用百度的语音识别胡搞出了个坑爹的1.0版，一个小程序。可以把一段文字转换成语音，也可以把一堆视频、音频文件转换成文字，功能如图：
Paste_Image.png
不知道简书会不会把代码给替换了，总之欢迎来喷，东拼西凑，加载一堆，代码混乱，没有多线程，没有异常处理，暴力修bug，写的什么玩意233
用到ffmpeg，需要把这个下载到同一目录下。
以下是代码：
import os
import sys
import math
import json
import base64
import urllib
import shutil
import subprocess
import configparser
import urllib.request
from tinytag import TinyTag
from subprocess import run
class BaiduRest:
"""百度接口模块"""
def __init__(self, cu_id, api_key, api_secert):
# token认证的url
self.token_url = "https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id=%s&client_secret=%s"
# 语音合成的resturl
self.getvoice_url = "http://tsn.baidu.com/text2audio?tex=%s&lan=zh&cuid=%s&ctp=1&tok=%s"
# 语音识别的resturl
self.upvoice_url = 'http://vop.baidu.com/server_api'
self.cu_id = cu_id
self.getToken(api_key, api_secert)
return
def getToken(self, api_key, api_secert):
# 1.获取token
token_url = self.token_url % (api_key,api_secert)
r_str = urllib.request.urlopen(token_url).read()
token_data = json.loads(r_str)
self.token_str = token_data['access_token']
print ("token获取完成...")
pass
def getVoice(self, text, filename):
# 2. 向Rest接口提交数据
get_url = self.getvoice_url % (urllib.parse.quote(text), self.cu_id, self.token_str)
voice_data = urllib.request.urlopen(get_url).read()
# 3.处理返回数据
voice_fp = open("output\\"+filename,'wb+')
voice_fp.write(voice_data)
voice_fp.close()
pass
def getText(self, filename):
# 2. 向Rest接口提交数据
repeat_times = 10
for x in range(1,10):
for y in range(repeat_times):
try:
data = {}
# 语音的一些参数
data['format'] = 'amr'
data['rate'] = 16000
data['channel'] = 1
data['cuid'] = self.cu_id
data['token'] = self.token_str
print("正在上传："+filename)
wav_fp = open("temp\\"+filename,'rb')
voice_data = wav_fp.read()
data['len'] = len(voice_data)
data['speech'] = base64.b64encode(voice_data).decode('utf-8')
post_data = json.dumps(data)
r_data = urllib.request.urlopen(self.upvoice_url,data=bytes(post_data,encoding="utf-8")).read()
wav_fp.close()
return r_data
break
except :
print ("正在重试...")
sleep(1)
bdr = BaiduRest("Black", api_key, api_secert)
pass
pass
pass
class VoiceProcessing:
"""声音处理模块"""
def __init__(self, filename, durtime):
self.filename = filename
self.durtime = durtime
return
def preprocess(self, filename,durtime):
wfilename = self.filename
if os.path.exists("ffmpeg.exe"):
#预处理
print ("音频文件预处理中...")
suflen = len(os.path.splitext(wfilename)[1])
endname = wfilename[:-suflen]+".mp3"
#转换
parameter = "ffmpeg.exe -n -i input\\"+wfilename+" -acodec mp3 -ac 1 -ar 16000 -vn temp\\"+endname
run(parameter,shell=True)
#预处理结束，获取信息
wavetag = TinyTag.get("temp\\"+endname)
durtime = round(wavetag.duration,2)
sepparts = math.ceil(durtime/59)
timeleft = durtime%59
print("总时长:"+str(durtime)+"上传分段为："+str(sepparts)+"末段时长："+str(timeleft))
os.remove("temp\\"+endname)
return durtime
else:
print ("请把ffmpeg放到本目录后重试...")
pass
def transvoid(self, filename, durtime):
tfilename = self.filename
#音频转换部分
rdurtime = VoiceProcessing(tfilename,0).preprocess(tfilename,0)
durtime = rdurtime
suflen = len(os.path.splitext(tfilename)[1])
#初始化数据
set_parts = 0
start_time = 0
end_time = 59
#进程
if durtime < 59:
end_time = durtime
endname = tfilename[:-suflen]+str(set_parts)+".amr"
print ("正在转换："+endname+"...")
parameter = "ffmpeg.exe -n -i input\\"+tfilename+" -ss "+str(start_time)+" -t "+str(end_time)+" -acodec amr_wb -ac 1 -ar 16000 -vn temp\\"+endname
run(parameter,shell=True)
return rdurtime
else:
while durtime >=59:
endname = tfilename[:-suflen]+str(set_parts)+".amr"
parameter = "ffmpeg.exe -n -i input\\"+tfilename+" -ss "+str(start_time)+" -t "+str(end_time)+" -acodec amr_wb -ac 1 -ar 16000 -vn temp\\"+endname
start_time = start_time + 59
end_time = end_time + 59
durtime -= 59
set_parts += 1
print("正在处理："+endname+"剩余时长："+str(durtime)+"分段："+str(set_parts))
run(parameter,shell=True)
pass
else:
endname = tfilename[:-suflen]+str(set_parts)+".amr"
end_time = durtime
parameter = "ffmpeg.exe -n -i input\\"+tfilename+" -ss "+str(start_time)+" -t "+str(end_time)+" -acodec amr_wb -ac 1 -ar 16000 -vn temp\\"+endname
run(parameter,shell=True)
print ("音频文件转换完成...")
