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
|
# EXTRACT ACOUSTIC FEATURES USING WORLD VOCODER
PYTHONPATH=`pwd` python scripts/extract_acoustic_feature.py \
--input1_directory './data/mat-holly-24000/wav/holly' \
--input2_directory './data/mat-holly-24000/wav/mat' \
--output1_directory './data/mat-holly-24000/feat/holly' \
--output2_directory './data/mat-holly-24000/feat/mat'
parser.add_argument('--input1_directory', '-i1', type=Path)
parser.add_argument('--input2_directory', '-i2', type=Path)
parser.add_argument('--output1_directory', '-o1', type=Path)
parser.add_argument('--output2_directory', '-o2', type=Path)
parser.add_argument('--pre_converter1_config', type=Path)
parser.add_argument('--pre_converter1_model', type=Path)
parser.add_argument('--sample_rate', type=int, default=base_voice_param.sample_rate)
parser.add_argument('--top_db', type=float, default=base_voice_param.top_db)
parser.add_argument('--pad_second', type=float, default=base_voice_param.pad_second)
parser.add_argument('--frame_period', type=int, default=base_acoustic_feature_param.frame_period)
parser.add_argument('--order', type=int, default=base_acoustic_feature_param.order)
parser.add_argument('--alpha', type=float, default=base_acoustic_feature_param.alpha)
parser.add_argument('--f0_estimating_method', type=str, default=base_acoustic_feature_param.f0_estimating_method)
parser.add_argument('--f0_floor1', type=float, default=71)
parser.add_argument('--f0_ceil1', type=float, default=800)
parser.add_argument('--f0_floor2', type=float, default=71)
parser.add_argument('--f0_ceil2', type=float, default=800)
parser.add_argument('--ignore_feature', nargs='+', default=['spectrogram', 'aperiodicity'])
parser.add_argument('--disable_alignment', action='store_true')
parser.add_argument('--enable_overwrite', action='store_true')
# TRAIN VOICE CONVERSION NETWORKS
/home/spawn/.virtualenv/yukarin/bin/python train.py \
./20180719133105.3.mat_2_holly.json \
./data/mat-holly-24000/conversion-net/pp-el0-holly-2-mat
# OPTIONAL: TRAIN ON MULTIPLE GPUS USING RECIPE.JSON (RUNS MULTIPLE SCREENS)
PYTHONPATH=`pwd` /home/spawn/.virtualenv/yukarin/bin/python scripts/launch.py data/mat-holly-24000/conversion-net/
# TEST VOICE CONVERSION NETWORKS
PYTHONPATH=`pwd` /home/spawn/.virtualenv/yukarin/bin/python scripts/voice_conversion_test.py \
--model_directory /home/spawn/code/become-yukarin/data/mat-holly-24000/conversion-net/ \
--input_wave_directory /home/spawn/code/become-yukarin/data/mat-holly-24000/wav/mat/ \
--gpu 0 \
pp-el8-mat-2-holly
PYTHONPATH=`pwd` /home/spawn/.virtualenv/yukarin/bin/python scripts/voice_conversion_test.py \
--model_directory /home/spawn/code/become-yukarin/data/mat-holly-24000/conversion-net/ \
--input_wave_directory /home/spawn/code/become-yukarin/data/mat-holly-24000/wav/holly/ \
--gpu 0 \
pp-el8-holly-2-mat
parser.add_argument('model_names', nargs='+')
parser.add_argument('-md', '--model_directory', type=Path, default=Path('/mnt/dwango/hiroshiba/become-yukarin/'))
parser.add_argument('-iwd', '--input_wave_directory', type=Path,
default=Path('/mnt/dwango/hiroshiba/become-yukarin/dataset/hiho-wave/hiho-pause-atr503-subset/'))
parser.add_argument('-it', '--iteration', type=int)
parser.add_argument('-g', '--gpu', type=int)
# EXTRACT SPECTROGRAM PAIRS
./split.sh holly_normal.wav 4 holly_normal_4
cd holly_normal_4
mkdir 24k
for i in *.wav
do
sox $i -r 24000 "24k/$i"
done
PYTHONPATH=`pwd` python scripts/extract_spectrogram_pair.py \
--input_directory './data/holly-24000/wav' \
--output_directory './data/holly-24000/spec'
parser.add_argument('--input_directory', '-i', type=Path)
parser.add_argument('--output_directory', '-o', type=Path)
parser.add_argument('--sample_rate', type=int, default=base_voice_param.sample_rate)
parser.add_argument('--top_db', type=float, default=base_voice_param.top_db)
parser.add_argument('--pad_second', type=float, default=base_voice_param.pad_second)
parser.add_argument('--frame_period', type=int, default=base_acoustic_feature_param.frame_period)
parser.add_argument('--order', type=int, default=base_acoustic_feature_param.order)
parser.add_argument('--alpha', type=float, default=base_acoustic_feature_param.alpha)
parser.add_argument('--f0_estimating_method', default=base_acoustic_feature_param.f0_estimating_method)
parser.add_argument('--enable_overwrite', action='store_true')
# TRAIN SUPER RESOLUTION NETWORK
/home/spawn/.virtualenv/yukarin/bin/python train_sr.py \
./holly_sr.json \
./data/holly-24000/net
# TEST SUPER RESOLUTION NETWORK
/home/spawn/.virtualenv/yukarin/bin/python ./scripts/super_resolution_test.py \
--model_directory /home/spawn/code/become-yukarin/data/holly-24000/ \
--input_wave_directory /home/spawn/code/become-yukarin/data/holly-24000/wav/ \
--gpu 0 \
net
parser.add_argument('model_names', nargs='+')
parser.add_argument('-md', '--model_directory', type=Path, default=Path('/mnt/dwango/hiroshiba/become-yukarin/'))
parser.add_argument('-iwd', '--input_wave_directory', type=Path,
default=Path('/mnt/dwango/hiroshiba/become-yukarin/dataset/yukari-wave/yukari-news/'))
parser.add_argument('-g', '--gpu', type=int)
# VOICE TRANSFORM
PYTHONPATH=`pwd` /home/spawn/.virtualenv/yukarin/bin/python scripts/process.py \
--conv_model_directory ./data/mat-holly-24000/conversion-net/pp-el8-mat-2-holly/ \
--sr_model_directory ./data/holly-24000/net/ \
--input_directory ./test_data \
--output_directory ./output \
--gpu 0
mkdir all
for i in *.wav
do
dir=${i%.*}
../br_misc/split.sh $i 12 $dir
cd $dir
for j in *.wav
do
mv "$j" "${dir}_${j}"
done
cd ..
mv $i/* all
done
cd output/net/
mkdir concat
|
