Speech Synthesis Markup Language (SSML)

You can send Speech Synthesis Markup Language (SSML) in your Text-to-Speech request to allow for more customization in your audio response by providing details on pauses, and audio formatting for acronyms, dates, times, abbreviations, or text that should be censored. See the Text-to-Speech SSML tutorial for more information and code samples.

The following shows an example of SSML markup and the Text-to-Speech synthesizes the text:

<speak>
  Here are <say-as interpret-as="characters">SSML</say-as> samples.
  I can pause <break time="3s"/>.
  I can play a sound
  <audio src="https://www.example.com/MY_MP3_FILE.mp3">didn't get your MP3 audio file</audio>.
  I can speak in cardinals. Your number is <say-as interpret-as="cardinal">10</say-as>.
  Or I can speak in ordinals. You are <say-as interpret-as="ordinal">10</say-as> in line.
  Or I can even speak in digits. The digits for ten are <say-as interpret-as="characters">10</say-as>.
  I can also substitute phrases, like the <sub alias="World Wide Web Consortium">W3C</sub>.
  Finally, I can speak a paragraph with two sentences.
  <p><s>This is sentence one.</s><s>This is sentence two.</s></p>
</speak>

Here is the synthesized text for the example SSML document:

Here are S S M L samples. I can pause [3 second pause]. I can play a sound [audio file plays].
I can speak in cardinals. Your number is ten.
Or I can speak in ordinals. You are tenth in line.
Or I can even speak in digits. The digits for ten are one oh.
I can also substitute phrases, like the World Wide Web Consortium.
Finally, I can speak a paragraph with two sentences. This is sentence one. This is sentence two.

The Text-to-Speech supports a subset of the available SSML tags, which are described in this topic.

For more information about how to create audio data from SSML input with the Text-to-Speech, see Creating Voice Audio Files.

Tips for using SSML

Depending on your implementation, you may need to escape quotation marks or quotes in the SSML payload that you send to Text-to-Speech. The following example shows how to format SSML input included within a JSON object.

"{
    'input':{
     'ssml':'<speak>The <say-as interpret-as=\"characters\">SSML</say-as>
          standard <break time=\"1s\"/>is defined by the
          <sub alias=\"World Wide Web Consortium\">W3C</sub>.</speak>'
    },
    'voice':{
      'languageCode':'en-us',
      'name':'en-US-Standard-B',
      'ssmlGender':'MALE'
    },
    'audioConfig':{
      'audioEncoding':'MP3'
    }
  }"

Reserve characters

Avoid using SSML reserve characters in the text that is to be converted to audio. When you need to use an SSML reserve character, prevent the character from being read as code by using its escape code. The following table shows reserved SSML characters and their associated escape codes.

Select a voice

You can set the voice in the VoiceSelectionParams object. See the Text-to-Speech SSML tutorial to see a code sample demonstrating use of the VoiceSelectionParams object.

You can use the <voice> tag to read SSML in multiple voices, but you must set the VoiceSelectionParams name to a compatible voice:

Support for SSML elements

The following sections describe the SSML elements and options that can be used in your Actions.

<speak>

The root element of the SSML response.

To learn more about the speak element, see the W3 specification.

Example

<speak>
  my SSML content
</speak>

<break>

An empty element that controls pausing or other prosodic boundaries between words. Using <break> between any pair of tokens is optional. If this element is not present between words, the break is automatically determined based on the linguistic context.

To learn more about the break element, see the W3 specification.

Attributes

Example

The following example shows how to use the <break> element to pause between steps:

<speak>
  Step 1, take a deep breath. <break time="200ms"/>
  Step 2, exhale.
  Step 3, take a deep breath again. <break strength="weak"/>
  Step 4, exhale.
</speak>

<say‑as>

This element lets you indicate information about the type of text construct that is contained within the element. It also helps specify the level of detail for rendering the contained text.

The <say‑as> element has the required attribute, interpret-as, which determines how the value is spoken. Optional attributes format and detail may be used depending on the particular interpret-as value.

Examples

The interpret-as attribute supports the following values:

• currency

  The following example is spoken as "forty two dollars and one cent". If the language attribute is omitted, it uses the current locale.

  <speak>
    <say-as interpret-as='currency' language='en-US'>$42.01</say-as>
  </speak>
      

• telephone

  See the interpret-as='telephone' description in the W3C SSML 1.0 say-as attribute values WG note.

  The following example is spoken as "one eight zero zero two zero two one two one two". If the "google:style" attribute is omitted, it speaks zero as letter O.

