分享提取脚本

[HuanLinOTO]

const { appendFileSync, writeFileSync, existsSync, createReadStream } = require("fs");
const readline = require('readline');
const cliProgress = require('cli-progress')
// init a progress bar
console.log("读取文件...");
const result = require("./resultv33.1.json")
console.log("读取完成");

var chars = ['标贝'];

var isready = {'标贝': 0};

const check = (char)=>{
    // console.log(char);
    if(isready[char] == undefined) {
        isready[char] = chars.push(char)-1;
    }
    return isready[char];
}

const npclist = ["派蒙","流浪者","珐露珊","莱依拉","纳西妲","妮露","坎蒂丝","赛诺","多莉","提纳里","柯莱","鹿野院平藏","久岐忍","夜兰","空","荧","神里绫人","八重神子","云堇","申鹤","荒泷一斗","五郎","优菈","阿贝多","托马","胡桃","达达利亚","雷电将军","珊瑚宫心海","埃洛伊","宵宫","神里绫华","枫原万叶","温迪","刻晴","莫娜","可莉","琴","迪卢克","七七","魈","钟离","甘雨","早柚","九条裟罗","凝光","菲谢尔","班尼特","丽莎","行秋","迪奥娜","安柏","重云","雷泽","芭芭拉","罗莎莉亚","香菱","凯亚","北斗","诺艾尔","砂糖","辛焱","烟绯"]

const bar1 = new cliProgress.SingleBar({}, cliProgress.Presets.shades_classic);
writeFileSync("genshin_train.txt",'')
writeFileSync("genshin_val.txt",'')
// start the progress bar with a total value of 200 and start value of 0
bar1.start(Object.keys(result).length, 0);
var num = 0;
for (const key in result) {
    num++;
    bar1.update(num)
    const item = result[key];
    if(npclist.indexOf(item.npcName) == -1) continue;
    if(item.npcName == undefined || item.npcName.indexOf("#") != -1 || item.npcName.indexOf("test") != -1 || item.npcName.indexOf("?") != -1 || item.npcName.indexOf("？") != -1) continue
    if(item.language != "CHS") continue;
    if(item.type == "Card") continue;
    if(item.text == undefined) continue;
    if(!existsSync(`V33_Merged_Chinese_Wav/Merged_Chinese_Wav${item.fileName.replace("Chinese","").replaceAll("\\","/")}`)) continue;
    // 是否作为val集
    if(Math.random()*100%25>>>0 == 1)
        appendFileSync("./genshin_val.txt",`wavs${item.fileName.replace("Chinese","").replaceAll("\\","/")}|${check(item.npcName)}|${item.text.replace(/<.*?>/g,'').replace(/{.*}/gm,'').replaceAll("\\n",'').replaceAll("#",'')}\n`);
    else
        appendFileSync("./genshin_train.txt",`wavs${item.fileName.replace("Chinese","").replaceAll("\\","/")}|${check(item.npcName)}|${item.text.replace(/<.*?>/g,'').replace(/{.*}/gm,'').replaceAll("\\n",'').replaceAll("#",'')}\n`);
}

bar1.stop()
console.log("标贝");
const readInterface = readline.createInterface({
    input: createReadStream('biaobei_train.txt'),
});

readInterface.on('line', function(line) {
    // console.log(line);
        // 是否作为val集
        if(Math.random()*100%25>>>0 == 1)
        appendFileSync("./genshin_val.txt",line.replace("|","|0|")+"\n");
    else
        appendFileSync("./genshin_train.txt",line.replace("|","|0|")+"\n");
});

writeFileSync("chars.txt",JSON.stringify(chars)+"\n"+JSON.stringify(isready))

默认使用了标贝数据集作为 speaker 0 辅助其他speaker 如果无需标贝可以去掉L47-L59 L9-L11改成

var chars = [];

var isready = {};

其中 L21 表示需要提取的 speaker 列表, 可自己修改

需要 nodejs 16 以上环境 使用前先

npm i cli-progress

[HuanLinOTO]

顺便给一下音频格式化的py脚本

from pydub import AudioSegment
import threading
import os
def classify(input,output):
    song = AudioSegment.from_file(input)
    song = song.set_channels(1)
    song = song.set_frame_rate(采样率)
    song.export(output, format='wav')
for root,dirs,files in os.walk("这里改成音频目录/"):
    for file in files:
        if(os.path.join(root,file).endswith(".wav")):
            print(os.path.join(root,file))
            threading. Thread(target=classify,args= (os.path.join(root,file),os.path.join(root,file)))
            classify(os.path.join(root,file),os.path.join(root,file))

[yz1392946854]

谢谢脚本，想请教一下什么是“标贝”数据集

[HuanLinOTO]

谢谢脚本，想请教一下什么是“标贝”数据集

由标贝公司出品的免费、开源已标注语音数据集

[zhamao114514]

大佬，脚本怎么用啊