Skip to content
SPECIFICTION
Get listed

@audio/denoise

audiojs.dev

Free plan Open source Web
Visit audiojs.dev
Spec score 8.4 #1 of 24 in Audio Restoration Software

What it is

@audio/denoise is a free web tool for cleaning recordings and listening to or saving the result. It handles targets including steady background noise, mains hum, clicks, pops, vinyl crackle, clipping, wind rumble, plosives, sibilance, room reverb, and noise between phrases. The page describes thirteen open-source restoration methods and a classifier. Automatic mode measures properties of a recording, selects a specialized method, and displays its plan; users can instead choose methods such as dehum, declick, declip, and gate. Audio is processed in the browser, and the page says it is not uploaded or sent off the device. The approach uses classical signal processing rather than AI noise removal. For music, the page identifies hum, clicks, crackle, and clipping as treatable, while advising light use of broadband denoising and comparison with the original. Machine-learning models may perform better on heavy speech-over-noise. The associated npm package processes mono Float32Array audio. Automatic mode does not choose dereverb, which must be selected explicitly.

Spec sheet

What it does
@audio/denoise removes common recording defects through automatic or manually selected restoration methods.Read 15 Oct 2026 · audiojs.dev
Purpose
The noise remover cleans recordings and lets users listen to and save the result.Read 7 Oct 2026 · audiojs.dev
Privacy
Processing occurs in the browser tab and the audio is stated never to leave the device.Read 15 Oct 2026 · audiojs.dev
License
The repository and npm metadata state that the package is released under the MIT license.Read 15 Oct 2026 · github.com
Automatic API
The denoise function can return cleaned samples, a classifier plan, or be forced to a named method.Read 15 Oct 2026 · github.com
Automatic classifier
The classifier evaluates hum, click density, rumble, noise-floor stability and sibilance before selecting a method.Read 15 Oct 2026 · audiojs.dev
Automatic mode
A classifier selects a specialized method based on measured properties of the recording and shows its plan.Read 7 Oct 2026 · audiojs.dev
Automatic mode limit
The repository says automatic mode does not select dereverb; users must choose it explicitly.Read 7 Oct 2026 · github.com
Browser workflow
Users can drop or paste a recording, inspect the selected method and plan, audition the result and save the cleaned file.Read 15 Oct 2026 · audiojs.dev
Classical DSP
The product states that its methods use classical signal processing rather than AI models.Read 15 Oct 2026 · audiojs.dev
Click repair
Clicks, pops and crackle are detected and reconstructed with LPC-based interpolation.Read 15 Oct 2026 · audiojs.dev
Clipping repair
The package can detect clipping levels automatically and reconstruct clipped waveform sections.Read 15 Oct 2026 · audiojs.dev
Developer demo
The repository includes a self-contained demo.html for selecting scenarios, choosing methods, viewing waveforms and hearing differences.Read 15 Oct 2026 · npmjs.com
Developer use
The repository documents installing the package with npm and processing mono Float32Array audio in and out.Read 7 Oct 2026 · github.com
Hiss processing
Hiss and fan noise can be treated with Wiener or MMSE-LSA processing, or OM-LSA when speech is present.Read 15 Oct 2026 · audiojs.dev
Hum processing
The dehum method tracks 50 or 60 Hz mains frequencies and their harmonics with adaptive notch or sinusoidal subtraction processing.Read 15 Oct 2026 · audiojs.dev
Input and output
The JavaScript API accepts and returns mono Float32Array audio; stereo should be processed channel by channel.Read 15 Oct 2026 · npmjs.com
Installation
Installation is documented with npm using `npm install @audio/denoise`.Read 15 Oct 2026 · github.com
Limit
The page says machine learning models can do better for heavy speech-over-noise.Read 7 Oct 2026 · audiojs.dev
Limit: automatic dereverb
Automatic routing does not select dereverb because the documentation says it lacks a reliable single-pass signature; dereverb must be selected explicitly.Read 15 Oct 2026 · github.com
Limit: data shape
The core API is sample-array based and does not provide a hosted upload service or graphical desktop editor in the package documentation.Read 15 Oct 2026 · npmjs.com
Limit: heavy speech noise
The maker cautions that machine-learning models can outperform these classical methods for heavy speech-over-noise cases.Read 15 Oct 2026 · audiojs.dev
Limit: single-pass operations
