Python 百度语音识别字幕生成器
2019-08-11 本文已影响0人
徐Teri
from moviepy.editor import AudioFileClip
from pydub import AudioSegment
from pydub.silence import detect_silence
from aip import AipSpeech
import os
import time
import re
import requests
import js2py
class baidu_Translate():
def __init__(self):
self.js = js2py.eval_js('''
var i = null;
function n(r, o) {
for (var t = 0; t < o.length - 2; t += 3) {
var a = o.charAt(t + 2);
a = a >= "a" ? a.charCodeAt(0) - 87 : Number(a),
a = "+" === o.charAt(t + 1) ? r >>> a: r << a,
r = "+" === o.charAt(t) ? r + a & 4294967295 : r ^ a
}
return r
}
var hash = function e(r,gtk) {
var o = r.match(/[\uD800-\uDBFF][\uDC00-\uDFFF]/g);
if (null === o) {
var t = r.length;
t > 30 && (r = "" + r.substr(0, 10) + r.substr(Math.floor(t / 2) - 5, 10) + r.substr( - 10, 10))
} else {
for (var e = r.split(/[\uD800-\uDBFF][\uDC00-\uDFFF]/), C = 0, h = e.length, f = []; h > C; C++)"" !== e[C] && f.push.apply(f, a(e[C].split(""))),
C !== h - 1 && f.push(o[C]);
var g = f.length;
g > 30 && (r = f.slice(0, 10).join("") + f.slice(Math.floor(g / 2) - 5, Math.floor(g / 2) + 5).join("") + f.slice( - 10).join(""))
}
var u = void 0,
u = null !== i ? i: (i = gtk || "") || "";
for (var d = u.split("."), m = Number(d[0]) || 0, s = Number(d[1]) || 0, S = [], c = 0, v = 0; v < r.length; v++) {
var A = r.charCodeAt(v);
128 > A ? S[c++] = A: (2048 > A ? S[c++] = A >> 6 | 192 : (55296 === (64512 & A) && v + 1 < r.length && 56320 === (64512 & r.charCodeAt(v + 1)) ? (A = 65536 + ((1023 & A) << 10) + (1023 & r.charCodeAt(++v)), S[c++] = A >> 18 | 240, S[c++] = A >> 12 & 63 | 128) : S[c++] = A >> 12 | 224, S[c++] = A >> 6 & 63 | 128), S[c++] = 63 & A | 128)
}
for (
var p = m,F = "+-a^+6", D = "+-3^+b+-f", b = 0;
b < S.length; b++) p += S[b],p = n(p, F);
return p = n(p, D),
p ^= s,
0 > p && (p = (2147483647 & p) + 2147483648),
p %= 1e6,
p.toString() + "." + (p ^ m)
}
''')
headers = {
'user-agent': 'Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/72.0.3626.96 Safari/537.36', }
self.session = requests.Session()
self.session.get('https://fanyi.baidu.com', headers=headers)
response = self.session.get('https://fanyi.baidu.com', headers=headers)
self.token = re.findall("token: '(.*?)',", response.text)[0]
self.gtk = '320305.131321201' # re.findall("window.gtk = '(.*?)';", response.text, re.S)[0]
def translate(self, query, from_lang='en', to_lang='zh'):
# langdetect
self.session.post('https://fanyi.baidu.com/langdetect', data={'query': query})
# clickEvent
self.session.get('https://click.fanyi.baidu.com/?src=1&locate=zh&action=query&type=1&page=1')
# translate
data = {
'from': from_lang,
'to': to_lang,
'query': query,
'transtype': 'realtime',
'simple_means_flag': '3',
'sign': self.js(query, self.gtk),
'token': self.token
}
response = self.session.post('https://fanyi.baidu.com/v2transapi', data=data)
json = response.json()
if 'error' in json:
pass
# return 'error: {}'.format(json['error'])
else:
return response.json()['trans_result']['data'][0]['dst']
class baidu_SpeechRecognition():
def __init__(self, dev_pid):
# 百度语音识别API
Speech_APP_ID = '自己的'
Speech_API_KEY = '自己的'
