Google Text-to-speech - 从 txt 文件的各行加载文本

问题描述

我在 Node.js 中使用 Google TextToSpeech API 从文本生成语音。我能够获得与为语音生成的文本同名的输出文件。但是，我需要稍微调整一下。我希望我可以同时生成多个文件。关键是，例如，我有 5 个单词（或句子）要生成，例如猫，狗，房子，天空，太阳。我想将它们分别生成到一个单独的文件中：cat.wav、dog.wav 等。

我还希望应用程序能够从 * .txt 文件中读取这些单词（每个单词/句子在 * .txt 文件的单独一行中）。

有这种可能吗？下面我粘贴了 * .js 文件代码和我正在使用的 * .json 文件代码。

*.js

const textToSpeech = require('@google-cloud/text-to-speech');
const fs = require('fs');
const util = require('util');
const projectId = 'forward-dream-295509'
const keyFilename = 'myauth.json'
const client = new textToSpeech.TextToSpeechClient({ projectId,keyFilename });
const YourSetting = fs.readFileSync('setting.json');
async function Text2Speech(YourSetting) {
  const [response] = await client.synthesizeSpeech(JSON.parse(YourSetting));
  const writeFile = util.promisify(fs.writeFile);
  await writeFile(JSON.parse(YourSetting).input.text + '.wav',response.audioContent,'binary');
  console.log(`Audio content written to file: ${JSON.parse(YourSetting).input.text}`);
}
Text2Speech(YourSetting);

*.json

{
  "audioConfig": {
    "audioEncoding": "LINEAR16","pitch": -2,"speakingRate": 1
  },"input": {
    "text": "Text to Speech" 
  },"voice": {
    "languageCode": "en-US","name": "en-US-Wavenet-D"
  }
}

我不太擅长编程。我在 google 上找到了一个关于如何执行此操作的教程，并对其稍作修改，以便保存文件的名称与生成的文本相同。

非常感谢您的帮助。阿雷克

解决方法

开始吧 - 我还没有测试过它，但这应该显示如何读取文本文件，拆分为每一行，然后使用设置的并发在它上面运行 tts。它使用您需要添加到项目中的 p-any 和 filenamify npm 包。请注意，谷歌可能有 API 节流或速率限制，我在这里没有考虑到 - 如果有问题，可以考虑使用 p-throttle 库。

// https://www.npmjs.com/package/p-map
const pMap = require('p-map');

// https://github.com/sindresorhus/filenamify
const filenamify = require('filenamify');


const textToSpeech = require('@google-cloud/text-to-speech');
const fs = require('fs');
const path = require('path');
const projectId = 'forward-dream-295509'
const keyFilename = 'myauth.json'
const client = new textToSpeech.TextToSpeechClient({ projectId,keyFilename });
const rawSettings = fs.readFileSync('setting.json',{ encoding: 'utf8'});

// base data for all requests (voice,etc)
const yourSetting = JSON.parse(rawSettings);

// where wav files will be put
const outputDirectory = '.';

async function Text2Speech(text,outputPath) {
    // include the settings in settings.json,but change text input
    const request = {
        ...yourSetting,input: { text }
    };
    const [response] = await client.synthesizeSpeech(request);

    await fs.promises.writeFile(outputPath,response.audioContent,'binary');
    console.log(`Audio content written to file: ${text} = ${outputPath}`);
    // not really necessary,but you could return something if you wanted to
    return response;
}

// process a line of text - write to file and report result (success/error)
async function processLine(text,index) {
    // create output path based on text input (use library to ensure it's filename safe)
    const outputPath = path.join(outputDirectory,filenamify(text) + '.wav');
    const result = {
        text,lineNumber: index,path: outputPath,isSuccess: null,error: null
    };
    try {
        const response = await Text2Speech(text,outputPath);
        result.isSuccess = true;
    } catch (error) {
        console.warn(`Failed: ${text}`,error);
        result.isSuccess = false;
        result.error = error;
    }
    return result;
}

async function processInputFile(filepath,concurrency = 3) {
    const rawText = fs.readFileSync(filepath,{ encoding: 'utf8'});
    const lines = rawText
        // split into one item per line
        .split(/[\r\n]+/)
        // remove surrounding whitespace
        .map(s => s.trim())
        // remove empty lines
        .filter(Boolean);
    
    const results = await pMap(lines,processLine,{ concurrency });
    console.log('Done!');
    console.table(results);
}

// create sample text file
const sampleText = `Hello World
cat
dog
another line of text`;
fs.writeFileSync('./my-text-lines.txt',sampleText);

// process each line in the text file,3 at a time
processInputFile('./my-text-lines.txt',3);

google-text-to-speech javascript json text-to-speech