  The "google:style='zero-as-zero'" attribute currently only works in EN locales.

        <speak>
          <say-as interpret-as='telephone' google:style='zero-as-zero'>1800-202-1212</say-as>
        </speak>
      

• verbatim or spell-out

  The following example is spelled out letter by letter:

  <speak>
    <say-as interpret-as="verbatim">abcdefg</say-as>
  </speak>
      

• date

  The format attribute is a sequence of date field character codes. Supported field character codes in format are {y, m, d} for year, month, and day (of the month) respectively. If the field code appears once for year, month, or day then the number of digits expected are 4, 2, and 2 respectively. If the field code is repeated then the number of expected digits is the number of times the code is repeated. Fields in the date text may be separated by punctuation and/or spaces.

  The detail attribute controls the spoken form of the date. For detail='1' only the day fields and one of month or year fields are required, although both may be supplied. This is the default when less than all three fields are given. The spoken form is "The {ordinal day} of {month}, {year}".

  The following example is spoken as "The tenth of September, nineteen sixty":

  <speak>
    <say-as interpret-as="date" format="yyyymmdd" detail="1">
      1960-09-10
    </say-as>
  </speak>
      

  The following example is spoken as "The tenth of September":

  <speak>
    <say-as interpret-as="date" format="dm">10-9</say-as>
  </speak>
      

  For detail='2' the day, month, and year fields are required and this is the default when all three fields are supplied. The spoken form is "{month} {ordinal day}, {year}".

  The following example is spoken as "September tenth, nineteen sixty":

  <speak>
    <say-as interpret-as="date" format="dmy" detail="2">
      10-9-1960
    </say-as>
  </speak>
      

• characters

  The following example is spoken as "C A N":

  <speak>
    <say-as interpret-as="characters">can</say-as>
  </speak>
      

• cardinal

  The following example is spoken as "Twelve thousand three hundred forty five" (for US English) or "Twelve thousand three hundred and forty five (for UK English)":

  <speak>
    <say-as interpret-as="cardinal">12345</say-as>
  </speak>
      

• ordinal

  The following example is spoken as "First":

  <speak>
    <say-as interpret-as="ordinal">1</say-as>
  </speak>
      

• fraction

  The following example is spoken as "five and a half":

  <speak>
    <say-as interpret-as="fraction">5+1/2</say-as>
  </speak>
      

• expletive or bleep

  The following example comes out as a beep, as though it has been censored:

  <speak>
    <say-as interpret-as="expletive">censor this</say-as>
  </speak>
      

• unit

  Converts units to singular or plural depending on the number. The following example is spoken as "10 feet":

  <speak>
    <say-as interpret-as="unit">10 foot</say-as>
  </speak>
      

• time

  The following example is spoken as "Two thirty P.M.":

  <speak>
    <say-as interpret-as="time" format="hms12">2:30pm</say-as>
  </speak>
      

  The format attribute is a sequence of time field character codes. Supported field character codes in format are {h,m, s, Z, 12, 24} for hour, minute (of the hour), second (of the minute), time zone, 12-hour time, and 24-hour time respectively. If the field code appears once for hour, minute, or second then the number of digits expected are 1, 2, and 2 respectively. If the field code is repeated then the number of expected digits is the number of times the code is repeated. Fields in the time text may be separated by punctuation and/or spaces. If hour, minute, or second are not specified in the format or there are no matching digits then the field is treated as a zero value. The default format is "hms12".

  The detail attribute controls whether the spoken form of the time is 12-hour time or 24-hour time. The spoken form is 24-hour time if detail='1' or if detail is omitted and the format of the time is 24-hour time. The spoken form is 12-hour time if detail='2' or if detail is omitted and the format of the time is 12-hour time.

To learn more about the say-as element, see the W3 specification.

<audio>

Supports the insertion of recorded audio files and the insertion of other audio formats in conjunction with synthesized speech output.

Attributes

The following are the currently supported settings for audio:

• Format: MP3 (MPEG v2)
  • 24K samples per second
  • 24K ~ 96K bits per second, fixed rate
• Format: Opus in Ogg
  • 24K samples per second (super-wideband)
  • 24K - 96K bits per second, fixed rate
• Format (deprecated): WAV (RIFF)
  • PCM 16-bit signed, little endian
  • 24K samples per second
• For all formats:
  • Single channel is preferred, but stereo is acceptable.
  • 240 seconds maximum duration. If you want to play audio with a longer duration, consider implementing a media response.
  • 5 megabyte file size limit.
  • Source URL must use HTTPS protocol.
  • Our UserAgent when fetching the audio is "Google-Speech-Actions".

