Hugging Face's logo

dkounadis
/
artificial-styletts2

Text-to-Audio
Audiocraft
English
audiogen
styletts2
shift
audeering
sound
audio-generation
text-to-speech
mimic3
Model card Files Community
4
Dionyssos commited on
Commit
8a2aca3
•
1 Parent(s): 6430cbc

draft various spk

Browse files
Files changed (4) hide show
  1. mimic3_make_harvard_sentences.py +8 -3
  2. tts_harvard.py +82 -44
  3. visualize_tts_plesantness.py +7 -7
  4. voices.json +1612 -0
mimic3_make_harvard_sentences.py CHANGED
@@ -75,6 +75,7 @@ list_voices = [
75
  ] # special - for human we load specific style file - no Mimic3 is run
76
 
77
 
 
78
  # ================================================== INTERFACE MODELS
79
  LABELS = [
80
  'arousal', 'dominance', 'valence',
@@ -296,7 +297,7 @@ for _id, _voice in enumerate(list_voices):
296
  total_audio_mimic3 = []
297
  total_audio_styletts2 = []
298
  ix = 0
299
- for list_of_10 in harvard_individual_sentences[:4]: # 77
300
 
301
  text = ' '.join(list_of_10['sentences'])
302
 
@@ -367,8 +368,12 @@ for _id, _voice in enumerate(list_voices):
367
 
368
  # MIMIC3 = = = = = = = = = = = = = = END
369
 
370
-
371
-
 
 
 
 
372
 
373
  style_vec = msinference.compute_style(style_path) # use mimic-3 as prompt
374
 
 
75
  ] # special - for human we load specific style file - no Mimic3 is run
76
 
77
 
78
+
79
  # ================================================== INTERFACE MODELS
80
  LABELS = [
81
  'arousal', 'dominance', 'valence',
 
297
  total_audio_mimic3 = []
298
  total_audio_styletts2 = []
299
  ix = 0
300
+ for list_of_10 in harvard_individual_sentences[:1000]: # 77
301
 
302
  text = ' '.join(list_of_10['sentences'])
303
 
 
368
 
369
  # MIMIC3 = = = = = = = = = = = = = = END
370
 
371
+ if 'en_US' in _str:
372
+ style_path = 'mimic3_english_4x/' + _str + '.wav'
373
+ elif ('de_DE' in _str) or ('fr_FR' in _str):
374
+ style_path = 'mimic3_foreign_4x/' + _str + '.wav'
375
+ else:
376
+ print(f'use human / generated style for {_str}')
377
 
378
  style_vec = msinference.compute_style(style_path) # use mimic-3 as prompt
379
 
tts_harvard.py CHANGED
@@ -1,42 +1,49 @@
1
- # Synthesize all Harvard Lists - 767 sentences as single .wav
2
  #
3
- # 1.
 
 
 
 
 
4
  #
5
- # './prompt_mimic3_english/'
6
- #
7
- # 2.
8
- #
9
- # './prompt_mimic3_english_4x/'
10
- #
11
- # 3.
12
- #
13
- # './prompt_human/'
14
- #
15
- # 4.
16
- #
17
- # './prompt_mimic3_foreign/'
18
- #
19
- # 5.
20
- #
21
- # './prompt_mimic3_foreign_4x/
22
- #
23
- #
24
- # ----> THE .wavs will be used for visualisation
25
 
26
  import soundfile
27
  import json
28
  import numpy as np
29
  import audb
30
  from pathlib import Path
31
- import os
32
  LABELS = ['arousal', 'dominance', 'valence']
33
 
34
 
35
- def load_human_speech(split=None):
 
 
36
  DB = [
37
  # [dataset, version, table, has_timdeltas_or_is_full_wavfile]
38
  # ['crema-d', '1.1.1', 'emotion.voice.test', False],
39
- ['emodb', '1.2.0', 'emotion.categories.train.gold_standard', False],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
40
  ]
41
 
42
  output_list = []
@@ -58,38 +65,65 @@ def load_human_speech(split=None):
58
  output_list += [f for f in a.index] # use file (no timedeltas)
59
  return output_list
60
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61
 
 
 
 
 
 
 
 
62
 
 
 
 
 
63
 
64
 
65
  # SYNTHESIZE mimic mimicx4 crema-d
66
  import msinference
67
- prompt_paths = {}
68
 
69
  with open('harvard.json', 'r') as f:
70
  harvard_individual_sentences = json.load(f)['sentences']
71
 
72
- for audio_prompt in [#'mimic3_english',
73
- #'mimic3_english_4x',
74
- 'human',
75
- 'mimic3_foreign',
76
- 'mimic3_foreign_4x']:
77
-
78
- if audio_prompt == 'human':
79
- prompt_paths = load_human_speech() # better emodb ?
80
- else:
81
- prompt_dir = '/data/dkounadis/artificial-styletts2/' + audio_prompt + '/'
82
- prompt_paths = [prompt_dir + f for f in os.listdir(prompt_dir)]
83
- prompt_paths = prompt_paths[:10]
84
- print(prompt_paths,'\n\n__________')
85
 
