If you need real-time, system-wide background noise suppression on Ubuntu, NoiseTorch is the most practical open-source option. It uses RNNoise and exposes a virtual microphone that any application can use. Once enabled, it works across Audacity, browsers, Teams, Zoom, OBS, and anything else that can select an input device.
Installation (Manual Binary)
NoiseTorch is not available via apt or Flathub. Install it from GitHub releases.
1. Download the latest release
Check available versions:
https://github.com/noisetorch/NoiseTorch/releases
Example:
wget https://github.com/noisetorch/NoiseTorch/releases/download/v0.12.2/NoiseTorch_x64_v0.12.2.tgz
2. Extract and move the binary
tar -xzf NoiseTorch_x64_v0.12.2.tgz
sudo mv noisetorch /usr/local/bin/noisetorch
Grant Required Capability
NoiseTorch needs real-time scheduling.
sudo setcap 'CAP_SYS_RESOURCE=+ep' /usr/local/bin/noisetorch
Verify:
getcap /usr/local/bin/noisetorch
Expected output:
/usr/local/bin/noisetorch cap_sys_resource=ep
Do not run NoiseTorch with sudo.
Run NoiseTorch
noisetorch
In the UI:
- Select your physical microphone
- Click Load
This creates a new input device:
- NoiseTorch Microphone / Filtered Microphone
You can confirm it exists with:
pavucontrol
Use It in Audacity (or Any App)
Applications will not auto-switch input devices.
In Audacity:
- Toolbar → Recording Device → NoiseTorch Microphone
- Restart Audacity if the device does not appear
From this point on, audio is cleaned before it reaches any application, which is why the suppression works system-wide.
Key Point
NoiseTorch does not post-process recordings. It filters the microphone signal itself, making it suitable for real-time use across the entire system without cloud processing or proprietary software.
Leave a Reply