The contents of the <audio> element are optional and are used if the audio file cannot be played or if the output device does not support audio. The contents may include a <desc> element in which case the text contents of that element are used for display. For more information, see the Recorded Audio section in the Responses Checklist.

The src URL must also be an https URL (Google Cloud Storage can host your audio files on an https URL).

To learn more about media responses, see the media response section in the Responses guide.

To learn more about the audio element, see the W3 specification.

Example

<speak>
  <audio src="cat_purr_close.ogg">
    <desc>a cat purring</desc>
    PURR (sound didn't load)
  </audio>
</speak>

<p>,<s>

Sentence and paragraph elements.

To learn more about the p and s elements, see the W3 specification.

Example

<p><s>This is sentence one.</s><s>This is sentence two.</s></p>

Best practices

• Use <s>...</s> tags to wrap full sentences, especially if they contain SSML elements that change prosody (that is, <audio>, <break>, <emphasis>, <par>, <prosody>, <say-as>, <seq>, and <sub>).
• If a break in speech is intended to be long enough that you can hear it, use <s>...</s> tags and put that break between sentences.

<sub>

Indicate that the text in the alias attribute value replaces the contained text for pronunciation.

You can also use the sub element to provide a simplified pronunciation of a difficult-to-read word. The last example below demonstrates this use case in Japanese.

To learn more about the sub element, see the W3 specification.

Examples

<sub alias="World Wide Web Consortium">W3C</sub>

<sub alias="にっぽんばし">日本橋</sub>

<mark>

An empty element that places a marker into the text or tag sequence. It can be used to reference a specific location in the sequence or to insert a marker into an output stream for asynchronous notification.

To learn more about the mark element, see the W3 specification.

Example

<speak>
Go from <mark name="here"/> here, to <mark name="there"/> there!
</speak>

<prosody>

Used to customize the pitch, speaking rate, and volume of text contained by the element. Currently the rate, pitch, and volume attributes are supported.

The rate and volume attributes can be set according to the W3 specifications. There are three options for setting the value of the pitch attribute:

To learn more about the prosody element, see the W3 specification.

Example

The following example uses the <prosody> element to speak slowly at 2 semitones lower than normal:

<prosody rate="slow" pitch="-2st">Can you hear me now?</prosody>

<emphasis>

Used to add or remove emphasis from text contained by the element. The <emphasis> element modifies speech similarly to <prosody>, but without the need to set individual speech attributes.

This element supports an optional "level" attribute with the following valid values:

• strong
• moderate
• none
• reduced

To learn more about the emphasis element, see the W3 specification.

Example

The following example uses the <emphasis> element to make an announcement:

<emphasis level="moderate">This is an important announcement</emphasis>

<par>

A parallel media container that allows you to play multiple media elements at once. The only allowed content is a set of one or more <par>, <seq>, and <media> elements. The order of the <media> elements is not significant.

Unless a child element specifies a different begin time, the implicit begin time for the element is the same as that of the <par> container. If a child element has an offset value set for its begin or end attribute, the element's offset will be relative to the beginning time of the <par> container. For the root <par> element, the begin attribute is ignored and the beginning time is when SSML speech synthesis process starts generating output for the root <par> element (i.e. effectively time "zero").

Example

<speak>
  <par>
    <media xml:id="question" begin="0.5s">
      <speak>Who invented the Internet?</speak>
    </media>
    <media xml:id="answer" begin="question.end+2.0s">
      <speak>The Internet was invented by cats.</speak>
    </media>
    <media begin="answer.end-0.2s" soundLevel="-6dB">
      <audio
        src="https://actions.google.com/.../cartoon_boing.ogg"/>
    </media>
    <media repeatCount="3" soundLevel="+2.28dB"
      fadeInDur="2s" fadeOutDur="0.2s">
      <audio
        src="https://actions.google.com/.../cat_purr_close.ogg"/>
    </media>
  </par>
</speak>

<seq>

A sequential media container that allows you to play media elements one after another. The only allowed content is a set of one or more <seq>, <par>, and <media> elements. The order of the media elements is the order in which they are rendered.

The begin and end attributes of child elements can be set to offset values (see Time Specification below). Those child elements' offset values will be relative to the end of the previous element in the sequence or, in the case of the first element in the sequence, relative to the beginning of its <seq> container.

