Cloud TTS (Plugin)
Plugin (
ai-tts): Optional plugin from the official catalog. Enable it under Integrations → Plugins (see Plugin system & developer guide).
AI → Cloud TTS (admin by default) turns text into speech with a cloud text-to-speech service. The result can be played on the page, downloaded as MP3, or saved straight into Sounds for IVRs, queues and time conditions — no recording studio, and every announcement in the same voice.
Providers
Set up one or more providers on the Providers tab:
| Provider | Credentials |
|---|---|
| Google Cloud Text-to-Speech | API key, or a service account JSON key (for organisations that forbid API keys) |
| Amazon Polly | access key ID + secret and region; give the IAM user only polly:SynthesizeSpeech and polly:DescribeVoices |
| Microsoft Azure AI Speech | key and region of a Speech resource |
| ElevenLabs | API key; voices come from your account's voice library |
| OpenAI | API key; fixed voices that speak every language |
Test connection checks the credentials. The voices are listed live from each provider (cached for 12 hours). When a request fails, the provider's own error message is shown.
Credentials are encrypted (libsodium) with a key kept outside the database — AIPBX_SETTINGS_KEY in the environment or /var/lib/aipbx/settings.key, created by install.sh — and are never sent back to the browser; the page only shows that a key is set (for a Google service account: its e-mail). AiPBX calls the providers' REST APIs directly; no vendor SDK is installed.
Synthesize
On the Synthesize tab choose provider, language, voice and speed and enter the text (up to 20,000 characters; long texts are split at sentence ends to fit each provider's limit). Then:
- Listen on the page,
- Download the MP3,
- Save as announcement — the audio is converted to 8 kHz mono WAV in the custom sounds folder and appears in Sounds, marked for Apply like an uploaded file.
The History lists every synthesis with its character count (the services bill per character), duration, who made it and the announcement it became. Syntheses and settings changes are written to the audit log.
Turkish system prompts
AiPBX's complete Turkish prompt set was generated with the Google provider configured here (voice tr-TR-Wavenet-C, 16 kHz masters) and ships as the asterisk-core-sounds-tr packages. To change or add a prompt, see Sounds & languages.