How silence is removed
A first ffmpeg pass scans the audio with silencedetect. A second pass trims every detected span out of the video and the audio at the same instants and joins what is left, so speech and picture stay aligned. The options control two things: detection threshold (how quiet the signal must be to count as silence) and minimum gap duration (how long a quiet stretch must last before it is cut). A short pad is left on each side of a cut so the first word after a pause is not clipped.
Related options
If you need a final handoff without audio, pair with mute video after the silence trim. If the output still feels too loose, use video cutter for strict section-level cuts.