Keep one continuous section of an audio file and discard the rest. Drag a range across the waveform, confirm it by ear, then export that range.
Marking the range
Dragging across the waveform sets the range, and the rail shows the start, end, length, and how much gets discarded. Select all resets it to the whole file, which is a sensible starting point when you only need to shave the end.
Clicking moves the playhead without changing the range. From there the arrow keys step 5 seconds, Shift with the arrow keys steps 1 second, and Home and End jump to the visible bounds. Once the playhead sits where a boundary belongs, Start here and End here move that edge to it, which is more exact than redrawing the range by hand.
Check the boundaries by ear
A trim that looks right on the waveform can still cut a breath, a reverb tail, or the attack of the first note. Turning on Loop repeats the selection so you hear the join between the end and the start over and over, which is where an awkward boundary becomes obvious.
Silence is deceptive at the start of a range. A quiet lead-in often reads as flat on the waveform while still carrying room tone, and cutting into it produces an abrupt shift in the noise floor that is more noticeable than the sound you removed.
The output keeps the source format
An MP3 in gives an MP3 out, and a WAV gives a WAV. Trimming is a cut, not a conversion, so there is no format menu here. Use the Audio Converter when the container has to change.
Trimming re-encodes
The selected range is decoded, then encoded again to produce a valid file. There is no way around this: cutting at an arbitrary point in a compressed file does not line up with the encoder’s frame boundaries.
For a lossy source such as MP3 or Ogg, that adds one generation of loss on top of whatever the file already had. Convert the source to WAV or FLAC first when the trimmed file is an intermediate step rather than the finished piece. The bitrate of the export is read from the source, so a 96 kbps MP3 does not come back at 192 kbps.
One range at a time
The export covers a single continuous range. Removing a section from the middle means exporting the part before it and the part after it as two files, then joining them elsewhere.
Trimming long recordings
Only the selected range is decoded and encoded, so pulling a 30 second clip out of a two-hour recording takes about as long as a 30 second file. The source is read incrementally rather than loaded whole, which keeps memory flat regardless of how long the recording is.