def AUDIO(s, audios, intervals, times, gain=None, epsilon=.4): for idx, audio in enumerate(audios): if idx == 0: a = AudioSegment.from_mp3(audio)[1000 * intervals[idx][0]:1000 * (intervals[idx][1] + epsilon)] s.renderer.file_writer.add_audio_segment(a, times[idx]) else: a = AudioSegment.from_mp3(audio)[1000 * intervals[idx][0]:1000 * (intervals[idx][1] + epsilon)] s.renderer.file_writer.add_audio_segment(a, times[idx], gain)
def download(url):
options = {
'format':
'bestaudio/best',
'postprocessors': [{
'key': 'FFmpegExtractAudio',
'preferredcodec': 'mp3',
'preferredquality': '192',
}],
'outtmpl':
'largeout.webm', # name the file the ID of the video
'logger':
MyLogger(),
'progress_hooks': [my_hook],
}
with youtube_dl.YoutubeDL(options) as ydl:
info = ydl.extract_info(url,
download=False) # don't download, much faster
description = info['description']
totalDuration = info['duration']
yt_name = info['title']
ydl.download([url])
entries = Parsers.catalystTextParse("catalystdesc.txt", totalDuration)
#entries = Parsers.dubstepParse(description, totalDuration)
print("Loading file...")
# load sound
sound = AudioSegment.from_mp3("largeout.mp3")
split_export(entries, sound, yt_name)
def get_second_part_wav(main_wav_path, start_time, end_time, part_wav_path):
'''
音频切片，获取部分音频 单位是秒级别
:param main_wav_path: 原音频文件路径
:param start_time: 截取的开始时间
:param end_time: 截取的结束时间
:param part_wav_path: 截取后的音频路径
:return:
'''
start_time = int(start_time) * 1000
end_time = int(end_time) * 1000
sound = AudioSegment.from_mp3(main_wav_path)
word = sound[start_time:end_time]
word.export(part_wav_path, format="wav")
def getVoice(self, text, filename):
# 2. 向Rest接口提交数据
data=parse.urlencode([
('tex',text),
('lan','zh'),
('cuid',self.cu_id),
('ctp',1),
('tok',self.token_str)
])
f=urlopen(Request(self.getvoice_url),data = data.encode('utf-8'))
voice_fp = open(self.locpla+filename,'wb')
voice_fp.write(f.read())
voice_fp.close()
f.close()
sound = AudioSegment.from_mp3(self.locpla + "turing.mp3")
sound.export(self.locpla + 'turing.wav', format="wav")
def wav_split(path):
file = vedio_to_wav(path)
main_wav_path = file
path = os.path.dirname(file) + '/'
sound_len = int(
MediaInfo.parse(main_wav_path).to_data()['tracks'][0]['duration'])
sound = AudioSegment.from_mp3(main_wav_path)
part_file_list = list()
min_ = sound_len / 1000
if min_ > 60:
n = int(min_ // 60)
print(type(n))
if n * 60 < min_:
n += 1
for i in range(n):
start_time = i * 60 * 1000 + 1
end_time = (i + 1) * 60 * 1000
if end_time > sound_len * 1000:
end_time = sound_len * 1000
word = sound[start_time:end_time]
part_file_name = 'part_sound_{}.wav'.format(i)
word.export(part_file_name, format="wav")
part_file_list.append(part_file_name)
return part_file_list
from scrampy.splice import parse_splits, update_splits, aud_from_log, expand_data, insert_gaps
from pydub.audio_segment import AudioSegment
# TODO: move to setup function (in a class)
# load load and audio
log = parse_splits('examples/aud/NTF/NotTheFall_sentence_splitpoints.txt')
#log['order'] = range(0, len(log))
audio = {'NTF': AudioSegment.from_mp3('examples/aud/NTF/NotTheFall.mp3')}
# shuffle
indx = [1,0]
#random.shuffle(indx)
new_log = log.iloc[indx].copy()
# make new audio
audlist = aud_from_log(new_log, **audio)
out = reduce(lambda x,y: x + y, audlist)
# update start and end points on new log
update_splits(new_log)
unscram_log = new_log.sort('order')
get_clip = lambda clip, log, ii: clip[log['start'][ii] : log['end'][ii]]
def test_update_splits():
"""First row of Updated splits has start value of 0"""
assert new_log.iloc[0]['start'] == 0
def test_new_log_indx_order():
"""New log starts with segment 1 (i.e. the second segment)"""