 
 
 
 
86
  total_audio = []
87
  ix = 0
88
- for list_of_10 in harvard_individual_sentences[:1]:
89
  # long_sentence = ' '.join(list_of_10['sentences'])
90
  # harvard.append(long_sentence.replace('.', ' '))
91
  for text in list_of_10['sentences']:
92
- style_vec = msinference.compute_style(prompt_paths[ix % len(prompt_paths)])
 
 
 
 
 
 
 
 
 
 
93
  print(ix, text)
94
  ix += 1
95
  x = msinference.inference(text,
@@ -100,6 +134,10 @@ for audio_prompt in [#'mimic3_english',
100
  embedding_scale=1)
101
 
102
  total_audio.append(x)
103
- total_audio = np.concatenate(total_audio) # -- concat 77x lists
104
- soundfile.write(f'{audio_prompt}_767_5.wav', total_audio, 24000)
105
- print(f'{audio_prompt}_767_5.wav')
 
 
 
 
 
1
+ # Synthesize all Harvard Lists 77x lists of 10x sentences to single .wav
2
  #
3
+ # 1. using mimic3 style
4
+ # Folder: 'prompt_mimic3/'
5
+ # 2. using mimic3 4x accelerated style
6
+ # Folder: 'prompt_mimic3speed/'
7
+ # 3. using crema-d style
8
+ # Folder: 'prompt_human/'
9
  #
10
+ # WAVS used from tts_paper_plot.py
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
 
12
  import soundfile
13
  import json
14
  import numpy as np
15
  import audb
16
  from pathlib import Path
17
+
18
  LABELS = ['arousal', 'dominance', 'valence']
19
 
20
 
21
+
22
+
23
+ def load_speech(split=None):
24
  DB = [
25
  # [dataset, version, table, has_timdeltas_or_is_full_wavfile]
26
  # ['crema-d', '1.1.1', 'emotion.voice.test', False],
27
+ ['librispeech', '3.1.0', 'train-clean-360', False],
28
+ # ['emodb', '1.2.0', 'emotion.categories.train.gold_standard', False],
29
+ # ['entertain-playtestcloud', '1.1.0', 'emotion.categories.train.gold_standard', True],
30
+ # ['erik', '2.2.0', 'emotion.categories.train.gold_standard', True],
31
+ # ['meld', '1.3.1', 'emotion.categories.train.gold_standard', False],
32
+ # ['msppodcast', '5.0.0', 'emotion.categories.train.gold_standard', False], # tandalone bucket because it has gt labels?
33
+ # ['myai', '1.0.1', 'emotion.categories.train.gold_standard', False],
34
+ # ['casia', None, 'emotion.categories.gold_standard', False],
35
+ # ['switchboard-1', None, 'sentiment', True],
36
+ # ['swiss-parliament', None, 'segments', True],
37
+ # ['argentinian-parliament', None, 'segments', True],
38
+ # ['austrian-parliament', None, 'segments', True],
39
+ # #'german', --> bundestag
40
+ # ['brazilian-parliament', None, 'segments', True],
41
+ # ['mexican-parliament', None, 'segments', True],
42
+ # ['portuguese-parliament', None, 'segments', True],
43
+ # ['spanish-parliament', None, 'segments', True],
44
+ # ['chinese-vocal-emotions-liu-pell', None, 'emotion.categories.desired', False],
45
+ # peoples-speech slow
46
+ # ['peoples-speech', None, 'train-initial', False]
47
  ]
48
 
49
  output_list = []
 
65
  output_list += [f for f in a.index] # use file (no timedeltas)
66
  return output_list
67
 
68
+
69
+
70
+
71
+
72
+ # Generate 77 wavs
73
+
74
+
75
+
76
+
77
+ with open('voices.json', 'r') as f:
78
+ df = json.load(f)['voices']
79
+ voice_names = [v['voice'] for k,v in df.items()]
80
+ synthetic_wav_paths = []
81
+ synthetic_wav_paths_AFFECT = []
82
+ for voice in voice_names:
83
 
84
+ synthetic_wav_paths.append(
85
+ '/data/dkounadis/shift/assets/wavs/style_vector/' + voice.replace('/', '_').replace('#', '_').replace(
86
+ 'cmu-arctic', 'cmu_arctic').replace('_low', '') + '.wav')
87
+ synthetic_wav_paths_AFFECT.append(
88
+ '/data/dkounadis/shift/assets/wavs/style_vector_v2/' + voice.replace('/', '_').replace('#', '_').replace(
89
+ 'cmu-arctic', 'cmu_arctic').replace('_low', '') + '.wav')
90
+
91
 
92
+ print(len(synthetic_wav_paths))
93
+
94
+
95
+ natural_wav_paths = load_speech()
96
 
97
 
98
  # SYNTHESIZE mimic mimicx4 crema-d
99
  import msinference
100
+
101
 
102
  with open('harvard.json', 'r') as f:
103
  harvard_individual_sentences = json.load(f)['sentences']
104
 