Speech_SECRET_KEY = '自己的'
self.dev_pid = dev_pid
self.SpeechClient = AipSpeech(Speech_APP_ID, Speech_API_KEY, Speech_SECRET_KEY)
self.TranslClient = baidu_Translate()
def load_audio(self, audio_file):
self.source = AudioSegment.from_wav(audio_file)
def speech_recognition(self, offset, duration, fanyi):
data = self.source[offset * 1000:duration * 1000].raw_data
result = self.SpeechClient.asr(data, 'wav', 16000, {'dev_pid': self.dev_pid, })
fanyi_text = ''
if fanyi:
try:
fanyi_text = self.TranslClient.translate(result['result'][0])
except:
pass
try:
return [result['result'][0], fanyi_text]
except:
# print('错误:',result)
return ['', '']
def cut_point(path, dbfs=1.25):
sound = AudioSegment.from_file(path, format="wav")
tstamp_list = detect_silence(sound, 600, sound.dBFS * dbfs, 1)
timelist = []
for i in range(len(tstamp_list)):
if i == 0:
back = 0
else:
back = tstamp_list[i - 1][1] / 1000
timelist.append([back, tstamp_list[i][1] / 1000])
min_len = 0.5
max_len = 5
result = []
add = 0
total = len(timelist)
for x in range(total):
if x + add < total:
into, out = timelist[x + add]
if out - into > min_len and out - into < max_len and x + add + 1 < total:
add += 1
out = timelist[x + add][1]
result.append([into, out])
elif out - into > max_len:
result.append([into, out])
else:
break
return result
def cut_text(text, length=38):
newtext = ''
if len(text) > length:
while True:
cutA = text[:length]
cutB = text[length:]
newtext += cutA + '\n'
if len(cutB) < 4:
newtext = cutA + cutB
break
elif len(cutB) > length:
text = cutB
else:
newtext += cutB
break
return newtext
return text
def progressbar(total, temp, text='&&', lenght=40):
content = '\r' + text.strip().replace('&&', '[{0}{1}]{2}%')
percentage = round(temp / total * 100, 2)
a = round(temp / total * lenght)
b = lenght - a
print(content.format('■' * a, '□' * b, percentage), end='')
def format_time(seconds):
sec = int(seconds)
m, s = divmod(sec, 60)
h, m = divmod(m, 60)
fm = int(str(round(seconds, 3)).split('.')[-1])
return "%02d:%02d:%02d,%03d" % (h, m, s, fm)
def separate_audio(file_path, save_path):
audio_file = save_path + '\\tmp.wav'
audio = AudioFileClip(file_path)
audio.write_audiofile(audio_file, ffmpeg_params=['-ar', '16000', '-ac', '1'], logger=None)
return audio_file
def file_filter(path, alldir=False):
key = ['mp4', 'mov']
if alldir:
dic_list = os.walk(path)
else:
dic_list = os.listdir(path)
find_list = []
for i in dic_list:
if os.path.isdir(i[0]):
header = i[0]
file = i[2]
for f in file:
for k in key:
if f.rfind(k) != -1:
find_list.append([header, f])
else:
for k in key:
if i.rfind(k) != -1:
find_list.append([path, i])
if find_list:
find_list.sort(key=lambda txt: re.findall(r'\d+', txt[1])[0])
return find_list
def countTime(s_time, now=True):
if now: s_time = (time.time() - s_time)
m, s = divmod(int(s_time), 60)
return '{}分{}秒'.format('%02d' % (m), '%02d' % (s))
def __line__print__(txt='-' * 10):
print('\n' + '-' * 10 + ' ' + txt + ' ' + '-' * 10 + '\n')
if __name__ == '__main__':
def StartHandle(timeList, save_path, srt_mode=2, result_print=False):
index = 0
total = len(timeList)
a_font = r'{\fn微软雅黑\fs14}'
