Audio, voicemail, and text-to-speech in VICIdial explained

Sooner or later every VICIdial system needs to say something on its own. A caller reaches a number when no agent is free, a campaign plays a greeting before the queue, or a missed call should land in a mailbox someone checks later. All of that runs on three connected pieces: voicemail boxes, the audio store, and text-to-speech. They overlap enough that newcomers tend to wire one up, hit a wall on another, and never see how they fit together.

This is the hub for that whole landscape. We will walk through how voicemail boxes are created and tuned, where audio prompts live and what formats they have to be in, and where text-to-speech slots in when you would rather type a sentence than record one. Each section links out to a focused walkthrough, so treat this page as the map and the linked posts as the turn-by-turn directions.

A quick note on scope before we start. None of this is about the recordings your agents make on live calls — that is a separate subsystem with its own retention rules. What follows is purely about the audio VICIdial plays out and the messages callers leave behind. Keep that distinction in mind and the admin screens make a lot more sense, because the same word, recording, gets used for two completely different things across the system.

Two kinds of voicemail box

VICIdial has voicemail boxes you never asked for and voicemail boxes you create on purpose. The first kind is automatic: every entry in the Phones admin section gets its own box so an extension can take messages. We break those down in how per-phone voicemail boxes work. The second kind is the general voicemail box, which you add by hand in the admin Voicemail section.

General boxes are the useful ones operationally. The most common job is acting as a message drop for calls into one specific DID (direct inward dialing) or Call menu, so an inbound caller can leave details when nobody picks up. They also cover the case where an Agent is simply unavailable and you want the customer to leave contact information for follow-up. We walk through both patterns in pointing a DID or call menu at a voicemail box and in taking a message when no agent is free.

Creating and editing a box

Adding a general box is short on purpose. The add screen only asks for the handful of fields a valid record needs, and the rest appear once you save and reopen the entry to modify it. The full step-by-step is in adding a voicemail box, and editing an existing one is covered in modifying a voicemail box.

Two rules trip people up at creation time. The voicemail ID has to be all numbers, at least two characters, and must not collide with an existing box or with the ID of a phone already on the system. And the box will not save at all unless you set both the ID and a password. We unpack both in the voicemail ID and password rules. The password, by the way, is what you key in when you dial the box to check messages, and it tops out at ten characters.

The fields that change behavior

Past the ID, password, and name, the modify screen is a row of optional fields, and each one changes how the box behaves. Here is the short tour, with a deeper post behind each.

• Active — toggles whether the box accepts and plays back messages at all. An inactive box silently drops on the floor. See the active setting.
• Email — optional address that each new message is mailed to, provided the system is set up to send mail. Covered in voicemail-to-email.
• Delete Voicemail After Email — removes the message from the server once it has been mailed out, default N. See delete-after-email.
• Show VM on Summary Screen — surfaces the box on the admin login summary, with new, old, and total counts. The table only appears once at least one box is set to Y. See showing voicemail on the summary screen.
• Voicemail Zone — the time zone stamped on each message. Default comes from System Settings. See the voicemail zone setting.
• Voicemail Options — an advanced field best left blank unless you know exactly what you are doing. See the voicemail options field.
• Admin User Group — scopes who can view this record, default --ALL--. See the admin user group.

There is also a greeting picker. The Voicemail Greeting field lets you point the box at an audio file from the audio store instead of the default beep, but it only appears once Allow Voicemail Greeting Chooser is enabled in System Settings first. We split that into enabling the greeting chooser and choosing a greeting so you can flip the gate and then use it.

How an inbound call lands in a box

It helps to see the path a call takes before it ever leaves a message. An inbound number routes through a DID or a call menu, the system checks whether an agent is available, and only when none is does the caller get handed to the voicemail box you attached. If that box is inactive, the message never records.

flowchart TD
  A["Inbound call to DID"] --> B["Call menu or DID route"]
  B --> C{"Agent available?"}
  C -->|"Yes"| D["Connect to agent"]
  C -->|"No"| E{"Voicemail box active?"}
  E -->|"No"| F["Call dropped"]
  E -->|"Yes"| G["Play greeting"]
  G --> H["Record message"]
  H --> I{"Email set?"}
  I -->|"Yes"| J["Mail message out"]
  I -->|"No"| K["Store on server"]

A related trick is the general message drop: rather than recording a fresh message, you can drop a pre-recorded one into a flow, which is handy when an Auto attendant or campaign needs a consistent outbound greeting. That pattern is in the general message-drop box, and it is a close cousin of Voicemail drop on the outbound side.