 
 
 
 
 
 
 
 
 
 
 
 
 
105
 
106
+
107
+
108
+
109
+ for audio_prompt in ['mimic3', 'mimic3_speed', 'human']:
110
  total_audio = []
111
  ix = 0
112
+ for list_of_10 in harvard_individual_sentences:
113
  # long_sentence = ' '.join(list_of_10['sentences'])
114
  # harvard.append(long_sentence.replace('.', ' '))
115
  for text in list_of_10['sentences']:
116
+ if audio_prompt == 'mimic3':
117
+ style_vec = msinference.compute_style(
118
+ synthetic_wav_paths[ix % 134])
119
+ elif audio_prompt == 'mimic3_speed':
120
+ style_vec = msinference.compute_style(
121
+ synthetic_wav_paths_AFFECT[ix % 134])
122
+ elif audio_prompt == 'human':
123
+ style_vec = msinference.compute_style(
124
+ natural_wav_paths[ix % len(natural_wav_paths)])
125
+ else:
126
+ print('unknonw list of style vecto')
127
  print(ix, text)
128
  ix += 1
129
  x = msinference.inference(text,
 
134
  embedding_scale=1)
135
 
136
  total_audio.append(x)
137
+ # concat before write
138
+ # -- for 10x sentenctes
139
+ print('_____________________')
140
+ # -- for 77x lists
141
+ total_audio = np.concatenate(total_audio)
142
+ soundfile.write(f'{audio_prompt}_770.wav', total_audio, 24000)
143
+ print(f'{audio_prompt}_full_770.wav')
visualize_tts_plesantness.py CHANGED
@@ -81,12 +81,12 @@ def _sigmoid(x):
81
 
82
  # for mimic3/mimic3speed/human - concat all 77 and run timeseries with 7s hop 3s
83
  for long_audio in [
84
- 'mimic3_english_767_5.wav',
85
- 'mimic3_english_4x_767_5.wav',
86
- 'human_767_5.wav',
87
- 'mimic3_foregin_767_5.wav',
88
- 'mimic3_foreign_4x_767_5.wav'
89
- ]:
90
  file_interface = f'timeseries_{long_audio.replace("/", "")}.pkl'
91
  if not os.path.exists(file_interface):
92
 
@@ -106,7 +106,7 @@ for long_audio in [
106
  # pool
107
  h = self.pool_model.sap_linear(x).tanh()
108
  w = torch.matmul(h, self.pool_model.attention)
109
- w = stylesoftmax(1)
110
  mu = (x * w).sum(1)
111
  x = torch.cat(
112
  [
 
81
 
82
  # for mimic3/mimic3speed/human - concat all 77 and run timeseries with 7s hop 3s
83
  for long_audio in [
84
+ # 'mimic3.wav',
85
+ # 'mimic3_speedup.wav',
86
+ 'human_770.wav', # 'mimic3_all_77.wav', #
87
+ 'mimic3_770.wav',
88
+ 'mimic3_speed_770.wav'
89
+ ]:
90
  file_interface = f'timeseries_{long_audio.replace("/", "")}.pkl'
91
  if not os.path.exists(file_interface):
92
 
