| Path | Content |
|---|---|
/var/lib/asterisk/sounds/custom/ |
sounds uploaded on Sounds (IVR, announcements, queues) |
/var/lib/asterisk/sounds/tr/ |
Turkish system prompts shipped in sounds/tr/ of the repository |
/var/lib/asterisk/moh/ |
music on hold |
/usr/share/asterisk/sounds/en/ |
English prompts (Debian package) |
Debian’s Asterisk searches prompts under /usr/share/asterisk/sounds. The installer links
/usr/local/share/asterisk/sounds (Debian’s custom directory) to /var/lib/asterisk/sounds/custom
and /usr/share/asterisk/sounds/tr to /var/lib/asterisk/sounds/tr. Without these links IVRs,
announcements and Turkish prompts fail with “File … does not exist in any format” and the call is hung
up. The links live outside dpkg-owned files, so package upgrades keep them.
Sounds should be 8 kHz mono WAV (16-bit PCM). Uploaded files with the same name as a shipped file are never overwritten by updates.
A call uses the language set on its inbound route. When a prompt is missing in that language Asterisk plays the English one, so a partly translated set gives mixed-language prompts.
The Turkish set ships at the repo root as one package per format,
asterisk-core-sounds-tr-<format>-<version>.tar.xz (with asterisk-core-sounds-tr-<version>.SHA256SUMS),
laid out like Asterisk’s own asterisk-core-sounds-* packages so it can be used on any Asterisk
server: extract the formats you need into the tr sounds directory.
tr-TR-Wavenet-C), generated with AiPBX Cloud TTS from 16 kHz masters: every
Asterisk core prompt (same names as asterisk-core-sounds-en, including voicemail, digits,
letters, phonetic, conference, queue and directory prompts) plus AiPBX’s own prompts.wav (8 kHz), ulaw, alaw, gsm and the wideband (HD) g722
and sln16. Asterisk picks the one matching the call’s codec, so no transcoding is needed.beep, ascending-2tone, …) come from the English set; silence/1..10 are generated.Installs and updates extract the newest package into /var/lib/asterisk/sounds/tr/, replacing the
prompts it contains (and removing other-format copies of the same names); other files in that
directory are left alone.
To change a text, edit sounds/core-sounds-tr.txt (name|text), regenerate the masters (Google
must be configured under AI → Cloud TTS) and rebuild the package:
sudo php web/bin/generate_tr_sounds.php --out=/tmp/tr-master # all prompts
sudo php web/bin/generate_tr_sounds.php --out=/tmp/tr-master vm-intro # or only some
scripts/build_tr_sounds.sh /tmp/tr-master 1.0.1
build_tr_sounds.sh needs every prompt’s master in the directory; to rebuild with only a few
changed prompts, first extract the current package’s .sln16 files back to WAV masters
(sox -t raw -r 16000 -e signed -b 16 -c 1 NAME.sln16 NAME.wav).
Asterisk has no Turkish number/date grammar, so SayNumber, dates and voicemail message counts
follow English word order; the texts are worded to stay natural in that order
(e.g. vm-youhave = “Posta kutunuzda”).