audio.js
Organization@audiojs
Open audio stack for JS
116
Published Tools
91
Total Stars
0
Weekly Downloads
74
GitHub Followers
72
Public Repos
Published Tools
1 MCP Server115 Skillsacross 2 categoriesCabinet simulation — measured-IR convolution (pair with @audio/measure-ir) or classical speaker-sim biquads
Full beat detection pipeline: onset detection → tempo estimation → beat grid
Tempo estimation via autocorrelation of onset detection function
Beat detection — umbrella for @audio/beat-* atoms (onset detection, tempo estimation, beat tracking)
Onset detection — spectral-flux and energy-flux onset detection functions (ODFs) + adaptive peak picking
Decode tracker modules — MOD, XM, S3M, IT, MPTM and more — with libopenmpt WASM
Decode the audio track of MP4, MOV, M4A, M4V and 3GP files (AAC, ALAC, MP3, FLAC, Opus, AMR, PCM) to PCM samples
Decode the audio stream of AVI files (PCM, MP3, AAC, µ-law, A-law) to PCM samples
Decode E-AC-3 (Dolby Digital Plus) and AC-3 with FFmpeg's eac3 decoder, WASM
Decode DTS (DTS Coherent Acoustics core) audio with libdca WASM
Decode Musepack (MPC, SV7 and SV8) audio with libmpcdec WASM
Decode Monkey's Audio (APE) with FFmpeg's ape decoder, WASM
Voice activity detection — frame-level speech decision (likelihood ratio over a minimum-statistics noise floor, anchored on voicing) plus per-bin speech-presence probability and decision-directed a-priori SNR
...moreDe-breath: breaths between phrases taken down, told by their length, level and the pauses either side; the room and the speech stay
...moreNoise reduction and restoration: gate, spectral subtraction, Wiener, OM-LSA, dehum, declick, decrackle, declip, dewind, deplosive, deesser, debreath, dereverb, debleed, dewow, desqueak
...moreMains hum, buzz and steady tone removal: the mains phase tracked, every line subtracted along it, then each tone held through the take (a whine, a test tone) found and subtracted; nothing found, no change
...moredenoise — content-aware auto-selector: finds the defect a recording carries, if any, and routes to its method
De-reverb: late reverberation off a voice, one channel: weighted prediction error (WPE, Nakatani et al. 2010) fitted over the take, then the late power the past carries, scaled by the take's own decays, taken by a gain; a dry take or music passes untouche
...moreDe-crackle: dense small impulses found as outliers of both the AR prediction error and the two-sided error, all of a window rebuilt at once by least-squares AR interpolation, looked for again until none is new, each event kept only where an impulse added
...moreLinear predictive coding — Levinson-Durbin AR analysis, prediction, forward extrapolation, and least-squares gap interpolation
...moreDe-click: clicks found as outliers of the AR prediction error, each taken for a gap or a damped resonance by BIC, the resonance taken off, the gap rebuilt as its posterior mean under AR interpolation, everywhere or in given regions
...moreWow & flutter correction — pitch-drift removal for tape/vinyl/cassette transfers
De-bleed: a source's spill taken out of another microphone, given that source's own track. A Kalman filter cancels it through the room path it learns, a Wiener gain takes the residual it predicts
...moreSpectral repair: rebuild dropouts and damaged time×frequency regions by AR interpolation, a sinusoidal bridge or a similarity transplant, routed by length and content (iZotope RX class)
...moreDe-plosive: close-mic 'p', 'b' pops found as sudden aperiodic bursts under 80 Hz and taken out as a linear-phase low band, ~14 ms look-ahead
...moreDe-clip: each side's rail found where the sound was cut flat, or where lossy coding spread the cut into a band, or soft saturation's curve fitted blind by maximum likelihood and inverted, every clipped sample rebuilt three times, AR (Janssen) and sparse (
...moreFeed-forward soft-knee compressor, gain reduction smoothed in the log domain (Giannoulis, Massberg & Reiss 2012)
Dynamics processing — umbrella for @audio/dynamics-* atoms (compressor, limiter, gate, expander, de-esser, ducker, softclip, compand, transient shaper)
...moreDe-esser: the sibilance band's level over the voice body's drives a cut held within a range, read 5 ms ahead, on the band over 3.5 kHz (linear phase), a band, or broadband; the same at any recording level
...moreExternal-sidechain compressor: main signal's gain is reduced in proportion
Noise shaping — error-feedback dithering for bit-depth reduction
Encode PCM audio samples to MP4/M4A (ISO BMFF) — pure JS muxer, remuxer and iTunes-tag writer
Encode PCM audio samples to AAC (ADTS): WebCodecs, else the Fraunhofer FDK encoder (WASM)
Encode PCM audio to Apple Lossless (ALAC) — pure JS port of Apple's reference encoder
Baxandall tone control — bass/treble shelving with reciprocal characteristics
Dynamic EQ — per-band peaking gain driven by band energy (Pro-Q3/soothe class, generalized de-esser)
FIR convolution EQ — linear-phase, arbitrary magnitude response (frequency-sampling design; firequalizer / match-EQ class)