The package describes itself as single-pass noise reduction, and several operations such as desqueak are documented as whole-clip rather than streaming processors.Read 15 Oct 2026 · github.com
Maker history
The home page dates audiojs maintenance to 2016 and describes the stack as cross-runtime JavaScript for browser, Node, Deno and Bun.Read 15 Oct 2026 · audiojs.dev
Maker scale
The audiojs home page states more than 300 open-source packages and over 9 million npm downloads per month.Read 15 Oct 2026 · audiojs.dev
Manual selection
Users can choose methods such as dehum, declick, declip and gate by hand.Read 7 Oct 2026 · audiojs.dev
Methods
The page describes thirteen open-source restoration methods plus a classifier.Read 7 Oct 2026 · audiojs.dev
Music handling
The maker says hum, clicks, crackle and clipping work on music, while broadband denoising should be applied lightly and compared with the original.Read 15 Oct 2026 · audiojs.dev
Music use
The page says hum, clicks, crackle and clipping can be treated in music, while broadband denoising should be used lightly and compared with the original.Read 7 Oct 2026 · audiojs.dev
Noise targets
Documented targets include fan and hiss noise, mains hum, clicks, pops, vinyl crackle, clipping, wind, plosives, sibilance, reverb and inter-phrase noise.Read 15 Oct 2026 · audiojs.dev
Noise types
It lists steady background noise, mains hum, clicks, pops, vinyl crackle, clipping, wind rumble, plosives, sibilance, room reverb and noise between phrases as targets.Read 7 Oct 2026 · audiojs.dev
Package
The GitHub repository describes @audio/denoise as a single-pass noise reduction package with thirteen specialized methods and an auto-classifier.Read 7 Oct 2026 · github.com
Package scope
The npm package describes 17 specialised methods plus an automatic classifier.Read 15 Oct 2026 · npmjs.com
Praise: measurable results
The documentation includes benchmark tables with SNR, PESQ, STOI, SI-SDR, DNSMOS and noise-reduction measurements on named test sets.Read 15 Oct 2026 · npmjs.com
Praise: method breadth
The package documentation presents separate processing paths for gates, hum, stationary and changing broadband noise, impulses, crackle, clipping, wind, plosives, sibilance, breaths, silence, reverb, bleed, wow and guitar squeaks.Read 15 Oct 2026 · github.com
Processing
The page says the processing uses classical signal processing rather than AI noise removal.Read 7 Oct 2026 · audiojs.dev
Public functions
Exports include denoise, classify, dehum, wiener, declick and lower-level STFT, VAD, speech-presence and noise-estimation utilities.Read 15 Oct 2026 · npmjs.com
Quality metrics
The package exposes SNR, segmental SNR, log-spectral distance, noise-reduction ratio and speech-attenuation measurements.Read 15 Oct 2026 · npmjs.com
Reverb processing
A dereverb stage is provided for room-reverberation reduction.Read 15 Oct 2026 · audiojs.dev
Streaming API
Wiener and related processors support repeated chunk calls with state retained between blocks and a final flush call.Read 15 Oct 2026 · github.com
TypeScript
npm identifies the package as including built-in TypeScript declarations.Read 15 Oct 2026 · npmjs.com
Version
The npm listing shows version 0.11.1 and identifies the package as public.Read 15 Oct 2026 · npmjs.com
Wind and plosives
Dedicated dewind and deplosive stages address low-frequency wind rumble and microphone plosive bursts.Read 15 Oct 2026 · audiojs.dev

Plans and prices

Free Free Runs in browser · no account · no upload Read 7 Oct 2026 · audiojs.dev
Free browser noise remover Free Local browser processing · no upload · save cleaned file Read 15 Oct 2026 · audiojs.dev

Every @audio/denoise plan and price →

Strong points and limits

Strong points

  • Free to use in the browser.
  • Audio is processed on-device, not uploaded.
  • Thirteen restoration methods plus a classifier.
  • Manual choices include dehum, declick, declip, and gate.
  • Supports hum, click, crackle, and declip repair.

Limits

  • No batch processing.
  • Automatic mode does not select dereverb.
  • Machine learning may do better on heavy speech-over-noise.
  • Broadband denoising should be used lightly.

Who it suits

It suits people who want browser-based restoration for recordings, especially when targeting specific noise or defects. Developers can also use the associated npm package with mono Float32Array audio.

Verdict

@audio/denoise specifies its restoration methods, processing approach, and on-device privacy clearly. Its stated limitation for heavy speech-over-noise and the need to select dereverb manually are worth noting.

Where Specifiction places it

Alternatives to @audio/denoiseAll alternatives →