return rdurtime
class filesmanage:
"""文件处理模块"""
def __init__(self, filename):
pass
def comptxt(filename):
#合并文本文档
meragefiledir = os.getcwd()+"\\temp"
finalfiledir = os.getcwd()+"\\output"
suflen = len(os.path.splitext(filename)[1])
txtname = filename[:-suflen]+".txt"
#列举
filenames=os.listdir("temp")
file=open(finalfiledir+"\\"+txtname,"w")
#遍历文件名
for files in filenames:
filepath=meragefiledir+'\\'+files
#遍历单个文件，读取行数
for line in open(filepath):
file.writelines(line)
file.write('\n')
file.close()
pass
if __name__ == "__main__":
#初始化加载API配置文件
config = configparser.ConfigParser()
if os.path.exists("apiconfig.cfg"):
config.read("ApiConfig.cfg")
api_key = config.get("api","api_key")
api_secert = config.get("api","api_secert")
print("API加载成功！")
else:
api_key = input("请输入Api_Key：\n")
api_secert = input("请输入Secrect_Key：\n")
config.write(open(r"ApiConfig.cfg","w"))
config.read("ApiConfig.cfg")
config.add_section("api")
config.set("api","api_key",api_key)
config.set("api","api_secert",api_secert)
config.write(open(r"ApiConfig.cfg","w"))
#建立文件夹
if not os.path.isdir("input"):
os.mkdir("input")
if not os.path.isdir("output"):
os.mkdir("output")
if not os.path.isdir("temp"):
os.mkdir("temp")
#内容初始化
bdr = BaiduRest("Black", api_key, api_secert)
print (" =======================================")
print (" | 视频&语音转成文字 v1.0版 by:Black |")
print (" | —————————————— |")
print (" | 将文本文件放入“input”， |")
print (" | 程序会自动将文字转变成语音。 |")
print (" | |")
print (" | 将视频或音频文件放入“input”， |")
print (" | 程序会将其中的语音转换成文字。 |")
print (" | |")
print (" | 转换完成的文件会出现在“output”。 |")
print (" | |")
print (" | 懒得修bug，出错就改文件名不要空格 |")
print (" =======================================")
input ("放入文件后，按回车键继续..")
print (" ----------------------------")
#先语音合成
procfile = os.listdir("input")
txtlist = []
for names in procfile:
if names.endswith(".txt"):
txtlist.append(names)
for txtname in txtlist:
readtxt = open(txtname,'r')
texts = readtxt.read()
readtxt.close()
suflen = len(os.path.splitext(txtname)[1])
mp3name = txtname[:-suflen]+".mp3"
bdr.getVoice(texts, mp3name)
#再语音识别
procfile = os.listdir("input")
wavlist = []
templist = []
count = [0,0,0]
#语音分段处理
for names in procfile:
if not names.endswith(".txt"):
wavlist.append(names)
for wavename in wavlist:
waveinfo = VoiceProcessing(wavename,0).transvoid(wavename,0)
#数据上传
sepparts = math.ceil(waveinfo/59)
suflen = len(os.path.splitext(wavename)[1])
for i in range(0,sepparts):
wavetemp = wavename[:-suflen]+str(i)+".amr"
txttemp = wavename[:-suflen]+".txt"
r_data = bdr.getText(wavetemp).decode('utf-8')
t_data = str(r_data)
f_data = eval(t_data)
print ("返回结果："+f_data['err_msg'])
#内容存储
if f_data['err_msg']=='success.':
word = f_data['result'][0].encode('utf-8')
if word!='':
if word[len(word)-3:len(word)]=='，':
word = f_data['result'][0]
with open("output\\%s"%txttemp,"a") as f:
f.write(str(word))
print (wavename+"转换成功!但内容可能为空.")
f.close()
count[0] +=1
count[1] +=1
else:
word = f_data['result'][0]
with open("output\\%s"%txttemp,"a") as f:
f.write(str(word))
f.close()
count[0] +=1
count[1] +=1
else:
print ("音频文件不存在或格式错误...")
count[2] +=1
else:
print ("未知错误，请重试...")
count[2] +=1
os.remove("temp\\%s"%wavetemp)
shutil.rmtree(r"temp")
print ("--------------转换完成！---------------")
print ("总计转换文件数："+str(count[0])+"成功数"+str(count[1])+"失败数"+str(count[2]))
网友评论