Getting Started with Speech Synthesis Markup Language (SSML)

SSML was designed by W3C to provide an XML-based markup language to assist in generating natural sounding synthesized speech. Plivo <Speak> XML now supports the generation of SSML based speech. SSML speech generation on Plivo is powered by Amazon Polly, the leader in SSML speech synthesis.

With normal text-to-speech, developers can only choose from a basic male or female voice in a subset of languages. Plivo SSML supports 27 languages and over 40 voices, and allows developers to also control pronunciation, pitch, volume, etc. Plivo’s root XML element for SSML tags is <Speak>, same as that for basic TTS. For example:

1
2
3
4
5
6
<Response>
    <Speak voice="MAN">Go Green, Go Plivo</Speak> //Basic Text-to-Speech
    <Speak voice="Polly.Joey">
        <emphasis level="moderate">Go Green, Go Plivo</emphasis> //Text-to-Speech using SSML
    </Speak>
</Response>

Amazon Polly voices can process text-to-speech for a maximum of 3000 characters using the <Speak> tag. For more information about SSML, see the W3C specifications.

Amazon Polly

Amazon Polly is a service that provides life-like text-to-speech across several languages and locales. SSML support on Plivo is powered by Amazon Polly.

To synthesize SSML speech on Plivo, simply specify one of the many Amazon Polly voices in the ‘voice’ attribute of Plivo’s <Speak> XML. Note that Polly voices must be namespaced with Polly..

For example:

1
2
3
4
5
<Response>
    <Speak voice="Polly.Joey">
        <emphasis level="moderate">Go Green, Go Plivo</emphasis>
    </Speak>
</Response>

A complete list of supported Polly voices is available here.

SSML Tags

The following SSML tags are supported for use in Plivo’s XML:

Note: The following AWS Polly specific tags are not supported for use with Plivo XML:

• <amazon:auto-breaths>
• <amazon:effect name=”drc”>
• <amazon:effect phonation=”soft”>
• <amazon:effect vocal-tract-length>
• <amazon: effect name=”whispered”>

SSML Voices

The SSML Voices are supported for use with Plivo XML:

Character Limit

To ensure quick synthesis, an upper cap of 3000 characters is enforced on the text that can be synthesized in one <Speak> XML.

Pricing

Support for SSML based speech synthesis is currently in Beta. While in Beta, SSML based speech synthesis is absolutely free.

SSML based speech synthesis will eventually be charged on the basis of the number of characters synthesized.

SSML Support In Plivo Server-Side SDKs

SSML tags are supported in all our Server-side SDKs, you can get started with our SDKs by checking the setup guides available here for your preferred language.

Examples

The below examples use the Joey voice for US English (en-US). Use the <Speak voice> tag to specify the voice for your text.

• Say-as

The say-as tag describes how to interpret the text.

1
2
3
4
5
6
<Response>
    <Speak voice="Polly.Joey">
        The date is
        <say-as interpret-as="date">20180626</say-as>
    </Speak>
</Response>

• W

The w tag is used to customize the pronunciation of words by specifying the part of speech.

1
2
3
4
5
6
7
8
9
10
11
12
<Response>
    <Speak voice="Polly.Joey">
    The word
    <say-as interpret-as="characters">read</say-as>
    <s>
        may be interpreted as either the present simple form
    </s>
    <w role="amazon:VB">read</w>
    <s>or the past participle form</s>
    <w role="amazon:VBD">read</w>
    </Speak>
</Response>