text to audio file on node.js

Question

Im looking for an optimized legal way to create an audio file from text on nodejs.

Now I see 5 variants:

1) simple hhtp request to google translate text-to-sppeach api. This variant isn't good, because each request demands on generated token e.g. 'tk:729008.879154' Without this it can fail. Besides that, this option is 'illegal'.

2) http request to google translate text-to-sppeach api from 'console browser' - puppeteer

Is there a way to generate right token key to make this request 'legal'?

3) use Web Speech Api in puppeteer to get binary data and save it to file? Or is there a way to work with Chromium/Chrome source code ?

4) Using any other technology/language library on machine with nodejs and to use js as interpreter to call commands in this technology/programm. Any ideas?

5) Any free public api with different languages support (dream api)?

Any advice will be highly appreciated.

Answer 1

Using text2wav.node.js, you don't need to depend on any external online service or separately installed host program. It's all self-contained FOSS. Besides, it supports 101 languages.

Answer 2

One possible approach is to wrap the eSpeak command line tool (Windows & Linux) http://espeak.sourceforge.net/. You can then wrap with Node.js.

const { exec } = require('child_process');

var outputFile = process.argv[2] || "output.wav";
var voice = process.argv[3] || "en-uk-north";
var text = process.argv[4] || "hello there buddy";
var command = `espeak.exe -v ${voice} -w ${outputFile} "${text}"`;

exec(command, (err, stdout, stderr) => {
  if (err) {
    console.log("Error occurred: ", err);
    return;
  }
});

This gives a fairly low quality output.

I've also played with the Bing Speech API and the output is very good, I've created a Node.js example. You would need to sign up for an API key but this is very easy (you to to https://azure.microsoft.com/en-us/try/cognitive-services/ and select "Speech").

const key = 'your api key here';

function synthesizeSpeech(apiKey)
{
    const fs = require('fs');
    const request = require('request');
    const xmlbuilder = require('xmlbuilder');
    const text = process.argv[2] || "The fault, dear Brutus, is not in our stars, But in ourselves, that we are underlings.";
    const outputFile = process.argv[3] || "speech.wav";

    var ssml_doc = xmlbuilder.create('speak')
        .att('version', '1.0')
        .att('xml:lang', 'en-au')
        .ele('voice')
        .att('xml:lang', 'en-au')
        .att('xml:gender', 'Female')
        .att('name', 'Microsoft Server Speech Text to Speech Voice (en-AU, HayleyRUS)')
        .txt(text)
        .end();
    var post_speak_data = ssml_doc.toString();

    console.log('Synthesizing speech: ', text);
    request.post({
        url: 'https://api.cognitive.microsoft.com/sts/v1.0/issueToken',
        headers: {
            'Ocp-Apim-Subscription-Key' : apiKey
        }
    }, function (err, resp, access_token) {
        if (err || resp.statusCode != 200) {
            console.log(err, resp.body);
        } else {
            try {
                request.post({
                    url: 'https://speech.platform.bing.com/synthesize',
                    body: post_speak_data,
                    headers: {
                        'content-type' : 'application/ssml+xml',
                        'X-Microsoft-OutputFormat' : 'riff-16khz-16bit-mono-pcm',
                        'Authorization': 'Bearer ' + access_token,
                        'X-Search-AppId': '9FCF779F0EFB4E8E8D293EEC544221E9',
                        'X-Search-ClientID': '0A13B7717D0349E683C00A6AEA9E8B6D',
                        'User-Agent': 'Node.js-Demo'
                    },
                    encoding: null
                }, function (err, resp, data) {
                    if (err || resp.statusCode != 200) {
                        console.log(err, resp.body);
                    } else {
                        try {
                            console.log('Saving output to file: ', outputFile);
                            fs.writeFileSync(outputFile, data);
                        } catch (e) {
                            console.log(e.message);
                        }
                    }
                });
            } catch (e) {
                console.log(e.message);
            }
        }
    });
}

synthesizeSpeech(key);

Also check out the MARY project here: http://mary.dfki.de/, this is an open source server that you can install, the voice output is very good, you could make calls to the server from node.js.

If you install the Mary Speech engine (quite easy):

"use strict";

const fs = require('fs');
const request = require('request');
const text = process.argv[2] || "The fault, dear Brutus, is not in our stars, But in ourselves, that we are underlings.";
const outputFile = process.argv[3] || "speech_mary_output.wav";

const options = {
    url: `http://localhost:59125/process?INPUT_TEXT=${text}!&INPUT_TYPE=TEXT&OUTPUT_TYPE=AUDIO&AUDIO=WAVE_FILE&LOCALE=en_US&VOICE=cmu-slt-hsmm`,
    encoding: null // Binary data.
}

console.log('Synthesizing speech (using Mary engine): ', text);
console.log('Calling: ', options.url);
request.get(options, function (err, resp, data) {
    if (err || resp.statusCode != 200) {
        console.log(err, resp.body);
    } else {
        try {
            console.log(`Saving output to file: ${outputFile}, length: ${data.length} byte(s)`);
            fs.writeFileSync(outputFile, data, { encoding: 'binary'});
        } catch (e) {
            console.log(e.message);
        }
    }
});

This will synthesize speech for you. No API key required!