 
106
  # pool
107
  h = self.pool_model.sap_linear(x).tanh()
108
  w = torch.matmul(h, self.pool_model.attention)
109
+ w = w.softmax(1)
110
  mu = (x * w).sum(1)
111
  x = torch.cat(
112
  [
voices.json ADDED
@@ -0,0 +1,1612 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "voices": {
3
+ "0": {
4
+ "voice": "en_US/vctk_low#p239",
5
+ "emotion": [
6
+ 0.001872760709375143,
7
+ 0.18340148776769638,
8
+ 0.3238210380077362
9
+ ],
10
+ "tgt_wav": "assets/wavs/en_US_vctk_p239.wav",
11
+ "affect_wav": "assets/wavs/en_US_vctk_p239_affect.wav",
12
+ "fig_file": "assets/en_US_vctk_p239.png",
13
+ "str_voice": "en_US_vctk_p239"
14
+ },
15
+ "1": {
16
+ "voice": "en_US/vctk_low#p236",
17
+ "emotion": [
18
+ 0.017000655410811305,
19
+ 0.20464985817670822,
20
+ 0.3406614363193512
21
+ ],
22
+ "tgt_wav": "assets/wavs/en_US_vctk_p236.wav",
23
+ "affect_wav": "assets/wavs/en_US_vctk_p236_affect.wav",
24
+ "fig_file": "assets/en_US_vctk_p236.png",
25
+ "str_voice": "en_US_vctk_p236"
26
+ },
27
+ "2": {
28
+ "voice": "en_US/vctk_low#p264",
29
+ "emotion": [
30
+ 0.0,
31
+ 0.14136632531881332,
32
+ 0.3482646197080612
33
+ ],
34
+ "tgt_wav": "assets/wavs/en_US_vctk_p264.wav",
35
+ "affect_wav": "assets/wavs/en_US_vctk_p264_affect.wav",
36
+ "fig_file": "assets/en_US_vctk_p264.png",
37
+ "str_voice": "en_US_vctk_p264"
38
+ },
39
+ "3": {
40
+ "voice": "en_US/vctk_low#p250",
41
+ "emotion": [
42
+ 0.0,
43
+ 0.1662377566099167,
44
+ 0.31436721980571747
45
+ ],
46
+ "tgt_wav": "assets/wavs/en_US_vctk_p250.wav",
47
+ "affect_wav": "assets/wavs/en_US_vctk_p250_affect.wav",
48
+ "fig_file": "assets/en_US_vctk_p250.png",
49
+ "str_voice": "en_US_vctk_p250"
50
+ },
51
+ "4": {
52
+ "voice": "en_US/vctk_low#p259",
53
+ "emotion": [
54
+ 0.1525903344154358,
55
+ 0.32237668335437775,
56
+ 0.3625539541244507
57
+ ],
58
+ "tgt_wav": "assets/wavs/en_US_vctk_p259.wav",
59
+ "affect_wav": "assets/wavs/en_US_vctk_p259_affect.wav",
60
+ "fig_file": "assets/en_US_vctk_p259.png",
61
+ "str_voice": "en_US_vctk_p259"
62
+ },
63
+ "5": {
64
+ "voice": "en_US/vctk_low#p247",
65
+ "emotion": [
66
+ 0.18136803060770035,
67
+ 0.3469696491956711,
68
+ 0.3744945675134659
69
+ ],
70
+ "tgt_wav": "assets/wavs/en_US_vctk_p247.wav",
71
+ "affect_wav": "assets/wavs/en_US_vctk_p247_affect.wav",
72
+ "fig_file": "assets/en_US_vctk_p247.png",
73
+ "str_voice": "en_US_vctk_p247"
74
+ },
75
+ "6": {
76
+ "voice": "en_US/vctk_low#p261",
77
+ "emotion": [
78
+ 0.018009521067142487,
79
+ 0.20371025055646896,
80
+ 0.3044930547475815
81
+ ],
82
+ "tgt_wav": "assets/wavs/en_US_vctk_p261.wav",
83
+ "affect_wav": "assets/wavs/en_US_vctk_p261_affect.wav",
84
+ "fig_file": "assets/en_US_vctk_p261.png",
85
+ "str_voice": "en_US_vctk_p261"
86
+ },
87
+ "7": {
88
+ "voice": "en_US/vctk_low#p263",
89
+ "emotion": [
90
+ 0.19148042052984238,
91
+ 0.3607906401157379,
92
+ 0.4013771116733551
93
+ ],
94
+ "tgt_wav": "assets/wavs/en_US_vctk_p263.wav",
95
+ "affect_wav": "assets/wavs/en_US_vctk_p263_affect.wav",
96
+ "fig_file": "assets/en_US_vctk_p263.png",
97
+ "str_voice": "en_US_vctk_p263"
98
+ },
99
+ "8": {
100
+ "voice": "en_US/vctk_low#p283",
101
+ "emotion": [
102
+ 0.0,
103
+ 0.15594127774238586,
104
+ 0.29933948814868927
105
+ ],
106
+ "tgt_wav": "assets/wavs/en_US_vctk_p283.wav",
107
+ "affect_wav": "assets/wavs/en_US_vctk_p283_affect.wav",
108
+ "fig_file": "assets/en_US_vctk_p283.png",
109
+ "str_voice": "en_US_vctk_p283"
110
+ },
111
+ "9": {
112
+ "voice": "en_US/vctk_low#p274",
