> ## Documentation Index
> Fetch the complete documentation index at: https://docs.astradial.com/llms.txt
> Use this file to discover all available pages before exploring further.

# IVR greetings and TTS

> Generate IVR greetings from text using Google Cloud TTS — multiple languages, voice styles, and regeneration without re-recording.

Astradial uses Google Cloud Text-to-Speech to generate IVR greetings from the text you type. No studio recording needed.

## How it works

1. You enter text in the greeting node of the IVR builder.
2. Pick a language and voice.
3. Click **Generate greeting** — Astradial calls Google TTS and saves a `.wav` file to the server.
4. When a call hits the IVR, Asterisk plays this `.wav` as the greeting.

The filename and reference are stored on the IVR record in the database — no manual file management required.

## Supported languages

Any Google Cloud TTS language code works. Common ones:

| Language        | Code    | Example voice      |
| --------------- | ------- | ------------------ |
| English (India) | `en-IN` | `en-IN-Wavenet-D`  |
| English (US)    | `en-US` | `en-US-Wavenet-D`  |
| English (UK)    | `en-GB` | `en-GB-Wavenet-A`  |
| Hindi (India)   | `hi-IN` | `hi-IN-Wavenet-A`  |
| Tamil (India)   | `ta-IN` | `ta-IN-Wavenet-A`  |
| Telugu (India)  | `te-IN` | `te-IN-Standard-A` |

<Note>
  The full list of Google TTS voices is at [cloud.google.com/text-to-speech/docs/voices](https://cloud.google.com/text-to-speech/docs/voices). Any voice name listed there is supported.
</Note>

## Voice types

* **Wavenet** voices (`*-Wavenet-*`) — highest quality, neural TTS. Recommended for production.
* **Standard** voices (`*-Standard-*`) — lower quality but faster and cheaper. Good for testing.
* **Neural2** voices — Google's newer TTS. Supported if the voice name is correct.

## Writing good IVR text

### Keep it short

Aim for under 15 seconds of audio. That's about 40 words at a normal speaking pace.

✅ "Welcome to Acme. Press 1 for sales, 2 for support, 3 for billing."

❌ "Hello and thank you for calling Acme Corporation, the leading provider of... \[30 seconds of options]"

### Speak the menu structure explicitly

Callers can't see your UI. Read every option out loud.

✅ "Press 1 for sales, 2 for support, or press star to repeat this menu."

❌ "Please select from the following options." (What options?)

### Pronunciation hints

Google TTS handles most text well. For tricky words:

* **Acronyms**: Space out letters: `A B N` instead of `ABN`.
* **Phone numbers**: Spell out the country code: "plus nine one" for +91.
* **Proper nouns**: If the TTS mispronounces a name, try phonetic spelling: "Rahool" instead of "Rahul".

## Previewing before saving

Click **Preview** next to the voice dropdown to hear the current text spoken. This plays in your browser and does NOT save the audio to the server.

Useful for testing phrasing and voice choice before the production generate call.

## Regenerating

You can regenerate any time by editing the text/voice/language and clicking **Generate greeting** again. The new `.wav` overwrites the old one.

<Warning>
  Regenerating does NOT automatically redeploy the dialplan. If your IVR is already live, click **Publish** after regenerating so the new greeting is picked up by Asterisk.
</Warning>

## Fallback if TTS fails

If Google TTS returns an error (quota, invalid voice name, network issue), the IVR record is still updated but no `.wav` is written. Callers will hear Asterisk's default "welcome" sound.

Fix: correct the language/voice/text and click **Generate greeting** again.

## Generating greetings via API

For scripted provisioning — see the [IVR API reference](/api-reference/ivrs#generate-greeting).