def get_words(file_path):
"""
TODO DOCUMENTATION
:param file_path:
:return:
"""
info_text = read_json(file_path)
words = [
k['text'].lower() for i in info_text for j in i['voices']
for k in j['trans']
]
return save_list(words, os.path.join(get_tmp_folder(), 'words.txt'))
def check_up(self):
"""
TODO DOCUMENTATION
:return:
"""
check_audio_file(self.audio_path)
self.audio_name = path.basename(self.audio_path).split('.')[0]
audio_conv_path = convert_audio(
self.audio_path, path.join(get_tmp_folder(), self.audio_name),
self.output_samplerate, self.output_n_channels,
self.output_bitdepth, get_working_format())
self.raw_audio = audio_conv_path
def generate_dic(words_path):
destination_folder = os.path.join(get_tmp_folder(), 'words.dic')
script = ['']
def get_sentences(file_path):
info_text = read_json(file_path)
sentences = [[k['text'].lower() for k in j['trans']] for i in info_text
for j in i['voices']]
return save_list(sentences, os.path.join(get_tmp_folder(),
'sentences.txt'), 1)
def noise_removal(origin_path, destination_path="", return_json=True):
"""
TODO DOCUMENTATION
:param return_json:
:param origin_path:
:param destination_path:
:return:
"""
gc.collect()
if destination_path == "":
destination_path = origin_path
original_audio = AudioSegment.from_wav(origin_path)
original_audio_splited = list(original_audio[::1])
audio_chunks = []
noise_chunks = []
max_noise_duration = get_audio_config()['max_noise_duration']
max_noise_level = get_audio_config()['max_noise_level']
aux_noise = []
for i, chunk in enumerate(original_audio_splited):
data = (i, chunk)
if chunk.dBFS > max_noise_level:
if len(aux_noise) > max_noise_duration:
noise_chunks += aux_noise
else:
audio_chunks += aux_noise
aux_noise.clear()
audio_chunks.append(data)
else: # SILENCE
aux_noise.append(data)
if len(noise_chunks) > 100000:
break
if len(aux_noise) > max_noise_duration:
noise_chunks += aux_noise
else:
audio_chunks += aux_noise
gc.collect()
try:
raw_noise = sum([*zip(*noise_chunks)][1], AudioSegment.empty())
except IndexError:
raw_noise = AudioSegment.empty()
raw_noise_path = os.path.join(get_tmp_folder(), 'raw_noise.wav')
raw_noise.export(raw_noise_path, format='wav')
noise_reduction = sox.Transformer()
noise_prof_path = os.path.join(get_tmp_folder(), 'noise.prof')
noise_reduction.noiseprof(raw_noise_path, noise_prof_path)
noise_remove = sox.Transformer()
noise_tolerance = get_audio_config()['noise_tolerance']
noise_remove.noisered(noise_prof_path, noise_tolerance)
gc.collect()
if destination_path == origin_path:
destination_path_splited = os.path.splitext(destination_path)
destination_path = destination_path_splited[0] + '_2' + destination_path_splited[1]
noise_remove.build(origin_path, destination_path)
os.remove(origin_path)
os.rename(destination_path, origin_path)
else:
noise_remove.build(origin_path, destination_path)
if return_json:
from itertools import groupby, count
audio_list = [(list(g)) for k, g in
groupby([*zip(*audio_chunks)][0], key=lambda i, j=count(): i - next(j))]
return [(l[0], l[-1]) for l in audio_list]
def mix_audio_trans(audio_info, trans):
for i in audio_info:
i['voices'] = get_text(trans, i['start_time'], i['end_time'])
return save_json(audio_info, path.join(get_tmp_folder(), 'info.json'))