113
+ "emotion": [
114
+ 0.14588971808552742,
115
+ 0.31555116176605225,
116
+ 0.34421128034591675
117
+ ],
118
+ "tgt_wav": "assets/wavs/en_US_vctk_p274.wav",
119
+ "affect_wav": "assets/wavs/en_US_vctk_p274_affect.wav",
120
+ "fig_file": "assets/en_US_vctk_p274.png",
121
+ "str_voice": "en_US_vctk_p274"
122
+ },
123
+ "10": {
124
+ "voice": "en_US/vctk_low#p286",
125
+ "emotion": [
126
+ 0.07937583327293396,
127
+ 0.2856627404689789,
128
+ 0.26015742123126984
129
+ ],
130
+ "tgt_wav": "assets/wavs/en_US_vctk_p286.wav",
131
+ "affect_wav": "assets/wavs/en_US_vctk_p286_affect.wav",
132
+ "fig_file": "assets/en_US_vctk_p286.png",
133
+ "str_voice": "en_US_vctk_p286"
134
+ },
135
+ "11": {
136
+ "voice": "en_US/vctk_low#p276",
137
+ "emotion": [
138
+ 0.0,
139
+ 0.15758457779884338,
140
+ 0.32494914531707764
141
+ ],
142
+ "tgt_wav": "assets/wavs/en_US_vctk_p276.wav",
143
+ "affect_wav": "assets/wavs/en_US_vctk_p276_affect.wav",
144
+ "fig_file": "assets/en_US_vctk_p276.png",
145
+ "str_voice": "en_US_vctk_p276"
146
+ },
147
+ "12": {
148
+ "voice": "en_US/vctk_low#p270",
149
+ "emotion": [
150
+ 0.062022265046834946,
151
+ 0.25552842766046524,
152
+ 0.29453884065151215
153
+ ],
154
+ "tgt_wav": "assets/wavs/en_US_vctk_p270.wav",
155
+ "affect_wav": "assets/wavs/en_US_vctk_p270_affect.wav",
156
+ "fig_file": "assets/en_US_vctk_p270.png",
157
+ "str_voice": "en_US_vctk_p270"
158
+ },
159
+ "13": {
160
+ "voice": "en_US/vctk_low#p281",
161
+ "emotion": [
162
+ 0.11574980244040489,
163
+ 0.2708873227238655,
164
+ 0.3113572597503662
165
+ ],
166
+ "tgt_wav": "assets/wavs/en_US_vctk_p281.wav",
167
+ "affect_wav": "assets/wavs/en_US_vctk_p281_affect.wav",
168
+ "fig_file": "assets/en_US_vctk_p281.png",
169
+ "str_voice": "en_US_vctk_p281"
170
+ },
171
+ "14": {
172
+ "voice": "en_US/vctk_low#p277",
173
+ "emotion": [
174
+ 0.0,
175
+ 0.14411474019289017,
176
+ 0.29423509538173676
177
+ ],
178
+ "tgt_wav": "assets/wavs/en_US_vctk_p277.wav",
179
+ "affect_wav": "assets/wavs/en_US_vctk_p277_affect.wav",
180
+ "fig_file": "assets/en_US_vctk_p277.png",
181
+ "str_voice": "en_US_vctk_p277"
182
+ },
183
+ "15": {
184
+ "voice": "en_US/vctk_low#p231",
185
+ "emotion": [
186
+ 0.006912603043019772,
187
+ 0.18807627260684967,
188
+ 0.2868843674659729
189
+ ],
190
+ "tgt_wav": "assets/wavs/en_US_vctk_p231.wav",
191
+ "affect_wav": "assets/wavs/en_US_vctk_p231_affect.wav",
192
+ "fig_file": "assets/en_US_vctk_p231.png",
193
+ "str_voice": "en_US_vctk_p231"
194
+ },
195
+ "16": {
196
+ "voice": "en_US/vctk_low#p238",
197
+ "emotion": [
198
+ 0.00611584447324276,
199
+ 0.1861497163772583,
200
+ 0.36705054342746735
201
+ ],
202
+ "tgt_wav": "assets/wavs/en_US_vctk_p238.wav",
203
+ "affect_wav": "assets/wavs/en_US_vctk_p238_affect.wav",
204
+ "fig_file": "assets/en_US_vctk_p238.png",
205
+ "str_voice": "en_US_vctk_p238"
206
+ },
207
+ "17": {
208
+ "voice": "en_US/vctk_low#p271",
209
+ "emotion": [
210
+ 0.14705300331115723,
211
+ 0.31942586600780487,
212
+ 0.3615124672651291
213
+ ],
214
+ "tgt_wav": "assets/wavs/en_US_vctk_p271.wav",
215
+ "affect_wav": "assets/wavs/en_US_vctk_p271_affect.wav",
216
+ "fig_file": "assets/en_US_vctk_p271.png",
217
+ "str_voice": "en_US_vctk_p271"
218
+ },
219
+ "18": {
220
+ "voice": "en_US/vctk_low#p257",
221
+ "emotion": [
222
+ 0.0,
223
+ 0.14702527225017548,
224
+ 0.3079967051744461
225
+ ],
226
+ "tgt_wav": "assets/wavs/en_US_vctk_p257.wav",
227
+ "affect_wav": "assets/wavs/en_US_vctk_p257_affect.wav",
