Generate audio prompts from text

The Generate with TTS button on the Audios tab turns any text into a VICIdial-compatible audio prompt (8 kHz mono WAV) using Microsoft Azure neural voices. The generated file lands straight into your VICIdial audio library — same place uploads land — and is ready to wire into a campaign, in-group, or call menu.

How to use it

1. Open your server's Audios tab and click Generate with TTS.

2. Pick a voice from the dropdown. Tap the ▶ play icon to hear a sample — sample playback is free, no charge.

3. Type or paste your text (up to 2,000 characters, around 2 minutes of audio).

4. Edit the suggested filename if you want — it auto-fills from voice + timestamp.

5. Watch the cost tile update as you type. It tells you exactly what you'll be charged before you commit.

6. Click Generate & save. The platform debits your wallet, calls Azure, and lands the WAV in your audio library. The whole flow takes a second or two.

What it costs

You're billed per character of text. The exact rate is set by your platform administrator (or your reseller, if you're a reseller customer). At the default rate of $25 per million characters, a typical 300-character greeting costs less than a cent. The cost tile in the modal always shows the real number for the active rate — there's no hidden markup.

What about the "Discard" button?

After generation you can hit Discard without saving if you don't like the result. That removes the file from your server, but the wallet charge stays. Reason: Azure has already done the work and billed us by the time you hear the audio. Discarding is a cleanup convenience, not a refund.

To avoid wasted charges, use the free ▶ sample button to audition voices first.

Limits and rules

• 2,000 characters per generation. If you have more text, split it into multiple prompts (call menus / IVR usually want short snippets anyway).

• Filenames must be unique. If you pick a name that already exists in your audio store, the generation fails with a "name collision" error. The charge still applies — Azure already synthesised the audio. Pick a different name and try again.

• Wallet must cover the cost. If your balance is short, the modal swaps the primary button for an "Add funds to wallet" link.

• WAV output is 8 kHz mono PCM — exactly what VICIdial's playback chain expects. No transcoding hassles on your side.

Where the file ends up

Same place every other audio prompt lives: VICIdial's audio_store_details table, and on disk under /var/www/html/<sounds_dir>/. Asterisk picks it up automatically on its next central-control sync (typically subsecond). You can reference the filename in any campaign / in-group / call menu the same way you'd reference an uploaded prompt.

What if Azure or our system has a problem?

If the synthesis call fails for any reason before the file lands, the charge is automatically refunded to your wallet (the refund shows as an "Adjustment" entry in your wallet history). The only path where a charge sticks without you keeping a file is the "name collision" case described above.