...moreAudio-facing filters — umbrella for @audio/filter-* atoms (biquads, comb, resonator, DC blocker, pre-emphasis, spectral tilt, analog ladder models)
...moreSpectral tilt filter: apply a constant dB/octave slope to the spectrum
The audio-side biquad — RBJ cookbook coefficients (Web Audio conventions) + stateful SOS kernel (process/step/cascade/filter)
...moreDerivative / integral — FFmpeg aderivative/aintegral first-difference & running-sum pair
Variable-bandwidth filter with per-sample smoothed coefficient trajectory
Moog 4-pole transistor ladder filter (-24 dB/oct lowpass with resonance; highpass, bandpass, 2-pole modes)
Oberheim SEM state-variable filter (–12 dB/oct multimode)
Loudness metering — umbrella for @audio/loudness-* atoms (BS.1770-4 LUFS, true peak, LRA, ReplayGain, DR, speech contrast, sound labeling)
...moreMusic information retrieval — umbrella for @audio/mir-* atoms (chroma, chord, key; MIREX-parity roadmap)
Source separation (stems): a song's by SCNet (MIT weights), Hybrid Transformer Demucs, Open-Unmix with multichannel Wiener refinement; a soundtrack's dialogue, music and effects by MRX (MIT weights) and TIGER (Apache-2.0); model-agnostic ONNX
...moreA sound made again from its mel spectrogram by Vocos: a vocoder's artifacts (a TTS's phase smear, its frames ringing on) give way to Vocos's own, through @audio/neural-runtime
...moreTimbre recognition: a small encoder maps one note to a vector, a catalog of preset vectors names the sound (Yamaha DX7 first), ranked by cosine through @audio/neural-runtime
...moreModel inference adapter — ONNX Runtime (Node/wasm/webgpu) behind one worklet-ready interface
Polyphonic note transcription with pitch bends: Spotify's Basic Pitch (ICASSP 2022) through ONNX Runtime, in Node and the browser
...morePitch detection — umbrella for @audio/pitch-* atoms (YIN, pYIN, McLeod, HPS, cepstrum, SWIPE, autocorrelation, AMDF)
Polyphase FIR resampling — streaming-friendly rational rate conversion (48↔16/24 kHz voice-agent path)
LPC source-filter pitch shift — repitched excitation residual through the unmodified all-pole vocal-tract filter
Canonical pitch-shifting algorithms: phase vocoder, peak-locked, transient-aware, PSOLA, WSOLA, granular, PaulStretch, SMS, HPSS, formant, sampler, hybrid
...moreTransient-aware phase vocoder pitch shift with attack preservation
Formant-preserving pitch shift via cepstral envelope re-imposition
OLA (Overlap-Add) pitch shift — simplest stretch+resample baseline
Hybrid pitch shift — runs phase-lock + WSOLA in parallel, crossfades by transient confidence
Delay-line (harmonizer) pitch shift — dual crossfading taps sweeping a modulated delay window
PSOLA (Pitch-Synchronous Overlap-Add) pitch shift — pitch-synchronous grains for monophonic voice
Sinusoidal analysis — STFT peaks (parabolic) + nearest-frequency partial tracking (McAulay-Quatieri 1986)
Mid/side encode-decode — L/R ↔ M/S matrix with width control
Structural binaural panner — Brown-Duda spherical-head ITD + head-shadow ILD (no HRIR data)
WSOLA time stretching — time-domain: speech and solo voices keep their pulse shape and attacks
PVSOLA time stretching: a phase-locked vocoder reset to the input's own waveform where it fits; speech slowed without flanging or phasiness
...morePhase vocoder time stretching — educational baseline without phase locking
Time stretching — umbrella for @audio/stretch-* atoms (WSOLA, phase vocoder variants, PaulStretch, PSOLA, SMS)
Transient-aware phase-locked vocoder (Röbel 2003) — preserves attacks on drums and plucks
Hybrid harmonic/percussive time stretch (Driedger) — HPSS split, phase-locked vocoder + OLA per layer
Sinusoidal Modeling Synthesis time stretching — harmonic/tonal material
Pitch-Synchronous Overlap-Add — time-domain stretch for speech and monophonic instruments
Paulstretch — Paul Nasca's extreme stretch (8×+) for ambient/drone material
Polyphonic voice allocator — wrap any voice with N-voice polyphony
FM synthesis — Chowning phase modulation, single-op or serial op stack, bell/epiano presets
Synthesis & signal generators — umbrella for @audio/synth-* atoms (noise colors, oscillator, chirp, DTMF, pluck, risset drum, rhythm track, envelope, LFO)
...moreRetune a voice along a pitch curve, its own glottal cycles re-spaced (TD-PSOLA): formants and consonants kept, streaming (Melodyne class)
...morePitch correction to a scale: each note's centre moved onto its note, its slow drift flattened, the voice's own cycles re-spaced along one pitch curve (Auto-Tune / Melodyne class)
...moreNote-target pitch correction — align melody to reference MIDI/notes (Melodyne class)
IEC 61672-1:2013 C-weighting: H(s) = K·s² / ((s+w1)²(s+w4)²)
RIAA playback (de-emphasis): H(s) = (1 + s*T2) / ((1 + s*T1) * (1 + s*T3))
IEC 61672-1:2013 / ANSI S1.42 B-weighting filter — moderate-loudness frequency response
IEC 537 D-weighting filter — aircraft-noise frequency response (resonant hump near 4 kHz)
ITU-R BS.468-4 noise weighting, exact matched-z realization: peaked at