NoiseTorch is a cross-platform audio processor and real-time noise filter designed to suppress ambient sound from audio streams. It functions as a virtual microphone noise suppressor and routing tool, capturing system audio sources and directing filtered signals into virtual input or output devices.
The application uses a recurrent neural network to distinguish between human speech and ambient noise. It provides a virtual denoising microphone that removes background noise from a selected input, alongside tools for filtering audio output streams.
The system includes capabilities for audio device management, such as source enumeration and signal routing. It features voice activation tuning with configurable amplitude thresholds and system-level resource limiting to maintain stable, low-latency performance.
The software includes mechanisms for managing updates through the verification of cryptographically signed remote archives.
