-
Notifications
You must be signed in to change notification settings - Fork 0
/
test.py
232 lines (216 loc) · 7.31 KB
/
test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
from model import Graph
import tensorflow as tf
from hyperparams import hyperparams
import numpy as np
import tqdm
from audio_taco import inv_mel_spectrogram
from utils import get_sp
from utils import get_f0
from utils import get_mel
import hparam as hparams
import pyworld as pw
from audio_taco import melspectrogram, preemphasis
import hparam as hparams
import librosa
import os
hp = hyperparams()
"""
合成要sp和f0
sp是网络预测的
找好以下内容：
原始的sp，ap，f0，编码过的sp
目标人的sp，ap，f0，编码过的sp
"""
def test():
fpath = './paralleldata/VCC2SF1/SF1|10001.wav'
aim_spkid = 'spk2'
new_wav_for_infer, sr = librosa.load(fpath, sr=hp.SR, mono=True, dtype=np.float32) # 返回音频信号值 & 采样率
ori_mel = melspectrogram(new_wav_for_infer, hparams) # 康的
# ori_mel = ori_mel.T
synwav = inv_mel_spectrogram(ori_mel, hparams)
synwav = synwav / np.abs(synwav).max() # 04.23这句话的意思是，将语音的范围控制在【-1，1】之间
print(f'synthesize done. path : ./convert_to_{aim_spkid}_test1.wav')
librosa.output.write_wav(f'./convert_to_{aim_spkid}_test22222.wav', synwav, sr=hp.SR)
if __name__ == '__main__':
test()
