语音翻译项目(Whisper，opencc，Baidu翻译API，edge_tts)

定制程序开发

heguannan @ 163.com

一、项目背景

国内天生有墙，海外与国内的视频平台不连通，视频搬运业务应运而生。 通过国内的视频搬运到海外平台或者将海外平台视频搬运到国内，以获得平台的报酬。

二、技术路线

当然，可以直接采用成熟的API直接做到 汉语语音<->英语语音。然而，大多数API都要收费，且价格不便宜。Google和Microsoft 一个月只有5条音频能够试用。不能满足搬运的需求。

本文通过

1.开源离线模型whisper 完成视频字幕的提取，可以获得视频的字幕开始时间，字幕结束时间和字幕内容

2.opencc进行简体和繁体中文转换

3.Baidu翻译API （每月免费100万个字符，能够满足需求） 进行汉英互相翻译

4.edge_tts：采用微软的免费API进行Text to Speech 进行配音。

三、具体代码

程序分为S1和S2。

S1从中文视频中提取中文字幕文件，由于识别的字幕可能有错，文件生成后，可以进行人工修改。

修改完成后，再运行S2，就可以得到英语配音。

# -*- coding: utf-8 -*-
"""
Created on Sun Jun 18 16:17:41 2023

@author: hegua

脚本1：
1_EA:获取音频/没有音频的视频并保存在Temp文件夹中。
2_TS:转换为繁体中文字幕并转化为简体
3_SS:保存中间文件

"""

from moviepy.editor import VideoFileClip
import whisper
import opencc




            
##1_EA
video_name = "0618"
video_path = rf"./Ori/"+ video_name + ".mp4"
# 读取视频文件
video = VideoFileClip(video_path)

# 提取视频和音频
audio = video.audio
video = video.without_audio()

temp_path = rf"./Temp/"+ video_name 
vwa_path = rf"./Res/"+ video_name 
# 保存提取出来的视频和音频
video.write_videofile(vwa_path+".mp4")
audio.write_audiofile(temp_path+".mp3")

##2_TS : 
whisper_path = temp_path +".mp3"
model = whisper.load_model("medium")
whisper_result = model.transcribe(audio = whisper_path,language = "Chinese")

converter = opencc.OpenCC("t2s.json")
idx = 0
while idx < len(whisper_result['segments']):
    t2s_res = converter.convert(whisper_result['segments'][idx]['text'])
    whisper_result['segments'][idx]['text'] = t2s_res
    idx += 1
   
##3_SS

# 字幕开始时间、结束时间和内容列表
subs = whisper_result['segments']

tup_list = []
for sub in subs:
    tup1 =(sub['start'],sub['end'],sub['text'])
    tup_list.append(tup1)
    
    
srt_path = temp_path +'.txt'
with open(srt_path, 'w') as f:
    for i, (start, end, text) in enumerate(tup_list):
        f.write(str(start)+","+str(end) +"," +text + '\n')
           
            

# -*- coding: utf-8 -*-
"""
Created on Sun Jun 18 17:49:47 2023

@author: hegua

## 1_读取修正的字幕文本文件
## 2_进行翻译
## 2_进行配音
## 3_生成音频文件

"""
import http.client
import hashlib
import urllib
import random
import json
import time

##利用微软的TTS API进行合成
import edge_tts
import asyncio
import nest_asyncio
nest_asyncio.apply()

##语音集合
from pydub import AudioSegment

async def TTS(TEXT,output,rate):
    voice = 'en-US-EricNeural'
    volume = '+30%'
    tts = edge_tts.Communicate(text=TEXT, voice=voice, rate=rate, volume=volume)
    print(output+'TTS保存成功')
    await tts.save(output)
    
    

def BaiduTrans(q,idx,fromLang = 'zh', toLang ='en'):
    appid = ''  # 填写你的appid
    secretKey = ''  # 填写你的密钥
    httpClient = None
    myurl = '/api/trans/vip/translate'
    salt = random.randint(32768, 65536)
    sign = appid + q + str(salt) + secretKey
    sign = hashlib.md5(sign.encode()).hexdigest()
    myurl = myurl + '?appid=' + appid + '&q=' + urllib.parse.quote(q) + '&from=' + fromLang + '&to=' + toLang + '&salt=' + str(
    salt) + '&sign=' + sign

    try:
        httpClient = http.client.HTTPConnection('api.fanyi.baidu.com')
        httpClient.request('GET', myurl)

        # response是HTTPResponse对象
        response = httpClient.getresponse()
        result_all = response.read().decode("utf-8")
        result = json.loads(result_all)
    except Exception as e:
        print (e)
    finally:
        if httpClient:
            httpClient.close()
            print(str(idx)+" 翻译成功")
            return result



video_name = '0618'

file_path = rf"./Temp/"+ video_name + '.txt'

dub_path = rf"./Temp/"+ video_name 

audio_segments = []
with open(file_path, 'r') as file:
    content = file.read()
        # 按照空行将内容分割为多个字幕记录
    records = content.split('\n')
    idx = 0
    while idx < len(records):
        if records[idx]:
            s = records[idx].split(',')[0]
            e = records[idx].split(',')[1]
            q = records[idx].split(',')[2]
            t = BaiduTrans(q,idx)
            text = t['trans_result'][0]['dst']
            start = float(s)
            end = float(e)
            #翻译的结果是t，进行Bark配音
            output = dub_path+'_'+str(idx)+".mp3"
            asyncio.run(TTS(text,output,'+5%'))
            ##读取语音，获取语音的成都
            # audio = AudioSegment.from_file(output)
            # duration = len(audio) / 1000  # 将毫秒转换为秒
            # if duration > end-start:
            #     rate = int(duration/(end-start)*100 + 10)
            #     rate = '+' + str(rate)+'%'
            #     asyncio.run(TTS(text,output,rate))
            audio_temp = {"file": output, "start":start , "end":end }
            # 将解析后的字幕记录添加到列表中
            audio_segments.append(audio_temp)
            #time.sleep(0.5)
        idx += 1
        

combined_audio = AudioSegment.silent(duration=0)
duration_all = 0 
for segment in audio_segments:
    audio = AudioSegment.from_file(segment["file"])
    duration = len(audio) / 1000
    duration_all = duration + duration_all
    end = segment["end"]
    if duration_all < end:
        audio = audio + AudioSegment.silent(duration= (end - duration_all)*1000)
        duration_all = end
    combined_audio += audio


combined_audio_path = rf"./Res/"+ video_name +'.wav'
combined_audio.export(combined_audio_path, format="wav")