Text to Speech

Overview

Vonage uses text-to-speech engines to allow you to play machine generated speech to your users. This can either be done via an NCCO with the use of the talk action, or by making a PUT request to an in-progress call.

Example

The following NCCO demonstrates an example use case:

[
  {
    "action": "talk",
    "text": "Thank you for calling. Please leave your message after the tone."
  }
]

Voice Options

Vonage offers a wide range of options for text-to-speech categorized into three tiers based on quality (from basic to high-end):

• Standard
• Premium
• Premier

Premier Voices

Vonage Premier Voices are based on Generative AI.

Currently, Premier Voices support google (Chirp3) and deepgram (Aura) as providers. The providerOptions object should include your selected options for the chosen provider. See the sections below for details.

Google

Vonage supports Google as a Premier TTS provider. Use google as the provider value.

The providerOptions object should include your selected Chirp3 options, such as the desired language, among other customization options. The following NCCO demonstrates an example of how to use the talk action with Google Chirp3:

[
  {
    "action": "talk",
    "text": "Thank you for calling. Please leave your message and we’ll get back to you as soon as we can",
    "provider": "google",
    "providerOptions": {
      "name": "en-US-Chirp3-HD-Charon",
      "language_code": "en-US"
    }
  }
]

Deepgram

Vonage supports Deepgram Aura (1 & 2) as a Premier TTS provider. Use deepgram as the provider value.

The providerOptions object should include your selected Deepgram Aura options, such as the desired voice model, among other customization options. It supports the following keys:

Any additional keys you supply are forwarded to Deepgram verbatim.

The following NCCO demonstrates an example of how to use the talk action with Deepgram Aura:

[
  {
    "action": "talk",
    "text": "Hello, thank you for calling. How can I help you today?",
    "bargeIn": false,
    "provider": "deepgram",
    "providerOptions": {
      "model": "aura-2-thalia-en",
      "speed": 1.0
    }
  },
  {
    "action": "input",
    "type": ["speech"],
    "eventUrl": ["https://example.com/input"]
  }
]

Standard and Premium

Our Standard and Premium offerings remain available using the syntax described in the following sections.

Locale

You should choose a locale that matches the language type of the text you are asking to be read, trying to read Spanish with an en-us voice for example will not produce good results.

You can set the language code (BCP-47) with a language parameter in the talk command, if you do not specify a language then Vonage will default to an en-US voice. For many languages Vonage offers a choice of voices with different styles. The style parameter maps to features such as vocal range, timbre and tessitura of the selected voice. You can experiment with different styles to find the one appropriate for your use case. You can choose a specific voice style by using the style parameter. By default, the first style (0) is used. Please note not all the voice styles support SSML, see the list of all the supported languages and SSML enabled styles below.

NCCO example with specific voice language and style:

[
  {
    "action": "talk",
    "text": "Obrigado pela sua chamada. Por favor, deixe sua mensagem após o sinal.",
    "language": "pt-PT",
    "style": 6
  }
]

Premium Voices

Some voice styles come with a premium alternative, which through the use of AI, have a more natural sound. To use a premium voice style, add the premium option in your NCCO:

[
  {
    "action": "talk",
    "text": "Obrigado pela sua chamada. Por favor, deixe sua mensagem após o sinal.",
    "language": "pt-PT",
    "style": 6,
    "premium": true
  }
]

Supported Languages

Download the list of voices as a JSON file.