ssml-document

Follow the W3C SSML (Speech Synthesis Markup Language) specification to build SSML documents that meet the requirements of all major voice service providers.

遵循 W3C SSML（语音合成标记语言）规范，构建符合大部分语音/数字人服务提供商要求的 SSML 文档。

Currently supports the generation of SSML documents for these service providers: Microsoft Azure / Aliyun / Tencent Cloud / Google Cloud / Amazon AWS / GuijiAI / Tencent YunXiaoWei / XMOV / Microsoft XiaoBing / Sensetime / Eta ...

当前支持生成这些服务商的SSML文档：微软云 / 阿里云 / 腾讯云 / 谷歌云 / 亚马逊云 / 火山云 / 硅基智能 / 腾讯云小微 / XMOV魔珐科技 / 微软小冰 / 商汤科技 / 艾塔 ...

Every month we check the development documents of these service providers to ensure that the library is not out of date.

每月我们会检查这些服务商的开发文档，以确保本库不会过时。

Initialization

npm i ssml-document --save

Features

Support the construction of SSML that meets the requirements of mainstream voice service providers 支持构建符合主流语音服务商要求的SSML
Supports parsing and construction of aggregated SSML (intermediate state) 支持聚合SSML（中间态）的解析和构建
Supports almost all SSML document tags 支持几乎全部的SSML文档标签
Support for selectively constructing or discarding tags and attributes based on service providers 支持根据服务商选择性构建或丢弃标签与属性
Support expression compilation to make content change dynamically 支持JavaScript表达式求值使内容动态变化
The library supports the following SSML tags and corresponding elements 该库支持以下SSML标签和对应的元素
- speak - Document
- voice - Voice
- prosody - Prosody
- p - Paragraph
- s - Sentence
- w - Word
- break - Break
- phoneme - Phoneme
- say-as - SayAs
- sub - Subsitute
- audio - Audio
- background-audio - BackgroundAudio
- express-as - ExpressAs
- emotion - Emotion
- effect - Effect
- emphasis - Emphasis
- lang - Language
- mark - Mark
- bookmark - Bookmark
- seq - Sequential
- par - Parallel
- lexicon - Lexicon
- auto-breaths - AutoBreaths
- silence - Silence

Basic Usage

Build microsoft-azure SSML:

构建微软云语音SSML：

const { Document, ServiceProvider } = require("ssml-document");
const doc = new Document();
const ssml = doc
.voice("zh-CN-XiaoxiaoNeural")
.prosody({ rate: 1.2, pitch: 1.1 })
.say("Hello World")
.pause(500)
.say("GO GO GO")
.sayAs("123456", { interpretAs: "digits" })
.up()
.up()
.render({
    pretty: true,  //format SSML
    provider: ServiceProvider.Microsoft
});
console.log(ssml);
/*
<?xml version="1.0"?>
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts">
  <voice name="zh-CN-XiaoxiaoNeural">
    <prosody pitch="110%" rate="120%">
      Hello World
      <break time="500ms"/>
      GO GO GO
      <say-as interpret-as="digits">123456</say-as>
    </prosody>
  </voice>
</speak>
*/

Build aliyun SSML:

构建阿里云语音SSML：

const { Document, ServiceProvider } = require("ssml-document");
const doc = new Document();
const ssml = doc
.voice("aixia")
.prosody({ rate: 1.2, pitch: 1.1 })
.say("我的身高")
.phoneme("长", {
    alphabet: "py",
    ph: "zhǎng"  //or zhang3
})
.say("高了")
.s("欢迎来到")
.sub("W3C", "万维网")
.up()
.up()
.up()
.render({ pretty: true, provider: ServiceProvider.Aliyun });
console.log(ssml);
/*
<speak voice="aixia" rate="100" pitch="50">
  我的身高
  <phoneme alphabet="py" ph="zhang3">长</phoneme>
  高了
  <s>
    欢迎来到
    <sub alias="万维网">W3C</sub>
  </s>
</speak>
*/

Build aggregated SSML for storage or transport:

构建用于存储或传输的聚合SSML：

const { Document } = require("ssml-document");
const doc = new Document();
const ssml = doc
.voice("aixia")
.say("Hello World")
.break(2000)
.say("Bye")
.up()
.render({ pretty: true });  //no provider
console.log(ssml);
/*
<?xml version="1.0"?>
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">
  <voice name="aixia">
    Hello World
    <break time="2s"/>
    Bye
  </voice>
</speak>
*/

Support for marking expected voice service providers in aggregated SSML:

支持在聚合SSML中标记期望的语音服务商：

const { Document } = require("ssml-document");
//Frontend code:
//前端代码：
const doc = new Document({ provider: ServiceProvider.Aliyun });
const ssml = doc
.voice("aida")
.say("Hello World")
.up()
.render({ pretty: true });  //no provider
console.log(ssml);
/*
<?xml version="1.0"?>
<speak provider="aliyun" version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">
  <voice name="aida">Hello World</voice>
</speak>
*/

//Backend code:
//后端代码：
const doc2 = Document.parse(ssml);
console.log(doc2.provider);  //aliyun
const ssml2 = doc2.render({ provider: doc2.provider });  //build aliyun SSML
console.log(ssml2);
//<speak voice="aida">Hello World</speak>

Get speech rate, pitch, volume and speaker parameters.

