Free tools for your next audio edit
Noise gate online. Lower the gaps. Keep the timing.
Reduce the level of quiet passages in a short MP3 or WAV. Set a threshold, reduction, attack, hold, and release, then compare the exported WAV with your original. Free browser processing, with no account or audio upload.
Noise gate
Free · localThe example uses a quiet tone beneath louder tone bursts, not speech or typing. It is loaded only when selected and processed by this gate when you prepare a WAV.
No account, credits, or audio upload. This is file processing, not a live microphone filter or AI source separation. It reduces level in quiet passages; typing mixed with louder speech can remain.
Compare three parts of your recording
How to use the noise gate
- 01 / LISTEN
Find a pause and a quiet word
Choose a mono or stereo MP3/WAV up to 50 MiB and five minutes. Listen to the original, including a pause, a quiet wanted phrase and an overlap. Prepare a shorter excerpt with the free trimmer if needed.
- 02 / ADJUST
Choose a threshold and gentle reduction
The gate opens when a sample in either channel reaches the threshold. Below it, hold and release determine when the linked gain falls. More reduction makes gaps quieter but can make transitions conspicuous. Lower the threshold if wanted sounds are being cut.
- 03 / REVIEW
Preview the exported WAV
Apply the gate, then compare the actual output with the original. Check consonants, breaths, tails and the first sound after a pause. Download the WAV only when the result suits your edit. Each preparation starts from the selected original.
A level decision, not source recognition
What a noise gate can and cannot remove
A noise gate can reduce low-level room sound between phrases. It does not identify hum, typing or voices by their meaning. Loud typing can open the gate, and a quieter distraction under a louder voice can remain while the gate is open. For recorded typing, the keyboard cleanup comparison helps choose between a local edit, gating and source isolation.
The Audacity noise-gate manual explains threshold, timing and the risk of cutting wanted audio. Its effect has different processing, including look-ahead. This tool has no look-ahead: it starts a linear attack after a crossing, so an opening transient can be reduced. Hold and release can also leave background audible after a phrase.
This gate uses the largest absolute sample across the channels as its detector and applies the same gain to both. It keeps the decoded sample-frame count and channel positions. It does not delete silent intervals, normalize levels, remove DC, repair clipping, or guarantee unchanged speech.
Try a known input
Hear quiet gaps and louder overlaps separately
The optional example contains a quiet 3 kHz tone throughout and two louder 440 Hz bursts. Select it in the tool, then apply the gate to create a new WAV with the displayed settings. Listen to an early gap and the middle of a burst: a gate can lower the gap while leaving both tones together during the louder passage.
This is an original synthesized recording, not typing, speech or an AI separation demonstration. Download the original test WAV or read its generation notes and CC0 license. The pre-existing FFmpeg example in those notes is a different gate implementation, not the output of this tool. More original files are available in the WAV test-file library.
A few useful answers
Free online noise gate questions
How do I apply a noise gate online for free?
Choose a mono or stereo MP3/WAV up to 50 MiB and five minutes. Set the gate controls, apply the gate, and preview the actual WAV before downloading. Processing is local and requires no account or credits. The WAV output must also fit within 50 MiB.
Will a noise gate remove keyboard clicks during speech?
Not necessarily. This gate responds to level, not the identity of a sound. A key press above the threshold can open it, and typing during an open speech passage passes through. It is more useful when unwanted sound stays below the threshold in pauses.
What threshold should I choose?
There is no universal threshold. Compare a pause, your quietest wanted words, and a loud overlap. Raise the threshold enough to reduce unwanted quiet passages, but lower it again if breaths, consonants or word endings become damaged. Default settings are only starting values.
What do attack, hold, release and reduction mean?
Attack ramps the gate toward full level when a sample crosses the threshold. Hold keeps it open after the last crossing, and release ramps it back down. Reduction sets the closed level, from 0 dB (no reduction) to 60 dB quieter. This gate uses linear ramps and has no look-ahead.
Are stereo channels gated independently?
No. The highest absolute sample across the channels controls one linked gain, applied to both. A loud event on either side can keep both sides open. This preserves their relative balance but does not separate sources or move a sound between channels.
What file is downloaded?
A 48 kHz, 16-bit PCM WAV with the decoded recording's frame count and channel count. Preview and download use the same export. There is no MP3 output here, no metadata copy, no trimming or automatic loudness normalization. Browser decoding and PCM quantization can change the original samples.
Does this clean a live microphone or use AI?
No. It processes a selected recording with a level-based gate. It does not install a virtual microphone, filter a call live, or isolate a described source. Saturalabs' separate AI workflow has account and download-credit requirements.
After preparing your audio