def text2wav(text, language='en', filename='temp', tld='cn'):
gTTS(text=text, tld=tld, lang=language).save(filename + ".mp3")
AudioSegment.from_mp3(filename + ".mp3").set_frame_rate(16000).export(
filename + ".wav", format="wav")
fw.setsampwidth(sampwidth)
fw.setframerate(framerate)
fw.writeframes(wave_data[left * windows:right *
windows].tostring())
fw.close()
i = nextleft_tem - 1
break
j = right_tem + 1
i = i + 1
if __name__ == '__main__':
# input mono channel wav file
os.chdir(project_path + '/resource/')
for test_speech in glob.glob('*.mp3'):
fr = AudioSegment.from_mp3(test_speech)
framerate, nframes, sampwidth = fr.frame_rate, fr.frame_count(
), fr.sample_width
index = fr.split_to_mono()
num = 1
# output record
file = open(
project_path + '/speech_separation/' +
os.path.splitext(os.path.basename(test_speech))[0] + '.txt', 'w')
for mp3_data in index:
exit_value = False
mp3_data_array = mp3_data.get_array_of_samples()
# wave_data = np.fromstring(mp3_data_array, dtype=np.short)
speech_separate(mp3_data_array)
num += 1
file.close()
def mp2Wav(source_file_path, destin_path):
sound = AudioSegment.from_mp3(source_file_path)
sound.export(destin_path, format='wav')
encoded_samples.append(encoded_sample)
encoded_audio = wave.open(watermarked_output_path, 'wb')
encoded_audio.setparams( (nchannels, sampwidth, framerate, nframes, comptype, compname) )
#设置参数
encoded_audio.writeframes(struct.pack("%dh" % len(encoded_samples), *encoded_samples))
#写入output路径
def watermark_to_bits(self,watermark, nbits=8):
watermark_bits = []
for byte in watermark:
for i in range(0,nbits):
watermark_bits.append( (byte & (2 ** i)) >> i )
return watermark_bits
if __name__ == "__main__":
cover_audio = "ori.wav"
output = "w.wav"
source = sys.argv[2]
sound = AudioSegment.from_mp3(source)
sound.export(cover_audio, format = 'wav')
if len(sys.argv) > 1:
message = sys.argv[1]
if len(sys.argv) > 3:
output = sys.argv[3]
# lsb_watermark(cover_audio, message, output)
#print(recover_lsb_watermark(output))
import os
import sys
import pandas
from scipy.io import wavfile
from python_speech_features.base import mfcc
from pydub.audio_segment import AudioSegment
# 音频的截取，通过时间片段来获取部分音频
# 因为音频以毫秒计时，在截取音频时，统统转为了毫秒
AudioSegment.converter = r'../../software/ffmpeg/bin/ffmpeg.exe'
# song = AudioSegment.from_file('./a.MP3', format='MP3') # 测试只支持wav格式，不支持MP3
sound = AudioSegment.from_mp3("a.wav") # 测试只支持wav格式，不支持MP3
part = sound[21500:100000]
part.export("./part.wav", format="wav")
print(sound)
# 比较大的音频文件，将分钟和秒进行结合，然后在自己拆分
def segeS(start_time, end_time):
start_time = (int(start_time.split(":")[0])*60 + int(start_time.split(":")[1])) * 1000
end_time = (int(end_time.split(':')[0])*60+int(end_time.split(':')[1]))*1000
# 格式
# start_time = "0:35"
# end_time = "0:38"
# https://blog.csdn.net/xuqingda/article/details/86540333
# python+ffmpeg实现wav文件切割
# 单文件切割成小文件，这样实现了数据容量的扩增。
import os
# _*_ coding: utf-8 _*_
# @Time : 2018/4/19 下午3:42
# @Author : yu yongsheng
# @FileName : channel_split_pydub2.py.py
# @Software :
# @Description: split_channel_pydub
import os
import time
import glob
import numpy as np
from pydub.audio_segment import AudioSegment
project_path = os.getcwd()
st = time.time()
os.chdir(project_path + '/speech_separate_data/')
for path in glob.glob('*.mp3'):
print path
mp3_file = AudioSegment.from_mp3(path)
# mp3_file = mp3_file.set_frame_rate(22050)
# each index in split_to_mono() represent a channel(0-left/1-right)
index = mp3_file.split_to_mono()
index[0].export(project_path + "/channel_data/" +
os.path.splitext(os.path.basename(path))[0] + '_left.wav',
format="wav")
index[1].export(project_path + "/channel_data/" +
os.path.splitext(os.path.basename(path))[0] + '_right.wav',
format="wav")
print time.time() - st