Noise Suppression for Voice is a real-time audio plugin that suppresses background noise in voice recordings and streams.
The plugin addresses the problem of unwanted background noise during voice communication and recording by applying RNNoise, a recurrent neural network-based noise suppression algorithm. It works by analyzing incoming audio to distinguish voice from noise sources such as computer fans, office environments, crowds, aircraft, vehicles, and construction sounds, then attenuating the noise while preserving voice quality. The tool operates on 16-bit audio at 48000 Hz sample rate and supports mono or multi-channel input. It includes a minimalistic graphical interface for adjusting three main parameters: VAD Threshold to control the voice detection sensitivity, VAD Grace Period to prevent cutting off word endings, and Retroactive VAD Grace Period to preserve word beginnings at the cost of added latency.
The plugin suits developers and end users who need noise suppression integrated into their audio workflow across multiple platforms and digital audio workstations. It is available as VST2, VST3, LV2, LADSPA, AU, and AUv3 plugins, making it compatible with most professional and consumer audio software on Windows, Linux, and macOS. The tool works best with reasonably good microphone input; poor microphone quality may result in the algorithm misclassifying voice as noise, potentially degrading rather than improving audio quality. Mild background noise is consistently suppressed, while louder sounds like mechanical keyboard clicks are reduced when voice is present but may remain audible during silence.
Development activity shows ongoing maintenance with releases available through the project's GitHub releases page. The plugin includes platform-specific setup guidance, such as integration instructions for Windows with Equalizer APO and detailed configuration examples for Linux systems using PipeWire, indicating attention to practical deployment across different operating systems.