获取语音语速、语调、音量以及发音人参数

const { Document } = require("ssml-document");
const doc = new Document();
doc.voice("aixia")
  .prosody({ rate: 1.2, pitch: 1.1 })
  .say("Example Text");
console.log(doc.rate, doc.pitch, doc.volume, doc.speaker);
//1.2 1.1 100 aixia

Building SSML based on elements:

基于元素构建SSML：

const { Document, ServiceProvider, elements } = require("ssml-document");
const { Prosody, Paragraph } = elements;
const doc = new Document();
const prosody = new Prosody({ rate: 1.2, pitch: 1.1 });
const p = new Paragraph();
p.appendChild("Hello World");
prosody.appendChild(p);
doc.appendChild(prosody);
const ssml = doc.render({ provider: ServiceProvider.Amazon });
console.log(ssml);
//<speak><prosody pitch="110%" rate="120%"><p>Hello World</p></prosody></speak>

Setting the compile attribute for tags enables JavaScript syntax expression evaluation:

给标签设置compile属性启用JavaScript语法表达式求值:

const { Document } = require("ssml-document");
const aggregationSSML = '<?xml version="1.0"?>\
<speak compile="true" version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">\
  <voice name="aixia">\
    Master, {{[\"today is a beautiful day\",\"have a good day from now\",\"the weather is not bad today\"][parseInt(Math.random() * 3)]}}.\
    <break time="500"/>\
    The current time is {{time}}.\
  </voice>\
</speak>';
const date = new Date();
const doc = Document.parse(aggregationSSML, {
    dataset: {
        time: date.getHours() + ":" + date.getMinutes()
    }
});
const ssml = doc.render({
    provider: doc.provider,
    pretty: true
});
console.log(ssml);
/*
<?xml version="1.0"?>
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">
  <voice name="aixia">
    Master, have a good day from now.
    <break time="500ms"/>
    The current time is 22:48.
  </voice>
</speak>
*/

Setting for/if/elif/else attributes for tags can also implement loops and conditional divergence. 给标签设置 for/if/elif/else 属性还能够实现循环和条件分歧。

const { Document } = require("ssml-document");
const aggregationSSML = '<?xml version="1.0"?>\
<speak compile="true" version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">\
  <voice name="aixia">\
    <s if="{{Math.random() > 0.5}}" for="{{s1}}">{{item}}<break time="100"></s>\
    <s else="true" for="{{s2}}">{{item}}<break time="100"></s>\
  </voice>\
</speak>';
const doc = Document.parse(aggregationSSML, {
    dataset: {
        s1: ["Oh", "my", "god"],
        s2: ["Holy", "crap"]
    }
});
const ssml = doc.render({
    provider: doc.provider,
    pretty: true
});
console.log(ssml);
/*
<?xml version="1.0"?>
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">
  <voice name="aixia">
    <s>
      Holy
      <break time="100ms"/>
    </s>
    <s>
      crap
      <break time="100ms"/>
    </s>
  </voice>
</speak>
*/

Elements

Document

Tag: speak
Support: w3c / microsoft / aliyun / tencent / google / amazon

Voice

Tag: voice
Support: w3c / microsoft / google

Prosody

Tag: prosody
Support: w3c / microsoft / google / amazon

Paragraph

Tag: p
Support: w3c / microsoft / aliyun / google / amazon

Sentence

Tag: s
Support: w3c / microsoft / aliyun / google / amazon

Word

Tag: w
Support: w3c / aliyun / google / amazon

Break

Tag: break
Support: w3c / microsoft / aliyun / tencent / google / amazon / yunXiaoWei

Phoneme

Tag: phoneme
Support: w3c / microsoft / aliyun / tencent / google / amazon / yunXiaoWei

SayAs

Tag: say-as
Support: w3c / microsoft / aliyun / tencent / google / amazon

Subsitute

Tag: sub
Support: w3c / aliyun / tencent / google / amazon

Audio

Tag: audio
Support: w3c / microsoft / aliyun* / google

BackgroundAudio

Tag: background-audio / backgroundAudio / backgroundaudio
Support: microsoft / aliyun*

ExpressAs

Tag: express-as
Support: microsoft / aliyun* / amazon*

Emotion

Tag: emotion
Support: microsoft* / aliyun / amazon*

Effect

Tag: effect
Support: aliyun* / amazon

Emphasis

Tag: emphasis
Support: w3c / microsoft / google / amazon

Language

Tag: lang
Support: w3c / microsoft / google

Mark

Tag: mark
Support: w3c / google / amazon

Bookmark

Tag: bookmark
Support: microsoft

Sequential

Tag: seq
Support: microsoft

Parallel

Tag: par
Support: w3c / google

Sequential

Tag: seq
Support: w3c / google

Lexicon

Tag: lexicon
Support: w3c / microsoft

AutoBreaths

Tag: auto-breaths
Support: amazon

Silence

Tag: silence
Support: microsoft

TIDC/ssml-document

ssml-document

Initialization

Features

Basic Usage

Elements

Document

Voice

Prosody

Paragraph

Sentence

Word

Break

Phoneme

SayAs

Subsitute

Audio

BackgroundAudio

ExpressAs

Emotion

Effect

Emphasis

Language

Mark

Bookmark

Sequential

Parallel

Sequential

Lexicon

AutoBreaths

Silence