# def synthesis(ori_path, aim_mel):
# aim_spkid = 'spk2'
# print('synthesizing ...')
# wav, sr = librosa.load(ori_path, sr=hp.SR, mono=True, dtype=np.float64)
# """
# ` hp.SR = 22050
# ` mel = librosa.feature.melspectrogram(y=wav, sr=sr, n_mels=80, n_fft=1024, hop_length=200)
#
# def mel_to_audio(M, sr=22050, n_fft=2048, hop_length=512, win_length=None,
# window='hann', center=True, pad_mode='reflect', power=2.0, n_iter=32,
# length=None, dtype=np.float32, **kwargs):
# """
# # synwav = librosa.feature.inverse.mel_to_audio(aim_mel, sr=hp.SR, n_fft=1024, hop_length=200)
# synwav = inv_mel_spectrogram(aim_mel, hparams)
# print(f'synthesize done. path : ./convert_to_{aim_spkid}_test1.wav')
# librosa.output.write_wav(f'./convert_to_{aim_spkid}_test1.wav', synwav, sr=hp.SR)
# def main():
# print("line 69")
# # spkid是代表你想转化到哪个人 fpath是原始音频
# fpath = './paralleldata/VCC2SF1/SF1|10001.wav'
#
# print(fpath) # ./data/convert/TMM1/M10007.wav
# fpath = fpath.strip()
# new_wav_for_infer, sr = librosa.load(fpath, sr=hp.SR, mono=True, dtype=np.float32) # 返回音频信号值 & 采样率
#
# ori_mel = melspectrogram(new_wav_for_infer, hparams) # 斌的：原始说话人的sp，接下来用np.random.normal生成同样形状的
# ori_mel = ori_mel.T
# print('源说话人的mel.shape：')
# print(ori_mel.shape) # sp 是二维的！！！
#
# mode = 'infer'
# G = Graph(mode=mode)
# print('{} graph loaded.'.format(mode))
# saver = tf.train.Saver()
# with tf.Session(config=tf.ConfigProto(log_device_placement=False, allow_soft_placement=True)) as sess:
# try:
# print(f'Try to load trained model in {hp.MODEL_DIR} ...')
# saver.restore(sess, tf.train.latest_checkpoint(hp.MODEL_DIR))
# except:
# raise Exception(f'Load trained model failed in {hp.MODEL_DIR}, please check ...')
# finally:
# ori_mel = np.reshape(ori_mel, (-1, hp.CODED_DIM)) # 斌的：原始说话人的sp
# ori_mel_batch = np.expand_dims(ori_mel, axis=0) # 模型训练的时候是三维度的 # ???;变成【1，None，80】
#
# # aim_spkid_batch = np.array([[aim_spkid]])
# for j in tqdm.tqdm(range(1)):
# aim_out = sess.run(G.aim_out, {G.ori_mel: ori_mel_batch})
#
# aim_out = np.array(aim_out, dtype=np.float64) # 转换出来的 mel，三维的（第一个是batch=1）
# predict_mel = np.reshape(aim_out, (-1, hp.CODED_DIM)) # [None, 80]，转换成二维mel
# print("line 103 predict_mel.shape = " + str(predict_mel.shape))
# predict_mel = predict_mel.T
# # ori_new_f0 = np.random.normal(source_mean_f0, 1.0, predict_mel.shape[0])
#
# print('Sp predicted done.')
# synthesis(fpath, aim_mel=predict_mel, )
#
# if __name__ == '__main__':
# main()
# import tensorflow as tf
#
# with tf.variable_scope('V1', reuse=None):
# a1 = tf.get_variable(name='a1', shape=[1], initializer=tf.constant_initializer(1))
# a2 = tf.Variable(tf.random_normal(shape=[2, 3], mean=0, stddev=1), name='a2')
# with tf.variable_scope('V2', reuse=True):
# a3 = tf.get_variable(name='a1', shape=[1], initializer=tf.constant_initializer(1))
# a4 = tf.Variable(tf.random_normal(shape=[2, 3], mean=0, stddev=1), name='a2')
#
# with tf.Session() as sess:
# sess.run(tf.initialize_all_variables())
# print(a1.name)
# print(a2.name)
# print(a3.name)
# print(a4.name)
# import tensorflow as tf
#
# output_file = './paralleldata/train_data/0.tfrecord'
# for example in tf.python_io.tf_record_iterator(output_file):
# print(tf.train.Example.FromString(example))
#
# with tf.Session() as sess:
# dataset = tf.data.TFRecordDataset(output_file) # 加载TFRecord文件
# dataset = dataset.map(parse_fn2) # 解析data到Tensor
# dataset = dataset.repeat(1) # 重复N epochs
# dataset = dataset.batch(3) # batch size
#
# iterator = dataset.make_one_shot_iterator()
# next_data = iterator.get_next()
#
# while True:
# try:
# position, label = sess.run(next_data)
# print(position)
# print(label)
# except tf.errors.OutOfRangeError:
# break
# def func1():
# try:
# return 1
#
# # print(2)
# finally:
# return 2
#
#
# def func2():
# try:
# raise ValueError()
# except:
# return 1
# finally:
# return 3
#
#
# print(func1())
# print(func2())
# # new_mel = librosa.feature.melspectrogram(y=wav[24001:48000], sr=sr, n_mels=80, hop_length=240) # n_mels 默认128,改为80维
# # win size800，hop200，fft 1024
# import tensorflow as tf
# import glob
#
# # tmp = tf.get_variable('global_step', initializer=0, dtype=tf.int32, trainable=False)
# # print(tmp)
#
# data_path = 'paralleldata/VCC2TM1/*'
#
# # for name in glob.glob('paralleldata/VCC2TM1/*'):
# # print("********")
#
# wav_files = glob.glob(data_path)
# wav_feats = dict()
# print(type(wav_files))
# print(wav_feats)
#
# def foo(num):
# print("starting...")
# while num<10:
# num=num+1
# yield num
# for n in foo(0):
# print(n)
# # for循环本身的原理就是调用next函数
# # 所以不再需要调用 next() 或者 send() 函数
# # print(type(foo(0))) # <class 'generator'>
# print('*******')
# # g = foo(0)
# # print(next(g))
# # print(next(foo(0)))
#
# # def foo():
# # print("starting...")
# # while True:
# # res = yield 4
# # print("res:",res)
# # g = foo()
# # print(next(g))
# # print("*"*20)
# # print(next(g))
#
# # class Fib(object):
# # def __init__(self):
# # pass
# #
# # def __call__(self, num):
# # a, b = 0, 1;
# # self.l = []
# #
# # for i in range(num):
# # self.l.append(a)
# # a, b = b, a + b
# # return self.l
# #
# # def __str__(self):
# # return str(self.l)
# #
# # __rept__ = __str__
# #
# #
# # f = Fib()
# # print(type(f(10).__repr__()))
# # print(f(10).__repr__())