forked from NVIDIA/waveglow
-
Notifications
You must be signed in to change notification settings - Fork 1
/
mel2samp.py
148 lines (131 loc) · 5.98 KB
/
mel2samp.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
# *****************************************************************************
# Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of the NVIDIA CORPORATION nor the
# names of its contributors may be used to endorse or promote products
# derived from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
# ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
# *****************************************************************************\
import os
import random
import argparse
import json
import torch
import torch.utils.data
import sys
from scipy.io.wavfile import read
import librosa
# We're using the audio processing from TacoTron2 to make sure it matches
sys.path.insert(0, 'tacotron2')
from tacotron2.layers import TacotronSTFT
MAX_WAV_VALUE = 32768.0
def files_to_list(filename):
"""
Takes a text file of filenames and makes a list of filenames
"""
with open(filename, encoding='utf-8') as f:
files = f.readlines()
files = [f.rstrip() for f in files]
return files
def load_wav_to_torch(full_path, use_librosa = False, sr=None):
"""
Loads wavdata into torch array
"""
if use_librosa :
data, sampling_rate = librosa.load(full_path, sr=sr)
else :
sampling_rate, data = read(full_path)
return torch.from_numpy(data).float(), sampling_rate
class Mel2Samp(torch.utils.data.Dataset):
"""
This is the main class that calculates the spectrogram and returns the
spectrogram, audio pair.
"""
def __init__(self, training_files, segment_length, filter_length,
hop_length, win_length, sampling_rate, mel_fmin, mel_fmax):
self.audio_files = files_to_list(training_files)
random.seed(1234)
random.shuffle(self.audio_files)
self.stft = TacotronSTFT(filter_length=filter_length,
hop_length=hop_length,
win_length=win_length,
sampling_rate=sampling_rate,
mel_fmin=mel_fmin, mel_fmax=mel_fmax)
self.segment_length = segment_length
self.sampling_rate = sampling_rate
def get_mel(self, audio):
audio_norm = audio / MAX_WAV_VALUE
audio_norm = audio_norm.unsqueeze(0)
audio_norm = torch.autograd.Variable(audio_norm, requires_grad=False)
melspec = self.stft.mel_spectrogram(audio_norm)
melspec = torch.squeeze(melspec, 0)
return melspec
def __getitem__(self, index):
# Read audio
filename = self.audio_files[index]
audio, sampling_rate = load_wav_to_torch(filename)
if sampling_rate != self.sampling_rate:
audio, sampling_rate = load_wav_to_torch(filename, True, self.sampling_rate)
'''
raise ValueError("{} SR doesn't match target {} SR".format(
sampling_rate, self.sampling_rate))
'''
# Take segment
if audio.size(0) >= self.segment_length:
max_audio_start = audio.size(0) - self.segment_length
audio_start = random.randint(0, max_audio_start)
audio = audio[audio_start:audio_start+self.segment_length]
else:
audio = torch.nn.functional.pad(audio, (0, self.segment_length - audio.size(0)), 'constant').data
mel = self.get_mel(audio)
audio = audio / MAX_WAV_VALUE
return (mel, audio)
def __len__(self):
return len(self.audio_files)
# ===================================================================
# Takes directory of clean audio and makes directory of spectrograms
# Useful for making test sets
# ===================================================================
if __name__ == "__main__":
# Get defaults so it can work with no Sacred
parser = argparse.ArgumentParser()
parser.add_argument('-f', "--filelist_path", required=True)
parser.add_argument('-c', '--config', type=str,
help='JSON file for configuration')
parser.add_argument('-o', '--output_dir', type=str,
help='Output directory')
args = parser.parse_args()
with open(args.config) as f:
data = f.read()
data_config = json.loads(data)["data_config"]
mel2samp = Mel2Samp(**data_config)
filepaths = files_to_list(args.filelist_path)
# Make directory if it doesn't exist
if not os.path.isdir(args.output_dir):
os.makedirs(args.output_dir)
os.chmod(args.output_dir, 0o775)
for filepath in filepaths:
audio, sr = load_wav_to_torch(filepath)
melspectrogram = mel2samp.get_mel(audio)
filename = os.path.basename(filepath)
new_filepath = args.output_dir + '/' + filename + '.pt'
print(new_filepath)
torch.save(melspectrogram, new_filepath)