b_font = r'{\fn微软雅黑\fs10}'
fanyi = False if srt_mode == 1 else True
file_write = open(save_path, 'a', encoding='utf-8')
for x in range(total):
into, out = timelist[x]
timeStamp = format_time(into - 0.2) + ' --> ' + format_time(out - 0.2)
result = baidufanyi.speech_recognition(into + 0.1, out - 0.1, fanyi)
if result_print:
if srt_mode == 0:
print(timeStamp, result[0])
else:
print(timeStamp, result)
else:
progressbar(total, x, '识别中...&& - {0}/{1}'.format('%03d' % (total), '%03d' % (x)), 44)
if len(result[0]) > 1:
index += 1
text = str(index) + '\n' + timeStamp + '\n'
if srt_mode == 0: # 仅中文
text += a_font + cut_text(result[1])
elif srt_mode == 1: # 仅英文
text += b_font + cut_text(result[0])
else: # 中文+英文
text += a_font + cut_text(result[1]) + '\n' + b_font + result[0]
text = text.replace('\u200b', '') + '\n\n'
file_write.write(text)
file_write.close()
if not result_print:
progressbar(total, total, '识别中...&& - {0}/{1}'.format('%03d' % (total), '%03d' % (total)), 44)
os.system('cls')
wav_path = os.environ.get('TEMP')
# 语音模型
pid_list = 1536, 1537, 1737, 1637, 1837, 1936
# 设置参数
print('[ 百度语音识别字幕生成器 - by Teri ]\n')
__line__print__('1 模式选择')
input_dev_pid = input('请选择识别模式:\n'
'\n (1)普通话,'
'\n (2)普通话+简单英语,'
'\n (3)英语,'
'\n (4)粤语,'
'\n (5)四川话,'
'\n (6)普通话-远场'
'\n\n请输入一个选项(默认3):')
__line__print__('2 字幕格式')
input_srt_mode = input('请选择字幕格式:\n'
'\n (1)中文,'
'\n (2)英文,'
'\n (3)中文+英文，'
'\n\n请输入一个选项(默认3):')
__line__print__('3 实时输出')
input_print = input('是否实时输出结果到屏幕? (默认:否/y:输出):').upper()
# 处理参数
dev_pid = int(input_dev_pid) - 1 if input_dev_pid else 3
dev_pid -= 1
srt_mode = int(input_srt_mode) if input_srt_mode else 3
srt_mode -= 1
re_print = True if input_print == 'Y' else False
# 输入文件
__line__print__('4 打开文件')
input_file = input('请拖入一个文件或文件夹并按回车:').strip('"')
video_file = []
if not os.path.isdir(input_file):
video_file = [input_file]
else:
file_list = file_filter(input_file)
for a, b in file_list:
video_file.append(a + '\\' + b)
# 执行确认
select_dev = ['普通话', '普通话+简单英语', '英语', '粤语', '四川话', '普通话-远场']
select_mode = ['中文', '英文', '中文+英文']
__line__print__('5 确认执行')
input('当前的设置:\n识别模式: {0}, 字幕格式: {1}, 输出结果: {2}\n当前待处理文件 {3} 个\n请按下回车开始处理...'.format(
select_dev[dev_pid],
select_mode[srt_mode],
'是' if re_print else '否',
len(video_file)
))
# 批量处理
total_file = len(video_file)
total_time = time.time()
baidufanyi = baidu_SpeechRecognition(pid_list[dev_pid])
for i in range(total_file):
item_time = time.time()
file_name = video_file[i].split('\\')[-1]
print('\n>>>>>>>> ...正在处理音频... <<<<<<<<', end='')
audio_file = separate_audio(video_file[i], wav_path)
timelist = cut_point(audio_file, dbfs=1.15)
if timelist:
print('\r>>>>>>>> 当前:{} 预计:{} <<<<<<<<'.format(
'%03d' % (i),
countTime(len(timelist) * 5, now=False)
))
srt_name = video_file[i][:video_file[i].rfind('.')] + '.srt'
baidufanyi.load_audio(audio_file)
StartHandle(timelist, srt_name, srt_mode, re_print)
print('\n{} 处理完成, 本次用时{}'.format(file_name, countTime(item_time)))
else:
print('音频参数错误')
# 执行完成
input('全部完成, 处理了{}个文件, 全部用时{}'.format(total_file, countTime(total_time)))
前往这里创建API：https://login.bce.baidu.com/?account=
点击-语音技术
点击-创建应用
随便填
得到KEY
修改这里
