A frequency-domain noise gate that removes quiet spectral content. After transforming the signal to the frequency domain, any FFT bin whose magnitude falls below the Threshold is zeroed, so steady low-level components such as hiss or hum are suppressed while louder content passes. Window length sets the FFT/analysis-window size (and the latency), a window selector chooses the analysis window — Hanning, Hamming, Blackman, Bartlett or None — and an Output control trims the level in dB. It takes one audio input and one audio output, in mono or stereo.
The input is buffered into overlapping, windowed blocks that are transformed with an FFT; each bin below the threshold is set to zero, the rest are kept, and the inverse transform is overlap-added back to the output before the output-gain trim. The transform runs on whole blocks, but samples still enter and leave one sample-frame per tick, in step with WaveWarp's sample-per-sample engine.