在线文本转语音工具的实现

本文介绍: 基于python的在线文本转语音工具

最近有文字转语音功能的需求，虽然也有一些免费工具，不过并没有找到好用的在线的文本转语音工具，因此便有了这篇文章。

这是一个基于微软edge浏览器大声朗读功能的python库，作者将其进行封装，因此我们可以借由它来实现文字转语音功能，那么首先我先介绍一下这个库。edge-tts的github地址

pip3 install edge_tts

import asyncio
import edge_tts

Content = "注意看，这是一段文字转语音的测试"
voiceName = "zh-TW-YunJheNeural"
output = "sound.mp3"
rate = '+5%'
volume = '+0%'
pitch = '+0Hz'

async def main():
    communicate = edge_tts.Communicate(text = Content, voice=voiceName ,rate=rate, volume=volume,pitch=pitch)
    await communicate.save(output)
  
asyncio.run(main())

asyncio.run 是 Python 标准库中 asyncio 模块提供的一个函数，用于运行一个异步函数直到它完成。它通常用于简化异步代码的执行，特别是在运行一个完整的异步应用程序时，而edge_tts所需要的就是这么一个异步执行，使用edge_tts.Communicate来生成音频内容，然后使用communicate.save来将文件保存到本地路径，其中text为所需要的文本，voice为所需要的音色，这个一会介绍，rate为语速，volume为音量，默认0就是满的，pitch为音调，后三个所接收的数据需要带上正负号数字百分比的形式，其中pitch需要正负号数字 Hz的形式。

voice为音色参数，对于音色的获取，需要在安装完库之后在cmd里输入

edge-tts --list-voices

来查询，我列举一下这个库所支持的中文音色：

           { "Value": "zh-CN-XiaoxiaoNeural", "text": "晓晓" },
            { "Value": "zh-CN-XiaoyiNeural", "text": "小艺" },
            { "Value": "zh-CN-YunjianNeural", "text": "云建" },
            { "Value": "zh-CN-YunxiNeural", "text": "云溪" },
            { "Value": "zh-CN-YunxiaNeural", "text": "云霞" },
            { "Value": "zh-CN-YunyangNeural", "text": "云阳" },
            { "Value": "zh-CN-liaoning-XiaobeiNeural", "text": "东北小贝" },
            { "Value": "zh-CN-shaanxi-XiaoniNeural", "text": "山西小妮" },
            { "Value": "zh-HK-HiuGaaiNeural", "text": "粤语小妹" },
            { "Value": "zh-HK-HiuMaanNeural", "text": "粤语小妹2" },
            { "Value": "zh-HK-WanLungNeural", "text": "粤语小哥" },
            { "Value": "zh-TW-HsiaoChenNeural", "text": "台湾小妹" },
            { "Value": "zh-TW-HsiaoYuNeural", "text": "台湾小妹2" },
            { "Value": "zh-TW-YunJheNeural", "text": "台湾小哥" }

import asyncio
import time
import edge_tts
import os

async def convert(text, file_name) -> None:
    start_time = time.time()
    communicate = edge_tts.Communicate(text, "zh-CN-XiaoxiaoNeural")
    await communicate.save(f"sounds/{file_name}.mp3")
    print(f"{file_name}用时：{int(time.time() - start_time)}秒")
async def main():
    # 创建输出文件夹
    if not os.path.exists("sounds"):
        os.makedirs("sounds")
    tasks = []
    # 遍历输入文件夹
    for root, dirs, files in os.walk("texts"):
        for file in files:
            # 将文件中的文本读出来
            with open(os.path.join(root, file), "r", encoding='utf-8') as text_file:
                text = text_file.read()
                # 获取文件名
                file_name, ext = os.path.splitext(file)
                # 加入任务列表
                tasks.append(convert(text, file_name))
    # 等待所有任务完成
    await asyncio.gather(*tasks)


asyncio.run(main())

from pydub import AudioSegment
#读取音频文件
audio1 = AudioSegment.from_file("file1.wav")
audio2 = AudioSegment.from_file("file2.wav")
#拼接音频文件
merged_audio0=audio1+audio2
#保存拼接后的音频文件
merged_audio.export("merged.wav", format="wav")

import asyncio
import edge_tts
import time
from pydub import AudioSegment
import hashlib

pwd = "/var/www/html/sound"

rate = '+0%'
volume = '+0%'
pitch = '+0Hz'
def calculate_8_digit_md5(input_string):
    md5 = hashlib.md5()
    md5.update(input_string.encode('utf-8'))
    md5_hash = md5.hexdigest()[:8]
    return md5_hash
def get_current_time():
    current_time = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime())
    return current_time


async def convert(text, file_name, count,voice_name) -> None:
    global result
    start_time = time.time()
    communicate = edge_tts.Communicate(text=text,voice=voice_name,rate=rate,volume=volume,pitch=pitch)
    await communicate.save(pwd+f"/cache/{file_name}{count}.mp3")
    print(f"{file_name}用时：{int(time.time() - start_time)}秒")
    
def split_and_terminate(input_string, chunk_size=3000):
    res = []
    start = 0
    while start < len(input_string):
        # 截取指定大小的片段
        chunk = input_string[start:start + chunk_size]
        if start + chunk_size < len(input_string):
            last_comma = chunk.rfind('，')
            last_period = chunk.rfind('。')
            last_punctuation = max(last_comma, last_period)
            if last_punctuation != -1:
                chunk = chunk[:last_punctuation + 1]
        res.append(chunk)
        start += len(chunk)
    return res

async def main():
    print("md5:"+md5)
    result = split_and_terminate(text)
    tasks = []
    for index, chunk in enumerate(result):
        tasks.append(convert(chunk, md5,str(index+1),"zh-CN-XiaoxiaoNeural"))
    await asyncio.gather(*tasks)

def save_string_to_file(data, file_path):
    try:
        with open(file_path, 'w', encoding='utf-8') as file:
            file.write(data)
        print(f"字符串已成功保存到文件: {file_path}")
    except Exception as e:
        print(f"保存文件时发生错误: {e}")

if __name__ == "__main__":
    global text,md5
    with open(pwd+"/texts/测试.txt", "r", encoding='utf-8') as text_file:
        text = text_file.read()
        md5 = calculate_8_digit_md5(text + str(time.time()))
    asyncio.run(main())
    print('音频下载完成')
    result = AudioSegment.from_file(pwd+"/cache/" + md5 + "1.mp3")
    print('第1段合成完成')
    count = len(split_and_terminate(text))
    for i in range(2,count+1):
        result += AudioSegment.from_file(pwd+"/cache/" + md5 + str(i) + ".mp3")
        print('第' + str(i) + '段合成完成')
    result.export(pwd+"/download/"+md5+".mp3", format="mp3")
    print(md5+".mp3生成成功")