web-audio-api

Pure JS implementation of Web Audio API — runs in Node.js, Deno, Bun, serverless, and edge runtimes.

Implemented

• AudioContext (state machine, suspend/resume/close, baseLatency)
• OfflineAudioContext (startRendering, oncomplete)
• AudioParam (automation events, a-rate/k-rate, cancelScheduledValues)
• AudioBuffer (via audio-buffer)
• AudioBufferSourceNode
• ConstantSourceNode
• OscillatorNode (sine/square/sawtooth/triangle/custom via PeriodicWave)
• GainNode
• StereoPannerNode
• PannerNode (3D spatial, equalpower, distance/cone models)
• DelayNode
• BiquadFilterNode (8 filter types, Audio EQ Cookbook)
• WaveShaperNode (with 2x/4x oversampling)
• IIRFilterNode (Direct Form II Transposed)
• ConvolverNode (time-domain, normalize)
• DynamicsCompressorNode (soft knee, envelope follower)
• ChannelSplitterNode / ChannelMergerNode
• AnalyserNode (radix-2 FFT, Blackman window)
• ScriptProcessorNode (deprecated but supported)

Install

npm install web-audio-api

Usage

import { AudioContext } from 'web-audio-api'

const ctx = new AudioContext({ sampleRate: 44100 })
ctx.outStream = writableStream // provide output destination

const osc = ctx.createOscillator()
osc.frequency.value = 440
osc.connect(ctx.destination)
osc.start()

Node.js with speaker

import { AudioContext } from 'web-audio-api'
import Speaker from 'speaker'

const ctx = new AudioContext()
ctx.outStream = new Speaker({
  channels: ctx.format.numberOfChannels,
  bitDepth: ctx.format.bitDepth,
  sampleRate: ctx.sampleRate
})

Piping to aplay (Linux)

import { AudioContext } from 'web-audio-api'
const ctx = new AudioContext()
ctx.outStream = process.stdout

node script.js | aplay -f cd

Architecture

Pull-based audio graph. AudioDestinationNode pulls from the graph via _tick(), each node pulls its inputs recursively. 128-sample blocks (spec render quantum). Buffer reuse throughout — no allocation in hot paths.

EventTarget ← Emitter ← DspObject ← AudioNode ← concrete nodes
                                    ← AudioParam
EventTarget ← Emitter ← AudioPort ← AudioInput / AudioOutput
EventTarget ← AudioContext

New nodes extend AudioNode, override _tick(), optionally separate DSP into static functions for future WASM swap.

Alternatives

• node-web-audio-api — Rust/napi-rs, native addon (Node.js only)
• web-audio-api-rs — pure Rust
• web-audio-engine — JS, archived 2019

License

MIT

🕉