Getting the cleanest result
SilentCut uses RNNoise, a small neural network trained to tell speech apart from background sound. It listens to your recording in 10-millisecond slices and turns down whatever it judges to be noise, which is why it handles fans, hiss and room tone well while leaving the voice in place.
Preview the noisiest part
Use the Preview from slider to jump to a stretch where the noise is obvious — a pause between sentences is ideal. Flip between Original and Cleaned while it plays; the playhead stays put so you hear the same moment both ways.
Turn it down if voices sound thin
Heavy noise reduction can make voices sound watery. If that happens, drop the strength to 60–70%. Some background sound comes back, but speech stays natural — usually the better trade for podcasts and interviews.
Clean first, then cut silence
Once the noise is gone, pauses become true silence, which makes automatic silence detection more accurate. Take the cleaned file into the SilentCut studio to remove dead air.