Example

<speak>
  <seq>
    <media begin="0.5s">
      <speak>Who invented the Internet?</speak>
    </media>
    <media begin="2.0s">
      <speak>The Internet was invented by cats.</speak>
    </media>
    <media soundLevel="-6dB">
      <audio
        src="https://actions.google.com/.../cartoon_boing.ogg"/>
    </media>
    <media repeatCount="3" soundLevel="+2.28dB"
      fadeInDur="2s" fadeOutDur="0.2s">
      <audio
        src="https://actions.google.com/.../cat_purr_close.ogg"/>
    </media>
  </seq>
</speak>

<media>

Represents a media layer within a <par> or <seq> element. The allowed content of a <media> element is an SSML <speak> or <audio> element. The following table describes the valid attributes for a <media> element.

Attributes

Time specification

A time specification, used for the value of `begin` and `end` attributes of <media> elements and media containers (<par> and <seq> elements), is either an offset value (for example, +2.5s) or a syncbase value (for example, foo_id.end-250ms).

• Offset value - Time offset value is an SMIL Timecount-value that allows values that match the regular expression: "\s\*(+|-)?\s\*(\d+)(\.\d+)?(h|min|s|ms)?\s\*"

  The first digit string is the whole part of the decimal number and the second digit string is the decimal fractional part. The default sign (i.e. "(+|-)?") is "+". The unit values correspond to hours, minutes, seconds, and milliseconds respectively. The default for the units is "s" (seconds).

• Syncbase value - A syncbase value is an SMIL syncbase-value that allows values that match the regular expression: "([-_#]|\p{L}|\p{D})+\.(begin|end)\s\*(+|-)\s\*(\d+)(\.\d+)?(h|min|s|ms)?\s\*"

  The digits and units are interpreted in the same way as an offset value.

<phoneme>

You can use the <phoneme> tag to produce custom pronunciations of words inline. Text-to-Speech accepts the IPA and X-SAMPA phonetic alphabets. See the phonemes page for a list of supported languages and phonemes.

Each application of the <phoneme> tag directs the pronunciation of a single word:

  <phoneme alphabet="ipa" ph="ˌmænɪˈtoʊbə">manitoba</phoneme>
  <phoneme alphabet="x-sampa" ph='m@"hA:g@%ni:'>mahogany</phoneme>

Stress markers

There are up to three levels of stress that can be placed in a transcription:

1. Primary stress: Denoted with /ˈ/ in IPA and /"/ in X-SAMPA.
2. Secondary stress: Denoted with /ˌ/ in IPA and /%/ in X-SAMPA.
3. Unstressed: Not denoted with a symbol (in either notation).

Some languages might have fewer than three levels or not denote stress placement at all. See the phonemes page to see the stress levels available for your language. Stress markers are placed at the start of each stressed syllable. For example, in US English:

Broad vs Narrow Transcriptions

As a general rule, keep your transcriptions more broad and phonemic in nature. For example, in US English, transcribe intervocalic /t/ (instead of using a tap):

There are some instances where using the phonemic representation makes your TTS results sound unnatural (for example, if the sequence of phonemes is anatomically difficult to pronounce).

One example of this is voicing assimilation for /s/ in English. In this case the assimilation should be reflected in the transcription:

Reduction

Every syllable must contain one (and only one) vowel. This means that you should avoid syllabic consonants and instead transcribe them with a reduced vowel. For example:

Syllabification

You can optionally specify syllable boundaries by using /./. Each syllable must contain one (and only one) vowel. For example:

Custom pronunciation dictionary

As an alternative to providing pronunciations inline with the phoneme tag, provide a dictionary of custom pronunciations in the speech synthesis RPC. When the custom pronunciation dictionary is in the request, the input text will automatically be transformed with the SSML phoneme tag.

As an example, the following request with text input and custom pronunciation will be transformed and will be equivalent to the SSML below.

Original Input:

input: {
  text: 'Hello world! It is indeed a beautiful world!',
  custom_pronunciations: {
    pronunciations: {
      phrase: 'world'
      phonetic_encoding: PHONETIC_ENCODING_IPA
      pronunciation: 'wɜːld'
    }
  }
}

Transformed input:

input: {
  ssml: '<speak>Hello <phoneme alphabet="ipa" ph="wɜːld">world</phoneme>! It is indeed a beautiful <phoneme alphabet="ipa" ph="wɜːld">world</phoneme>!</speak>'
}

Durations

Text-to-Speech supports <say-as interpret-as="duration"> to correctly read durations. For example, the following example would be verbalized as "five hours and thirty minutes":

<say-as interpret-as="duration" format="h:m">5:30</say-as>

The format string supports the following values:

<voice>

The <voice> tag allows you to use more than one voice in a single SSML request. In the following example, the default voice is an English male voice. All words will be synthesized in this voice except for "qu'est-ce qui t'amène ici", which will be verbalized in French using a female voice instead of the default language (English) and gender (male).