Checking and counting messages

Every box tracks two numbers that live on the Asterisk server itself: new messages and old messages. New is what has arrived and not been listened to, old is what someone has already heard. The split matters because it drives notifications and the summary counts, and we explain it in new versus old messages.

To actually hear messages, you dial the box from a Hardphone (deskphone) or Softphone registered on the system and enter the password. The full dial-in procedure is in checking voicemail by phone. If you would rather not pick up a handset at all, the email and summary-screen options above let you read counts without dialing in.

One detail that catches people out: those counts live on the Asterisk server, not in the web database, so they reflect the real state of the mailbox on disk rather than a cached number. If a box shows new messages but the summary screen looks empty, the usual culprit is that Show VM on Summary Screen is still set to N, because that table will not render at all until at least one box opts in. Set one box to Y and the rest of your flagged boxes appear alongside it.

The audio store

Greetings, prompts, and hold music all come from one place: the audio store. It is a central upload point on the web server that pushes your files out to every VICIdial server, so in a multi-server cluster you upload once and the prompt appears everywhere within a couple of minutes. The audio chooser links scattered across call menus, in-groups, Music on hold entries, and campaign fields all read from this same store.

If your install does not show an Audio Store link yet, it needs turning on in System Settings first — Central Sound Control Active, the sounds web server, and the active voicemail server. We cover the difference between the concept and the activation in what the audio store is and setting up the audio store. Uploading a file with a name that already exists simply replaces the old version, which is the cleanest way to update a prompt in place.

Recording a prompt does not require any upload at all. From a registered phone you can dial 8168, follow the prompts, and the recording lands in the store within a couple of minutes. The actual upload-from-computer flow is in uploading an audio prompt. Here is roughly how a file becomes a usable prompt.

flowchart LR
  A["Source audio"] --> B["Convert to allowed format"]
  B --> C["Upload to audio store"]
  C --> D{"Format valid?"}
  D -->|"No"| E["Red warning, re-format"]
  E --> B
  D -->|"Yes"| F["File copied to all servers"]
  F --> G["Pick via audio chooser"]

File formats and converting audio

This is the single most common stumbling block, so it earns its own section. The audio store accepts exactly two formats, and anything else is rejected at upload:

• .wav files that are PCM mono, 16-bit, 8000 Hz (one channel, uncompressed)
• .gsm files that are 8-bit, 8000 Hz

After a .wav upload the page validates the format and shows a red warning if it is wrong, at which point you reformat and try again. The why behind these two formats, including how they relate to the G.711 codec telephony Codec, is in the allowed file formats and GSM versus WAV. When your source is a stereo 44.1 kHz file from a recording app, you have to down-convert it first, and we show the exact ffmpeg and sox commands in converting audio for VICIdial.

None of this needs to slow down getting started. If you are evaluating VICIdial and would rather skip the install entirely, a managed box on Managed hosting arrives with the audio store already switched on — see pricing to spin one up in under 40 seconds and start uploading prompts the same minute.

Where text-to-speech fits

Sometimes you do not want to record or upload anything — you want VICIdial to speak a sentence you typed. That is what the Text-to-Speech section is for. Each TTS entry has an ID, a name, a voice (default Allison-8kHz), and the text that gets synthesized into an audio file and played to the customer.

The text field is more capable than it looks. It accepts Speech Synthesis Markup Language, so a tag like a one-second break drops a pause exactly where you want one, and it accepts lead variables in the same style you use in an Agent script — first name, last name, title, and many others — so the prompt can greet each caller by name. The catch is that TTS (text to speech) relies on Cepstral being installed and licensed on every server that uses it, plus the TTS option enabled in System Settings. So the choice is usually: record or upload a fixed prompt through the audio store, or type a dynamic one through TTS when you have Cepstral in place.

Putting it together

The pattern that ties all of this together is simple once you see it. Voicemail boxes decide what happens when nobody answers. The audio store decides what callers hear, whether that is a greeting on a box, a prompt in a call menu, or a track in your hold music. And text-to-speech is the shortcut for when typing beats recording. Get the file formats right, gate the greeting chooser before you reach for it, and keep an eye on which boxes are active, and the awkward edge cases mostly disappear.

From here, pick the spoke that matches what you are doing right now — adding a box, wiring email, fixing a rejected upload, or standing up TTS — and come back to this page when you need the next piece. If you would rather have the whole stack pre-configured instead of assembling it by hand, a managed VICIdial server ships with voicemail, the audio store, and System Settings already sorted; check the pricing and you can be uploading your first greeting in under a minute.