation points use play-only controls. Exponential mappings require positive endpoints; mapping ranges may descend, while random ranges must be ordered. Variation is sampled once per property per voice, rather than continuously.

voice.set() validates all keys before any update. Unknown names, invalid numbers, out-of-range values, and play-only updates fail atomically. Updates after Stop or completion fail with ended-voice. voice.parameters is a frozen snapshot of requested values, not a readback of the current smoothed value.

Seeded replayCopy Anchor Link

Seeds are unsigned 32-bit integers, including zero. Playback without an explicit seed selects one and exposes it on voice.seed. The same normalized recipe, controls, seed, and sample rate reproduce variation and generated noise. This does not promise identical oscillator/filter samples across browsers or sample rates. Noise uses one bounded mono second per layer with a 20 ms loop crossfade; live controls reuse sources and buffers.

Catalog workbenches use seed 42 initially. Its copy action captures the next Play's parameters and seed; changing a seed does not retrigger an active voice.