Apache-2.0 · Zero runtime dependencies · WebAssembly
librosa-compatible analysis, broadcast-grade mastering and mixing, built-in instruments, and a realtime headless-DAW runtime — for C++, Python, and the browser. The same engine runs native and WASM: no Python runtime, no GPL/AGPL, no model weights.
Try it locally
Every demo runs the same Apache-2.0 processors locally in your browser via WebAssembly.
Audio player with real-time chroma & spectrum visualization.
Hit streaming loudness targets with presets. WAV export.
Structure, harmony, melody, loudness, and spectral views.
Up to eight stem tracks with scene JSON and WAV bounce.
Local microphone voice changer with library character presets.
Estimate room geometry, reverb, and source distance in 3D from a recording.
Play the built-in polyphonic synth from your keyboard or a USB MIDI keyboard.
Step-sequence three tracks, mix them, and bounce the loop to WAV.
Use it in your code
import { init, detectBpm, detectKey, analyze } from '@libraz/libsonare'
await init()
const bpm = detectBpm(samples, sampleRate)
const key = detectKey(samples, sampleRate) // { name: "C major", confidence: 0.95 }
const result = analyze(samples, sampleRate)What ships in the box
BPM, key, chords with Viterbi/HMM smoothing, beats, downbeats, time signature, sections, timbre, dynamics, YIN/pYIN pitch, tempogram/PLP, NNLS chroma, LUFS, room-acoustic estimates, and geometric room tools. Defaults are validated against generated reference values in CI.
Feature map →76 named mastering DSP processors against published references — loudness, true peak, crossovers, clippers, tube saturation, oversampling — plus a real-time-safe mixer with channel strips, buses, sends, metering, scene presets, editing DSP, and creative FX.
Mastering processors →A 7-engine synth with a data-free GM fallback bank and a GS-compatible SoundFont 2 player — MIDI never renders silent. Author projects with audio and MIDI tracks, undo/redo, MIDI sequencing, SMF import/export, offline bounce, and a sample-accurate realtime engine that runs in the browser via AudioWorklet.
Built-in instruments →C++, C, Python, Node, WASM, and CLI under one Apache-2.0 license.