228
+ "fig_file": "assets/en_US_vctk_p257.png",
229
+ "str_voice": "en_US_vctk_p257"
230
+ },
231
+ "19": {
232
+ "voice": "en_US/vctk_low#p273",
233
+ "emotion": [
234
+ 0.019663178827613592,
235
+ 0.20706159621477127,
236
+ 0.31418660283088684
237
+ ],
238
+ "tgt_wav": "assets/wavs/en_US_vctk_p273.wav",
239
+ "affect_wav": "assets/wavs/en_US_vctk_p273_affect.wav",
240
+ "fig_file": "assets/en_US_vctk_p273.png",
241
+ "str_voice": "en_US_vctk_p273"
242
+ },
243
+ "20": {
244
+ "voice": "en_US/vctk_low#p284",
245
+ "emotion": [
246
+ 0.24391663074493408,
247
+ 0.37824292480945587,
248
+ 0.38162167370319366
249
+ ],
250
+ "tgt_wav": "assets/wavs/en_US_vctk_p284.wav",
251
+ "affect_wav": "assets/wavs/en_US_vctk_p284_affect.wav",
252
+ "fig_file": "assets/en_US_vctk_p284.png",
253
+ "str_voice": "en_US_vctk_p284"
254
+ },
255
+ "21": {
256
+ "voice": "en_US/vctk_low#p329",
257
+ "emotion": [
258
+ 0.0,
259
+ 0.13082702457904816,
260
+ 0.3300574719905853
261
+ ],
262
+ "tgt_wav": "assets/wavs/en_US_vctk_p329.wav",
263
+ "affect_wav": "assets/wavs/en_US_vctk_p329_affect.wav",
264
+ "fig_file": "assets/en_US_vctk_p329.png",
265
+ "str_voice": "en_US_vctk_p329"
266
+ },
267
+ "22": {
268
+ "voice": "en_US/vctk_low#p361",
269
+ "emotion": [
270
+ 0.03181833028793335,
271
+ 0.2181432768702507,
272
+ 0.3542213588953018
273
+ ],
274
+ "tgt_wav": "assets/wavs/en_US_vctk_p361.wav",
275
+ "affect_wav": "assets/wavs/en_US_vctk_p361_affect.wav",
276
+ "fig_file": "assets/en_US_vctk_p361.png",
277
+ "str_voice": "en_US_vctk_p361"
278
+ },
279
+ "23": {
280
+ "voice": "en_US/vctk_low#p287",
281
+ "emotion": [
282
+ 0.23462356626987457,
283
+ 0.3727245330810547,
284
+ 0.3918970078229904
285
+ ],
286
+ "tgt_wav": "assets/wavs/en_US_vctk_p287.wav",
287
+ "affect_wav": "assets/wavs/en_US_vctk_p287_affect.wav",
288
+ "fig_file": "assets/en_US_vctk_p287.png",
289
+ "str_voice": "en_US_vctk_p287"
290
+ },
291
+ "24": {
292
+ "voice": "en_US/vctk_low#p360",
293
+ "emotion": [
294
+ 0.18893346190452576,
295
+ 0.33515486121177673,
296
+ 0.35971683263778687
297
+ ],
298
+ "tgt_wav": "assets/wavs/en_US_vctk_p360.wav",
299
+ "affect_wav": "assets/wavs/en_US_vctk_p360_affect.wav",
300
+ "fig_file": "assets/en_US_vctk_p360.png",
301
+ "str_voice": "en_US_vctk_p360"
302
+ },
303
+ "25": {
304
+ "voice": "en_US/vctk_low#p374",
305
+ "emotion": [
306
+ 0.13329850882291794,
307
+ 0.30131587386131287,
308
+ 0.3387715220451355
309
+ ],
310
+ "tgt_wav": "assets/wavs/en_US_vctk_p374.wav",
311
+ "affect_wav": "assets/wavs/en_US_vctk_p374_affect.wav",
312
+ "fig_file": "assets/en_US_vctk_p374.png",
313
+ "str_voice": "en_US_vctk_p374"
314
+ },
315
+ "26": {
316
+ "voice": "en_US/vctk_low#p376",
317
+ "emotion": [
318
+ 0.04164157807826996,
319
+ 0.22950325906276703,
320
+ 0.2736595422029495
321
+ ],
322
+ "tgt_wav": "assets/wavs/en_US_vctk_p376.wav",
323
+ "affect_wav": "assets/wavs/en_US_vctk_p376_affect.wav",
324
+ "fig_file": "assets/en_US_vctk_p376.png",
325
+ "str_voice": "en_US_vctk_p376"
326
+ },
327
+ "27": {
328
+ "voice": "en_US/vctk_low#p310",
329
+ "emotion": [
330
+ 0.022892486304044724,
331
+ 0.20763949304819107,
332
+ 0.3794092684984207
333
+ ],
334
+ "tgt_wav": "assets/wavs/en_US_vctk_p310.wav",
335
+ "affect_wav": "assets/wavs/en_US_vctk_p310_affect.wav",
336
+ "fig_file": "assets/en_US_vctk_p310.png",
337
+ "str_voice": "en_US_vctk_p310"
338
+ },
339
+ "28": {
340
+ "voice": "en_US/vctk_low#p304",
341
+ "emotion": [
