Audio Noise Reducer
Reduce background noise from audio in the browser using spectral subtraction and a noise gate. Drag-and-drop or pick a file, decode with Web Audio API, estimate the noise floor from a leading silent region, subtract the average noise spectrum from each FFT window, and silence samples below a dBFS threshold. Then re-encode as a 16-bit PCM WAV (pure-JS encoder). 4 strength presets (light/medium/strong/aggressive), 4 noise-gate presets (-30/-40/-50/-60 dBFS), 5 FFT sizes, hanning windowing, magnitude/phase separation, IFFT reconstruction, file-size estimator, text + CSV report, history (localStorage), shareable URL, summary stats (input level, noise level, output level, % noise reduced). 100% client-side.
About Audio Noise Reducer
Reduce background noise from audio in the browser using spectral subtraction and a noise gate. Drag-and-drop or pick a file, decode with Web Audio API, estimate the noise floor from a leading silent region, subtract the average noise spectrum from each FFT window, and silence samples below a dBFS threshold. Then re-encode as a 16-bit PCM WAV (pure-JS encoder). 4 strength presets (light/medium/strong/aggressive), 4 noise-gate presets (-30/-40/-50/-60 dBFS), 5 FFT sizes, hanning windowing, magnitude/phase separation, IFFT reconstruction, file-size estimator, text + CSV report, history (localStorage), shareable URL, summary stats (input level, noise level, output level, % noise reduced). 100% client-side. Everything runs locally in your browser — your data never leaves your device.
How to use
- Enter your input in the tool above.
- Adjust any options to your preference.
- Use the Copy or Download buttons to save the result.
- Everything happens locally — your data never leaves your browser.
FAQ
How does the audio noise reducer work?
Drop or pick an audio file. We decode it with the Web Audio API, split it into small overlapping windows (e.g. 1024 samples), apply a Hanning window, and run an FFT on each. The first ~100ms of the signal is treated as a noise sample: we average its magnitude spectrum to estimate the noise floor. For every subsequent window we subtract a scaled copy of the noise floor from the magnitude (clamped at zero), keep the original phase, and run an inverse FFT. We then overlap-add the windows back together. An optional noise gate silences samples below a dBFS threshold. The result is re-encoded as a 16-bit PCM WAV file (pure-JS RIFF encoder, no library).
What strength and noise-gate presets are available?
Four strength presets control how aggressively the noise floor is subtracted: light (25), medium (50), strong (75), and aggressive (100). Higher values remove more noise but may introduce musical-noise artifacts. Four noise-gate threshold presets are also available: -30 dBFS (gentle), -40 dBFS (default), -50 dBFS (aggressive), and -60 dBFS (only the quietest hiss). Anything quieter than the threshold is silenced.
What FFT sizes can I choose?
Five power-of-two FFT sizes: 256, 512, 1024 (default), 2048, and 4096. Smaller sizes track transient noise better; larger sizes give finer frequency resolution but smear attacks. A 50% hop size and Hanning window with overlap-add reconstruction guarantee perfect reconstruction when no subtraction is applied.
What extra features does this tool have compared to others?
(1) 4 strength presets (light/medium/strong/aggressive). (2) 4 noise-gate threshold presets (dBFS). (3) 5 FFT sizes. (4) Noise estimation from a sample window (average magnitude spectrum). (5) Spectral subtraction algorithm (subtract scaled noise floor, clamp to zero). (6) Noise-gate applier (silence samples below threshold). (7) Hanning window generator. (8) Magnitude/phase calculator from complex FFT. (9) Reconstruction from magnitude/phase via inverse FFT. (10) Pure-JS radix-2 Cooley-Tukey FFT + IFFT. (11) File-size estimator. (12) WAV encoder (44-byte RIFF header + 16-bit PCM). (13) Text report generator. (14) CSV report generator. (15) History (localStorage, last 20). (16) Shareable URL settings. (17) Summary stats (input level, noise level, output level, % noise reduced, clipping prevented).
Is my data sent anywhere?
No. The audio file you load, decoded samples, FFT data, and output WAV never leave your browser. There is no upload. Noise-reduction metadata in history is stored in localStorage on this device only.
Related tools
Audio Bitrate Calculator
Calculate audio bitrate, file size, duration, and quality metrics entirely in the browser. Supports PCM/WAV, FLAC, ALAC, MP3, AAC, OGG/Vorbis, and Opus. Compute bitrate from sample rate × bit depth × channels (PCM), estimate FLAC/ALAC compression (50–60% of PCM), map lossy quality presets (low/medium/high/voice/music) to typical bitrates, compute file size from bitrate × duration, calculate streaming bandwidth requirement, reverse-calculate required bitrate from target file size, and score quality 0–100. Includes 7 sample rates, 5 bit depths, 4 channel layouts, 7 formats, 5 quality presets, side-by-side format comparison, text/CSV exports, history (localStorage), shareable URL, summary stats. 100% client-side.
Audio Converter
Convert audio files between formats entirely in the browser. Decode with Web Audio API, re-encode to WAV (PCM 16-bit), WebM/OGG (via MediaRecorder), or MP3 (via MediaRecorder with WAV fallback). Multi-format output, 4 bitrate presets, 6 sample-rate presets, mono/stereo channel selection, pure-JS 44-byte RIFF WAV header builder, Float32 to 16-bit PCM converter, file-size estimator, quality score calculator (0-100), lossless-vs-lossy compatibility checker, smart filename generator, conversion history (localStorage), shareable URL, drag-and-drop, and summary stats. 100% client-side.
Audio Equalizer
Apply a 10-band graphic EQ to audio entirely in the browser. Drag-and-drop or pick a file, decode with Web Audio API, adjust the gain (-12 dB to +12 dB) on each ISO-standard band (31, 62, 125, 250, 500, 1k, 2k, 4k, 8k, 16k Hz), apply via a chain of peaking BiquadFilterNodes in an OfflineAudioContext for batch processing, and re-encode as a 16-bit PCM WAV (pure-JS encoder). 10+ preset library (flat, bass-boost, treble-boost, vocal-boost, loudness, rock, pop, jazz, classical, podcast), 3-band & 5-band simplified modes, 7 gain presets per band, default Q factor 1.41, gain validator, filter-chain config builder, file-size estimator, text + CSV report, history (localStorage), shareable URL, summary stats. 100% client-side.