Example of text to speech with VOICELOID(c)

.gitignore

.env
node_modules

.node-version

8.1.2

README.md

— docomoの音声合成APIを利用して無料でVOICELOIDっぽい声を生成してみる | WEB EGG

Getting started

1. docomo Developer supportでAPIキーを作成する
2. 以下のコマンドを実行

npm install
export DOCOMO_API_KEY=XXXXXXXXXXXX
./playbook-to-voices 豚野郎_sm30193805.csv -p 豚野郎_preset.csv -o ./talk.wav

debuggable-fluent-ffmpeg.js

const debug = require('debug')('ffmpeg')
const ffmpeg = require('fluent-ffmpeg')

ffmpeg.prototype._prepare = (function (org) {
  return function(callback, readMetadata) {
    org.call(this, (err, args) => {
      debug(args.join(' '))
      callback(err, args)
    }, readMetadata)
  }
}(ffmpeg.prototype._prepare))

module.exports = ffmpeg

debuggable-isomorphic-fetch.js

const debug = require('debug')('fetch')
const fetch = require('isomorphic-fetch')

module.exports = async (url, opts) => {
  debug(`${opts.method} ${url}`)
  const response = await fetch(url, opts)
  debug(`${response.status} ${response.statusText}`)
  if (!response.ok) {
    debug(response.headers)
  }
  return response
}

package-lock.json

{
  "name": "playbook-to-voices",
  "version": "0.1.0",
  "lockfileVersion": 1,
  "requires": true,
  "dependencies": {
    "async": {
      "version": "2.5.0",
      "resolved": "https://registry.npmjs.org/async/-/async-2.5.0.tgz",
      "integrity": "sha512-e+lJAJeNWuPCNyxZKOBdaJGyLGHugXVQtrAwtuAe2vhxTYxFTKE73p8JuTmdH0qdQZtDvI4dhJwjZc5zsfIsYw==",
      "requires": {
        "lodash": "4.17.4"
      }
    },
    "bluebird": {
      "version": "3.5.0",
      "resolved": "https://registry.npmjs.org/bluebird/-/bluebird-3.5.0.tgz",
      "integrity": "sha1-eRQg1/VR7qKJdFOop3ZT+WYG1nw="
    },
    "commander": {
      "version": "2.11.0",
      "resolved": "https://registry.npmjs.org/commander/-/commander-2.11.0.tgz",
      "integrity": "sha512-b0553uYA5YAEGgyYIGYROzKQ7X5RAqedkfjiZxwi0kL1g3bOaBNNZfYkzt/CL0umgD5wc9Jec2FbB98CjkMRvQ=="
    },
    "csv-parse": {
      "version": "1.2.0",
      "resolved": "https://registry.npmjs.org/csv-parse/-/csv-parse-1.2.0.tgz",
      "integrity": "sha1-BHtzhoq5qFdG6IX2N/ntD7ZFpCU="
    },
    "debug": {
      "version": "2.6.8",
      "resolved": "https://registry.npmjs.org/debug/-/debug-2.6.8.tgz",
      "integrity": "sha1-5zFTHKLt4n0YgiJCfaF4IdaP9Pw=",
      "requires": {
        "ms": "2.0.0"
      }
    },
    "encoding": {
      "version": "0.1.12",
      "resolved": "https://registry.npmjs.org/encoding/-/encoding-0.1.12.tgz",
      "integrity": "sha1-U4tm8+5izRq1HsMjgp0flIDHS+s=",
      "requires": {
        "iconv-lite": "0.4.18"
      }
    },
    "fluent-ffmpeg": {
      "version": "2.1.2",
      "resolved": "https://registry.npmjs.org/fluent-ffmpeg/-/fluent-ffmpeg-2.1.2.tgz",
      "integrity": "sha1-yVLeIkD4EuvaCqgAbXd27irPfXQ=",
      "requires": {
        "async": "2.5.0",
        "which": "1.2.14"
      }
    },
    "iconv-lite": {
      "version": "0.4.18",
      "resolved": "https://registry.npmjs.org/iconv-lite/-/iconv-lite-0.4.18.tgz",
      "integrity": "sha512-sr1ZQph3UwHTR0XftSbK85OvBbxe/abLGzEnPENCQwmHf7sck8Oyu4ob3LgBxWWxRoM+QszeUyl7jbqapu2TqA=="
    },
    "is-stream": {
      "version": "1.1.0",
      "resolved": "https://registry.npmjs.org/is-stream/-/is-stream-1.1.0.tgz",
      "integrity": "sha1-EtSj3U5o4Lec6428hBc66A2RykQ="
    },
    "isexe": {
      "version": "2.0.0",
      "resolved": "https://registry.npmjs.org/isexe/-/isexe-2.0.0.tgz",
      "integrity": "sha1-6PvzdNxVb/iUehDcsFctYz8s+hA="
    },
    "isomorphic-fetch": {
      "version": "2.2.1",
      "resolved": "https://registry.npmjs.org/isomorphic-fetch/-/isomorphic-fetch-2.2.1.tgz",
      "integrity": "sha1-YRrhrPFPXoH3KVB0coGf6XM1WKk=",
      "requires": {
        "node-fetch": "1.7.1",
        "whatwg-fetch": "2.0.3"
      }
    },
    "lodash": {
      "version": "4.17.4",
      "resolved": "https://registry.npmjs.org/lodash/-/lodash-4.17.4.tgz",
      "integrity": "sha1-eCA6TRwyiuHYbcpkYONptX9AVa4="
    },
    "ms": {
      "version": "2.0.0",
      "resolved": "https://registry.npmjs.org/ms/-/ms-2.0.0.tgz",
      "integrity": "sha1-VgiurfwAvmwpAd9fmGF4jeDVl8g="
    },
    "node-fetch": {
      "version": "1.7.1",
      "resolved": "https://registry.npmjs.org/node-fetch/-/node-fetch-1.7.1.tgz",
      "integrity": "sha512-j8XsFGCLw79vWXkZtMSmmLaOk9z5SQ9bV/tkbZVCqvgwzrjAGq66igobLofHtF63NvMTp2WjytpsNTGKa+XRIQ==",
      "requires": {
        "encoding": "0.1.12",
        "is-stream": "1.1.0"
      }
    },
    "ssml-builder": {
      "version": "0.2.4",
      "resolved": "https://registry.npmjs.org/ssml-builder/-/ssml-builder-0.2.4.tgz",
      "integrity": "sha1-dN+9OPmvU+0q0mlKWDb09mY38WM="
    },
    "uuid": {
      "version": "3.1.0",
      "resolved": "https://registry.npmjs.org/uuid/-/uuid-3.1.0.tgz",
      "integrity": "sha512-DIWtzUkw04M4k3bf1IcpS2tngXEL26YUD2M0tMDUpnUrz2hgzUBlD55a4FjdLGPvfHxS6uluGWvaVEqgBcVa+g=="
    },
    "whatwg-fetch": {
      "version": "2.0.3",
      "resolved": "https://registry.npmjs.org/whatwg-fetch/-/whatwg-fetch-2.0.3.tgz",
      "integrity": "sha1-nITsLc9oGH/wC8ZOEnS0QhduHIQ="
    },
    "which": {
      "version": "1.2.14",
      "resolved": "https://registry.npmjs.org/which/-/which-1.2.14.tgz",
      "integrity": "sha1-mofEN48D6CfOyvGs31bHNsAcFOU=",
      "requires": {
        "isexe": "2.0.0"
      }
    }
  }
}

