Kleervox removes background noise and speaker echo from your microphone in real time, then hands every call app a clean virtual microphone. It works with PipeWire and PulseAudio, and there is no driver to install.
The usual Linux answer to microphone noise is a manual setup: load a suppression module, wire a null sink, remap a source, then repeat the whole thing after every reboot or audio server update. It works, but it is fragile and hard to explain to a teammate.
A gate simply cuts audio below a threshold, which clips the start of words and leaves noise audible whenever you speak. A neural model separates voice from everything else continuously, so quiet speech survives and steady hum does not.
Per-app suppression has to be configured again in each program, and many Linux clients expose no such setting at all. A system-level virtual microphone is selected once per app and then behaves like ordinary hardware.
The virtual device is created by the running audio server when the app starts and removed when it exits, so there is no module to reload and no config file to keep in sync.
The Debian package installs from a terminal, and the virtual microphone appears the first time you start the app.
Download the .deb from the download page and install it with sudo apt install ./kleervox_*_amd64.deb.
Open Kleervox and choose your physical microphone — a headset, USB mic, or the built-in laptop input. Never select the virtual microphone here.
In Zoom, Google Meet, Discord, Teams, or OBS, open the audio settings and choose the virtual microphone as the input. Keep your headphones as the output device.
Confirm on a real call that your voice sounds natural and that keyboard and fan noise are gone. Adjust the cancellation strength only if speech starts to sound processed.
Because the cleaned audio is presented as a normal input device, anything that lets you pick a microphone can use it — including browser-based calls, where per-app suppression settings are usually unavailable.
New to how the processing itself works? Read the overview of AI noise cancellation, or the practical guide to reducing background noise in a call center. On a Windows machine instead, see noise cancellation for Windows.
If you take calls on speakers rather than headphones, the other person’s voice leaves your speakers and returns through your microphone. That returning sound is speech, so a noise model will not remove it — it is trained to protect speech.
Echo cancellation handles that path independently and is available on Linux systems running PipeWire. It can stay on while noise cancellation is off, or the other way round.
Live microphone audio is never uploaded for cleaning. The model runs locally, which matters for confidential calls and for anyone who chose Linux partly to keep that control.
It also means the cleaning path does not depend on your connection — only the call app itself needs the network.
Try it with your own microphone in the browser before installing anything.
Yes. Kleervox runs natively on Linux and cleans the microphone in real time, then exposes the result as a virtual microphone that any call or recording app can select. Many noise-cancellation products ship Windows and macOS builds only, which is why Linux users often fall back to manual filter chains.
Both. PipeWire is preferred and is the default on current Ubuntu, Fedora, and most modern desktops. Systems still running PulseAudio are supported through the equivalent null-sink and remap-source setup. The app detects which audio server is running and configures itself accordingly.
No. On Linux the virtual microphone is created at runtime by the audio server, so there is no kernel module, no third-party audio cable, and no signed driver to approve. Start Kleervox and the "Kleervox Microphone" input appears; quit the app and it is removed again.
Open Kleervox, select your real microphone inside it, then open Discord's Voice & Video settings and choose "Kleervox Microphone" as the input device. Discord's own suppression can stay off, since the signal it receives is already cleaned.
Yes. Add an Audio Input Capture source in OBS and select the Kleervox virtual microphone. Because the cleaning happens before OBS receives the signal, the recorded or streamed audio is already free of keyboard clicks, fans, and room noise.
Yes. Echo cancellation is a separate stage from noise cancellation and is available on Linux systems running PipeWire. It targets the remote participant’s voice leaking from your speakers back into your microphone. Headphones remain the most complete way to eliminate that path.
No. Noise and echo cancellation run entirely on your own machine. Live microphone audio is not sent to a server for processing, so the cleaning keeps working even on an unstable connection.
A Debian package covers Ubuntu, Debian, Linux Mint, Pop!_OS, and other Debian-derived distributions, and an AppImage runs on distributions outside that family. The only real requirement is a working PipeWire or PulseAudio audio server, which is standard on current desktop installs.
The browser demo mixes your microphone with sample background noise so you can switch between the original and cleaned signal. It runs on your machine too — nothing is uploaded.
Test noise cancellation free