<speak>And then she asked, <voice language="fr-FR" gender="female">qu'est-ce qui
t'amène ici</voice><break time="250ms"/> in her sweet and gentle voice.</speak>

Alternatively, you can use a <voice> tag to specify an individual voice (the voice name on the supported voices page) rather than specifying a language and/or gender:

<speak>The dog is friendly<voice name="fr-CA-Wavenet-B">mais la chat est
mignon</voice><break time="250ms"/> said a pet shop
owner</speak>

When you use the <voice> tag, Text-to-Speech expects to receive either a name (the name of the voice you want to use) or a combination of the following attributes. All three attributes are optional but you must provide at least one if you don't provide a name.

• gender: One of "male", "female" or "neutral".
• variant: Used as a tiebreaker in cases where there are multiple possibilities of which voice to use based on your configuration.
• language: Your desired language. Only one language can be specified in a given <voice> tag. Specify your language in BCP-47 format. You can find the BCP-47 code for your language in the language code column on the supported voices and languages page.

You can also control the relative priority of each of the gender, variant, and language attributes using two additional tags: required and ordering.

• required: If an attribute is designated as required and not configured properly, the request will fail.
• ordering: Any attributes listed after an ordering tag are considered as preferred attributes rather than required. The Text-to-Speech API considers preferred attributes on a best effort basis in the order they are listed after the ordering tag. If any preferred attributes are configured incorrectly, Text-to-Speech might still return a valid voice but with the incorrect configuration dropped.

Examples of configurations using the required and ordering tags:

<speak>And there it was <voice language="en-GB" gender="male" required="gender"
ordering="gender language">a flying bird </voice>roaring in the skies for the
first time.</speak>

<speak>Today is supposed to be <voice language="en-GB" gender="female"
ordering="language gender">Sunday Funday.</voice></speak>

<lang>

You can use <lang> to include text in multiple languages within the same SSML request. All languages will be synthesized in the same voice unless you use the <voice> tag to explicitly change the voice. The xml:lang string must contain the target language in BCP-47 format (this value is listed as "language code" in the supported voices table). In the following example "chat" will be verbalized in French instead of the default language (English):

<speak>The french word for cat is <lang xml:lang="fr-FR">chat</lang></speak>

Text-to-Speech supports the <lang> tag on a best effort basis. Not all language combinations produce the same quality results if specified in same SSML request. In some cases, a language combination might produce an effect that is detectible but subtle or perceived as negative. Known issues:

• Japanese with Kanji characters is not supported by the <lang> tag. The input is transliterated and read as Chinese characters.
• Semitic languages such as Arabic, Hebrew, and Persian are not supported by the <lang> tag and will result in silence. If you want to use any of these languages we recommend using the <voice> tag to switch to a voice that speaks your desired language (if available).

SSML timepoints

The Text-to-Speech API supports the use of timepoints in your created audio data. A timepoint is a timestamp (in seconds, measured from the beginning of the generated audio) that corresponds to a designated point in the script. You can set a timepoint in your script using the <mark> tag. When the audio is generated, the API then returns the time offset between the beginning of the audio and the timepoint.

There are two steps to setting a timepoint:

1. Add a <mark> SSML tag to the point in the script that you want a timestamp for.
2. Set TimepointType to SSML_MARK. If this field is not set, timepoints are not returned by default.

The following example returns two timepoints:

• timepoint_1: Indicates the time (in seconds) that the word "Mark" appears in the generated audio.
• timepoint_2: Indicates the time (in seconds) that the word "see" appears in the generated audio.

<speak>Hello <mark name="timepoint_1"/> Mark. Good to <mark
name="timepoint_2"/> see you.</speak>

Styles

1. en-US-Neural2-F
2. en-US-Neural2-J

Use the <google:style> tag to control what style to use. Only use the tag around full sentences.

Example:

<speak><google:style name="lively">Hello I'm so happy today!</google:style></speak>

The name field supports the following values:

1. apologetic
2. calm
3. empathetic
4. firm
5. lively