werman/noise-suppression-for-voice — reverse-engineered prompt

Build me a simple real time voice noise suppression plugin that people can use as a mic cleanup tool for calls, streaming, and recording. It should be based on RNNoise style speech denoising and be aimed at removing things like fan noise, office noise, car or train noise, and some keyboard noise, without pretending to fix a bad mic. I want it to work in the common plugin formats people actually use on Windows, Linux, and Apple devices, including VST, LV2, LADSPA, AU, and AUv3 if possible.

Keep the app very minimal, with a small GUI that exposes the important controls and some live diagnostic stats. The controls should include a voice detection threshold, a grace period so word endings do not get chopped off, and a retroactive grace option even if that adds latency. Make it clear that the audio input must be 48000 Hz and handle mono or multiple channels, with mono as the normal case. If you need to, look up current docs online for plugin setup on systems like Equalizer APO, PipeWire, and PulseAudio.