Threshold finds quiet audio. Minimum silence duration decides which gaps qualify. Padding keeps space at both edges of a detected gap. Change one control at a time, then listen around the marked cuts.
Start with a short section you can judge
Choose a recording, open Silence Remover and listen to a passage containing both speech and a pause. Keep the original available for comparison. A quiet word ending can sit close to the room tone, so the waveform alone cannot tell you whether it should be removed.
AudioCut starts with a threshold of −40 dB, a minimum silence duration of 0.6 seconds and 0.15 seconds of padding at each edge. These are starting settings, not a prescription for every microphone or room. The controls allow −60 to −20 dB, 0.2 to 3 seconds, and 0 to 0.5 seconds of padding.
Threshold chooses the level to look below
A threshold closer to zero includes louder sounds. Moving from −40 dB to −30 dB can mark more material, including soft speech. Moving to −50 dB makes detection more selective.
If almost nothing is marked, listen to the background in a pause before raising the threshold. A fan, traffic or microphone hiss may be keeping that passage above the detection level. Silence removal does not subtract that noise from the spoken parts.
Duration qualifies the gap; padding protects its edges
Minimum duration is the length a quiet run must reach before it qualifies. Increasing it can preserve short natural pauses. It does not set the length of every remaining pause.
After a gap qualifies, padding leaves a margin at both edges. For a one-second gap with 0.1 seconds at each edge, approximately 0.8 seconds is selected for removal. If the padding consumes the gap, there may be nothing left to cut. Detection works in small windows, and manual boundaries can change the final cut.
0.1 sRemove approximately 0.8 sKeep
0.1 s
Keep the edges of the existing gap. No new silence is inserted.
Match the adjustment to the symptom
Use the table to choose one change, then replay the same passage. Changing all three controls together makes it harder to identify the cause.
| What you notice | Try first | Listen for |
|---|---|---|
| Soft endings disappear | Lower threshold; increase padding | The complete word at each edge |
| Short breaths or pauses vanish | Increase minimum duration | A natural sentence rhythm |
| Long waits remain | Check the room tone and threshold | A quiet run that actually qualifies |
| A cut sounds abrupt | Restore it or move its boundary | A smooth join with meaning intact |
Review the joins before exporting
Compare Original and Processed audio around each questionable edit. Restore a selected section when it belongs in the recording. Check the beginning and ending of the export range as well as the middle.
Export creates a new MP3 file. A shorter result is not automatically a better one: an interviewee’s pause can communicate uncertainty or emphasis. Keep the pauses that serve the recording.
Questions about this task
Does padding add new silence?
No. It keeps part of an existing detected gap at each edge. It does not insert generated silence.
Should all recordings use the same threshold?
No. Recording level, room noise and soft speech differ. Start conservatively and compare a representative passage.