package.json

{
  "name": "playbook-to-voices",
  "version": "0.1.0",
  "description": "Convert playbook to voices with VOICELOAD(c)",
  "main": "index.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  },
  "keywords": [
    "text2speech",
    "text_to_speech",
    "speechsynthesize"
  ],
  "author": "Leko <[email protected]>",
  "license": "MIT",
  "dependencies": {
    "bluebird": "^3.5.0",
    "commander": "^2.11.0",
    "csv-parse": "^1.2.0",
    "debug": "^2.6.8",
    "fluent-ffmpeg": "^2.1.2",
    "isomorphic-fetch": "^2.2.1",
    "ssml-builder": "^0.2.4",
    "uuid": "^3.1.0"
  }
}

playbook-to-voices

#!/usr/bin/env node
const fs = require('fs')
const querystring = require('querystring')
const commander = require('commander')
const Promise = require('bluebird')
const uuid = require('uuid/v4')
const parse = Promise.promisify(require('csv-parse'))
const debug = require('debug')
const SSML = require('ssml-builder')
const ffmpeg = require('./debuggable-fluent-ffmpeg')
const fetch = require('./debuggable-isomorphic-fetch')
const pkg = require('./package.json')

const unlink = Promise.promisify(fs.unlink)
const readFile = Promise.promisify(fs.readFile)
const writeFile = Promise.promisify(fs.writeFile)
const PRESET_NAME_DEFAULT = 'default'
const PRESET_DEFAULT = { rate: 1, pitch: 1, range: 1, volume: 1 }

SSML.prototype.voice = function (name, text) {
  this._elements.push(`<voice name="${name}">` + text + '</voice>')
  return this
}
SSML.prototype.ssml = (function (org) {
  return function () {
    const xmlPrefix = '<?xml version="1.0" encoding="utf-8" ?>\n'

    return xmlPrefix + org.call(this)
      .replace('<speak>', '<speak version="1.1">')
      .replace(/> </g, '><')
  }
})(SSML.prototype.ssml)

const parseCSV = async (path) => {
  debug('playbook:csv')(path)
  return readFile(path, 'utf-8').then(parse)
}

