An outcue is the last few spoken words of a piece of audio — the verbal cue a live broadcast host listens for so they know when to come back on the mic. Librarian can transcribe those words automatically and store them in a new field on each song record, so AutoCast can display the cue right on the deck row during the show.
Right-click any folder in the Station Folders sidebar and choose Detect outcues. Librarian walks every audio file in that folder, reads the last 15 seconds, and runs Apple's on-device speech recognizer against it. The recognizer's transcription gets trimmed to its last six words and written to a new outcue column in your library database. AutoCast then renders that string as a small Q: line next to the start time on each deck row.
Recognition is fully on-device. Nothing leaves your Mac. The audio doesn't get uploaded anywhere; the recognizer's language model is built into macOS itself.
Not every spot ends with a spoken cue. PSAs sometimes fade out on music; sweepers might end on a sound-design flourish; some commercials trail off into a long musical tag. When Librarian's recognizer can't pull usable words out of the last few seconds, it writes the literal text [INDISTINCT] into the outcue field, and AutoCast renders that as Q: Indistinct (or music out) on the deck row. That tells the host: "no verbal cue available here — listen for the music to fade and watch the clock."
The right-click option is a toggle. The first time you click it on a folder, Librarian starts monitoring that folder and immediately scans every file in it. From then on, the menu item reads Stop monitoring for outcues when you right-click that folder, so you can turn the monitoring off the same way you turned it on.
Your monitored-folder set is remembered between launches — you don't have to re-enable it every time you open Librarian. The monitoring is per-folder, not recursive: choosing it on PSAs watches the immediate PSAs folder, not any sub-folders inside it.
The first time you choose Detect outcues on any folder, macOS will ask permission for Librarian to use speech recognition. Click OK. You'll only see this prompt once, ever.
On-device speech recognition is good but not perfect. Names, brand URLs, and unusual words sometimes come out slightly off (you might see "Forster" instead of "Forester," or "rady.gov" where the announcer said "ready.gov"). When you spot a mis-hear, just edit the outcue field in the Attribute Editor like any other attribute, and click Apply.
Librarian remembers that you edited it and will not overwrite your correction on subsequent automatic re-scans — even when monitoring is enabled. Your correction sticks.
The one exception: if the underlying audio file itself gets replaced or re-recorded (for example, the same PSA now has a different read), Librarian notices the change and re-runs the recognizer, on the assumption that the closing words may have changed too. Your previous correction is overwritten by the new transcription, since the script it described is no longer in the file.
AutoCast subscribes to your library and pulls in the outcue field automatically. On each deck row, when a file has a non-empty outcue, AutoCast draws a small Q: line next to the playback time:
Q: "the sale will run from 8 AM to 5 PM" — with curly quotes around the recognized words.[INDISTINCT] render as Q: Indistinct (or music out) — no quotes, signaling "no verbal cue, watch the music."Q: line at all.When the cue is wider than the deck-row space available, AutoCast truncates from the front of the cue (with a leading ellipsis), preserving the final words — those are the actual cue the host listens for. Six words usually fits cleanly without truncation.
outcue field in the Attribute Editor. Your edit will stick.