Open the repository's speech folder. Each language is a folder inside it
(en, hax, …) with a clips.tsv listing every clip the station needs.
Recordings are written to speech/<code>/recordings/<id>.wav; a replaced take
is kept in recordings/previous/.
Then run python3 -I tools/speech.py --lang <code> process and bank.
R or Space record / stop · P play · B play as broadcast · Enter keep and go to the next clip still to record · Esc discard · ↑ ↓ move between clips. Leave a moment of silence before and after speaking; the station trims it.