const parsePlaybook = async (path) => {
  const playbook = await parseCSV(path)
  return playbook.slice(1)
}
const parsePresets = async (presetPath) => {
  const presetRows = await parseCSV(presetPath)
  return presetRows.slice(1).reduce((acc, row) => {
    const [ voice, name, rate, pitch, range, volume ] = row
    debug('playbook:preset')(`${voice}(${name || PRESET_NAME_DEFAULT}): ${JSON.stringify({ rate, pitch, range, volume })}`)
    return Object.assign(acc, {
      [voice]: {
        [name || PRESET_NAME_DEFAULT]: Object.assign({}, PRESET_DEFAULT, { rate, pitch, range, volume })
      }
    })
  }, {})
}

const appendVoice = (ssml, voice, presetName, txt, presets) => {
  const map = {
    月読アイ: (ssml, txt) => ssml.voice('anzu', txt),
    弦巻マキ: (ssml, txt) => ssml.voice('maki', txt),
    結月ゆかり: (ssml, txt) => ssml.voice('sumire', txt),
  }

  let preset = PRESET_DEFAULT
  if (!map[voice]) {
    throw new Error(`Unknown voice: ${voice}`)
  }
  if (presets[voice] && presets[voice][presetName || PRESET_NAME_DEFAULT]) {
    preset = presets[voice][presetName || PRESET_NAME_DEFAULT]
  }
  const filteredPreset = Object.entries(preset).filter(([name, val]) => !!val)
  if (filteredPreset.length <= 0) {
    throw new Error(`Empty preset: ${voice}, ${presetName || PRESET_NAME_DEFAULT}`)
  }

  const attribtues = filteredPreset
    .map(([name, val]) => `${name}="${val}"`)
    .join(' ')
  debug('playbook:voice')(`${voice}(${attribtues})「${txt}」`)
  map[voice](ssml, `<prosody ${attribtues}>${txt}</prosody>`)
  return ssml
}

const generatePresets = async (presetPath) => {
  if (presetPath) {
    return parsePresets(presetPath)
  } else {
    return Promise.resolve({})
  }
}

const toSSML = (genPresets) => async (voices) => {
  const presets = await genPresets
  const ssml = voices.reduce((acc, [voice, presetName, text]) => {
    return appendVoice(acc, voice, presetName, text, presets)
  }, new SSML())

  return ssml.ssml()
}

const textToSpeech = async (ssml) => {
  const ENDPOINT = 'https://api.apigw.smt.docomo.ne.jp/aiTalk/v1/textToSpeech'
  const query = querystring.stringify({
    APIKEY: process.env.DOCOMO_API_KEY,
  })

  return fetch(`${ENDPOINT}?${query}`, {
    method: 'POST',
    body: ssml,
    headers: {
      'Content-Type': 'application/ssml+xml',
      'Accept': 'audio/L16',
    }
  })
}

const storeTemporary = async (response) => {
  if (response.ok) {
    const path = `/tmp/${uuid()}`
    debug('playbook:storeTemporary')(path)
    return response.buffer()
      .then(buff => writeFile(path, buff))
      .then(() => path)
  } else {
    return response.text()
      .then(text => Promise.reject(text))
  }
}

const toWav = (destPath) => async (pcmPath) => {
  return new Promise((resolve, reject) => {
    const cmd = ffmpeg()
      .input(pcmPath)
      .inputOptions(['-ac 1', '-ar 16000'])
      .inputFormat('s16be')
      .output(destPath)
      .on('end', () => {
        console.log(destPath)
        unlink(pcmPath).then(resolve)
      })
      .on('error', reject)

    cmd.run()
  })
}

commander
  .version(pkg.version)
  .arguments('<playbook>')
  .option('-p, --presets [path]', 'Define prosody presets')
  .option('-o, --output [path]', 'Set output path')
  .action((playbook) => {
    if (!playbook) {
      commander.outputHelp()
      return
    }

    const output = commander.output || playbook.replace('.csv', '') + '.wav'

    parsePlaybook(playbook)
      .then(toSSML(generatePresets(commander.presets)))
      .then(textToSpeech)
      .then(storeTemporary)
      .then(toWav(output))
      .catch(e => {
        console.error(e)
        process.exit(1)
      })
  })
  .parse(process.argv)

グレートエレキファイア.csv

グレートエレキファイア_preset.csv

豚野郎_preset.csv

豚野郎_sm30193805.csv

voice,preset,text
結月ゆかり,,皆さんこんにちは、結月ゆかりです
弦巻マキ,,"<phoneme ph=""ツル’／マ’キ"">弦巻</phoneme>マキです"
月読アイ,,ゆっくり霊夢です
結月ゆかり,,突然ですけど私、スーパーハカーになりました！
月読アイ,,この人いきなり何言ってんだ…
弦巻マキ,,なろうと思って簡単になれるものじゃないぞ
弦巻マキ,,あとハカーじゃなくてハッカーね
結月ゆかり,,ゆかりさんの華麗なハッキング技術で
結月ゆかり,,お前たちの個人情報を丸裸にしてやる！
結月ゆかり,,具体的には<phoneme ph=""パソコン"">PC</phoneme>の<phoneme ph=""ディー"">D</phoneme>ドライブの中身を晒してやる！
月読アイ,,やめてください！社会的に死ぬ人が出るのでやめてください！