Phrase Studio (TTS)
Offline Piper-powered TTS studio for MikoPBX: generate IVR/voicemail phrases on the PBX, reuse them from a searchable history, and one-click drop the result into the standard Sound Files editor.
The Phrase Studio (TTS) module is an offline voice-phrase generator for MikoPBX, powered by the Piper TTS neural engine. All synthesis happens on the PBX itself — no caller text leaves the server, there are no per-character cloud charges, and there is no GDPR exposure for prompts and IVR scripts.
What it does
Inline TTS studio with text editor, voice picker, sample-rate switch and live audio preview.
Searchable generation history — click any row to repopulate the form with that phrase's text + voice and re-record it instantly.
25+ languages out of the box: English (multiple regional accents), Russian, German, French, Italian, Spanish, Polish, Portuguese, Czech, Ukrainian, Chinese and more.
Per-voice install / uninstall — pull only the languages you actually use; the rest stays off disk.
High-quality preview MP3 (22 kHz / 128 kbit/s) — no more "talking through a bucket" sound that the default 8 kHz telephony pipeline produces.
Filenames are auto-derived from the phrase text via Cyrillic→Latin transliteration (e.g. "Здравствуйте, спасибо за ваш звонок" →
Zdravstvujte_spasibo_za_vas_zvonok.wav).REST API v3 for scripted IVR build pipelines.
Bootstrapping the engine and voices
The module ships empty: the Piper binary and the voice models are downloaded on demand. The first two steps are normally automatic — enabling the module triggers a background download of the Piper engine and a refresh of the voice catalogue — so in the common case you can jump straight to step 2:
(Auto on enable, manual fallback) Open the Engine tab. The architecture-matched Piper binary (≈ 25 MB, statically linked) is downloaded and unpacked into
db/piper/automatically when the module is enabled. If the auto-bootstrap was skipped (no internet during enable, air-gapped install) or you want to re-pull the binary after a release, press Install engine — the same button is also exposed as Update engine and forces a re-download even if a working binary is already on disk.Switch to the Voices tab, select the languages you need and press Install next to each one. Models are 30–60 MB each and live in
db/voices/. The download runs asynchronously in a detached background process: the row immediately shows an Installing… spinner and flips to Installed (or Failed with a diagnostic message) once the download completes. You can navigate away or reload the page — progress is tracked on the database row, not in the browser, so the spinner picks up where it left off.Go back to the Studio tab and start synthesizing.
Inline use in the Sound Files editor
When you create or edit a custom sound file in MikoPBX, a "Generate from text (Phrase Studio)" segment appears right inside the standard form:
Reuse a previously generated phrase — pick an entry from the history dropdown; the source file in the form is replaced immediately and the player starts playing the new audio.
Generate and use as source — type new text, pick a voice and rate, and the result automatically becomes the source file of the current sound entry.
The high-quality MP3 plays in the standard MikoPBX player; the Asterisk-format derivatives (
ulaw,alaw,gsm,g722,sln) are produced at the rates Asterisk requires for telephony playback.
The block is intentionally hidden for MOH (music-on-hold) entries — Phrase Studio is for speech, not background music.
Quality knobs
Sample rate — Native (22 kHz) for studio-quality preview and HD codecs (g.722, opus); Telephony (8 kHz mono) for legacy g.711 trunks.
Maximum text length — defaults to 800 characters (≈ 60 seconds of speech). Longer phrases are rejected to protect the worker.
Cache size limit — defaults to 500 phrases; the oldest entries are pruned automatically.
REST API v3
All endpoints are auto-discovered via PHP 8 attributes and live under /pbxcore/api/v3/module-phrase-studio/. Authentication: localhost or Bearer token.
GET
engine
Engine status (installed, version)
POST
engine:install
Install the Piper binary. Pass {"force":true} in the body to force a re-download (Update flow)
DELETE
engine
Remove the Piper binary
GET
voices
Catalogue + installed flag per voice. Optional query: language=ru-ru, installed_only=true
POST
voices:install
Async. Queues a voice download and returns 202 Accepted with install_status="installing" — clients poll GET voices until the row flips to installed / failed
DELETE
voices/{id}
Remove a voice model
GET
phrases
Phrase history
POST
phrases
Generate (or return cached) phrase
GET
phrases/{id}:download
Stream the WAV (HEAD = duration probe)
POST
phrases/{id}:promoteToTmp
Stage WAV for Sound Files import
DELETE
phrases/{id}
Remove a single history entry
Voice rows returned by GET voices carry the async-install status fields the UI uses to drive its loader: install_status ("" / "installing" / "installed" / "failed") and install_error (set only when status is failed). A row stuck in installing for over five minutes is auto-flipped to failed with a synthetic timeout message, so a polling loop will eventually converge without manual cleanup.
Example — generate a phrase from a script:
Example — install a voice asynchronously and poll until ready:
Requirements
MikoPBX 2026.1.223+
PHP 8.4 on the module side
≈ 50 MB free disk for the engine plus a similar amount per installed voice
Outbound HTTPS access at first install (engine + voice download)
Module source code and issue tracker live on GitHub: mikopbx/ModulePhraseStudio. Share ideas and pull requests in our developer chat: @mikopbx_dev.
Last updated
Was this helpful?