342
+ 0.14982996135950089,
343
+ 0.2977369725704193,
344
+ 0.2675946205854416
345
+ ],
346
+ "tgt_wav": "assets/wavs/en_US_vctk_p304.wav",
347
+ "affect_wav": "assets/wavs/en_US_vctk_p304_affect.wav",
348
+ "fig_file": "assets/en_US_vctk_p304.png",
349
+ "str_voice": "en_US_vctk_p304"
350
+ },
351
+ "29": {
352
+ "voice": "en_US/vctk_low#p340",
353
+ "emotion": [
354
+ 0.0,
355
+ 0.15842513740062714,
356
+ 0.3057490587234497
357
+ ],
358
+ "tgt_wav": "assets/wavs/en_US_vctk_p340.wav",
359
+ "affect_wav": "assets/wavs/en_US_vctk_p340_affect.wav",
360
+ "fig_file": "assets/en_US_vctk_p340.png",
361
+ "str_voice": "en_US_vctk_p340"
362
+ },
363
+ "30": {
364
+ "voice": "en_US/vctk_low#p347",
365
+ "emotion": [
366
+ 0.1295829713344574,
367
+ 0.28821905702352524,
368
+ 0.32179172337055206
369
+ ],
370
+ "tgt_wav": "assets/wavs/en_US_vctk_p347.wav",
371
+ "affect_wav": "assets/wavs/en_US_vctk_p347_affect.wav",
372
+ "fig_file": "assets/en_US_vctk_p347.png",
373
+ "str_voice": "en_US_vctk_p347"
374
+ },
375
+ "31": {
376
+ "voice": "en_US/vctk_low#p330",
377
+ "emotion": [
378
+ 0.0,
379
+ 0.15657036751508713,
380
+ 0.29501572251319885
381
+ ],
382
+ "tgt_wav": "assets/wavs/en_US_vctk_p330.wav",
383
+ "affect_wav": "assets/wavs/en_US_vctk_p330_affect.wav",
384
+ "fig_file": "assets/en_US_vctk_p330.png",
385
+ "str_voice": "en_US_vctk_p330"
386
+ },
387
+ "32": {
388
+ "voice": "en_US/vctk_low#p308",
389
+ "emotion": [
390
+ 0.01279856450855732,
391
+ 0.18664204329252243,
392
+ 0.34837856888771057
393
+ ],
394
+ "tgt_wav": "assets/wavs/en_US_vctk_p308.wav",
395
+ "affect_wav": "assets/wavs/en_US_vctk_p308_affect.wav",
396
+ "fig_file": "assets/en_US_vctk_p308.png",
397
+ "str_voice": "en_US_vctk_p308"
398
+ },
399
+ "33": {
400
+ "voice": "en_US/vctk_low#p314",
401
+ "emotion": [
402
+ 0.0,
403
+ 0.14651427417993546,
404
+ 0.2921117842197418
405
+ ],
406
+ "tgt_wav": "assets/wavs/en_US_vctk_p314.wav",
407
+ "affect_wav": "assets/wavs/en_US_vctk_p314_affect.wav",
408
+ "fig_file": "assets/en_US_vctk_p314.png",
409
+ "str_voice": "en_US_vctk_p314"
410
+ },
411
+ "34": {
412
+ "voice": "en_US/vctk_low#p317",
413
+ "emotion": [
414
+ 0.0,
415
+ 0.14166613668203354,
416
+ 0.3167337030172348
417
+ ],
418
+ "tgt_wav": "assets/wavs/en_US_vctk_p317.wav",
419
+ "affect_wav": "assets/wavs/en_US_vctk_p317_affect.wav",
420
+ "fig_file": "assets/en_US_vctk_p317.png",
421
+ "str_voice": "en_US_vctk_p317"
422
+ },
423
+ "35": {
424
+ "voice": "en_US/vctk_low#p339",
425
+ "emotion": [
426
+ 0.0,
427
+ 0.14359427988529205,
428
+ 0.2953350394964218
429
+ ],
430
+ "tgt_wav": "assets/wavs/en_US_vctk_p339.wav",
431
+ "affect_wav": "assets/wavs/en_US_vctk_p339_affect.wav",
432
+ "fig_file": "assets/en_US_vctk_p339.png",
433
+ "str_voice": "en_US_vctk_p339"
434
+ },
435
+ "36": {
436
+ "voice": "en_US/vctk_low#p311",
437
+ "emotion": [
438
+ 0.19333276897668839,
439
+ 0.3314017504453659,
440
+ 0.3488151580095291
441
+ ],
442
+ "tgt_wav": "assets/wavs/en_US_vctk_p311.wav",
443
+ "affect_wav": "assets/wavs/en_US_vctk_p311_affect.wav",
444
+ "fig_file": "assets/en_US_vctk_p311.png",
445
+ "str_voice": "en_US_vctk_p311"
446
+ },
447
+ "37": {
448
+ "voice": "en_US/vctk_low#p294",
449
+ "emotion": [
450
+ 0.016178522258996964,
451
+ 0.21839848160743713,
452
+ 0.30392274260520935
453
+ ],
454
+ "tgt_wav": "assets/wavs/en_US_vctk_p294.wav",
455
+ "affect_wav": "assets/wavs/en_US_vctk_p294_affect.wav",
456
+ "fig_file": "assets/en_US_vctk_p294.png",
