A song does not end when the file ends. It ends when the last useful musical moment passes, which is often ten or fifteen seconds earlier. Everything after that is a fade nobody needs to hear.
Per-track markers let you say exactly where that point is.
Two markers per track
Intro end marks where the vocal starts — the point a presenter must talk up to and not past. Outro start marks where the next item should begin mixing in.
Both are stored in milliseconds on the track. You can stamp either from the current playback position while the song plays, which is the natural way to set them: listen, press, done.
They appear as thin vertical marks overlaid on the progress bar — green for intro end, orange for outro start — so an operator can see the shape of a track at a glance.
What the engine does with them
The outro marker is not decorative. The engine extends its preload lookahead so the handover lines up with OutroStartMs rather than the end of the file. The crossfade begins on the marker you set.
This is the difference between a station that mixes and one that merely fades. The tail of a track dissolving under the start of the next one sounds accidental. A mix that begins on a deliberate point sounds like a decision.
Detecting them automatically
Marking a 1,200-track library by hand is not realistic, so detection runs in layers:
- 20 ms RMS frames identify leading and trailing silence.
- An energy envelope pass refines the intro and outro positions.
- A peak-transient pass fills in tightly trimmed files that have no silence tails at all.
Each result carries a confidence score, plus a review flag when the analysis is uncertain. The method used is recorded on the track, so you can tell a hand-set marker from a detected one.
Run it across a multi-selection from the playlist, or across the whole library from Automation → Analyze Library Cue Points. Manual cues can be preserved rather than overwritten — the operator's ear outranks the algorithm, and the software should assume so.
Start with the ones that matter
You do not need every track marked. Start with your power rotation, where the same songs play most often, and let detection handle the long tail.