457
+ "str_voice": "en_US_vctk_p294"
458
+ },
459
+ "38": {
460
+ "voice": "en_US/vctk_low#p305",
461
+ "emotion": [
462
+ 0.023888904601335526,
463
+ 0.172433540225029,
464
+ 0.3457363098859787
465
+ ],
466
+ "tgt_wav": "assets/wavs/en_US_vctk_p305.wav",
467
+ "affect_wav": "assets/wavs/en_US_vctk_p305_affect.wav",
468
+ "fig_file": "assets/en_US_vctk_p305.png",
469
+ "str_voice": "en_US_vctk_p305"
470
+ },
471
+ "39": {
472
+ "voice": "en_US/vctk_low#p266",
473
+ "emotion": [
474
+ 0.001926269382238388,
475
+ 0.17083801329135895,
476
+ 0.31496064364910126
477
+ ],
478
+ "tgt_wav": "assets/wavs/en_US_vctk_p266.wav",
479
+ "affect_wav": "assets/wavs/en_US_vctk_p266_affect.wav",
480
+ "fig_file": "assets/en_US_vctk_p266.png",
481
+ "str_voice": "en_US_vctk_p266"
482
+ },
483
+ "40": {
484
+ "voice": "en_US/vctk_low#p335",
485
+ "emotion": [
486
+ 0.025709589943289757,
487
+ 0.21180857717990875,
488
+ 0.3755295127630234
489
+ ],
490
+ "tgt_wav": "assets/wavs/en_US_vctk_p335.wav",
491
+ "affect_wav": "assets/wavs/en_US_vctk_p335_affect.wav",
492
+ "fig_file": "assets/en_US_vctk_p335.png",
493
+ "str_voice": "en_US_vctk_p335"
494
+ },
495
+ "41": {
496
+ "voice": "en_US/vctk_low#p334",
497
+ "emotion": [
498
+ 0.11676321178674698,
499
+ 0.2832619696855545,
500
+ 0.2786499857902527
501
+ ],
502
+ "tgt_wav": "assets/wavs/en_US_vctk_p334.wav",
503
+ "affect_wav": "assets/wavs/en_US_vctk_p334_affect.wav",
504
+ "fig_file": "assets/en_US_vctk_p334.png",
505
+ "str_voice": "en_US_vctk_p334"
506
+ },
507
+ "42": {
508
+ "voice": "en_US/vctk_low#p318",
509
+ "emotion": [
510
+ 0.040294808335602283,
511
+ 0.24496475607156754,
512
+ 0.31459513306617737
513
+ ],
514
+ "tgt_wav": "assets/wavs/en_US_vctk_p318.wav",
515
+ "affect_wav": "assets/wavs/en_US_vctk_p318_affect.wav",
516
+ "fig_file": "assets/en_US_vctk_p318.png",
517
+ "str_voice": "en_US_vctk_p318"
518
+ },
519
+ "43": {
520
+ "voice": "en_US/vctk_low#p323",
521
+ "emotion": [
522
+ 0.02241475321352482,
523
+ 0.17832424491643906,
524
+ 0.2849506735801697
525
+ ],
526
+ "tgt_wav": "assets/wavs/en_US_vctk_p323.wav",
527
+ "affect_wav": "assets/wavs/en_US_vctk_p323_affect.wav",
528
+ "fig_file": "assets/en_US_vctk_p323.png",
529
+ "str_voice": "en_US_vctk_p323"
530
+ },
531
+ "44": {
532
+ "voice": "en_US/vctk_low#p351",
533
+ "emotion": [
534
+ 0.0,
535
+ 0.13668810948729515,
536
+ 0.32634344696998596
537
+ ],
538
+ "tgt_wav": "assets/wavs/en_US_vctk_p351.wav",
539
+ "affect_wav": "assets/wavs/en_US_vctk_p351_affect.wav",
540
+ "fig_file": "assets/en_US_vctk_p351.png",
541
+ "str_voice": "en_US_vctk_p351"
542
+ },
543
+ "45": {
544
+ "voice": "en_US/vctk_low#p333",
545
+ "emotion": [
546
+ 0.004481316544115543,
547
+ 0.20113587379455566,
548
+ 0.2926645800471306
549
+ ],
550
+ "tgt_wav": "assets/wavs/en_US_vctk_p333.wav",
551
+ "affect_wav": "assets/wavs/en_US_vctk_p333_affect.wav",
552
+ "fig_file": "assets/en_US_vctk_p333.png",
553
+ "str_voice": "en_US_vctk_p333"
554
+ },
555
+ "46": {
556
+ "voice": "en_US/vctk_low#p313",
557
+ "emotion": [
558
+ 0.0,
559
+ 0.11234099045395851,
560
+ 0.3038019388914108
561
+ ],
562
+ "tgt_wav": "assets/wavs/en_US_vctk_p313.wav",
563
+ "affect_wav": "assets/wavs/en_US_vctk_p313_affect.wav",
564
+ "fig_file": "assets/en_US_vctk_p313.png",
565
+ "str_voice": "en_US_vctk_p313"
566
+ },
567
+ "47": {
568
+ "voice": "en_US/vctk_low#p316",
569
+ "emotion": [
570
+ 0.13346926122903824,
571
+ 0.2938437908887863,
572
+ 0.3289429396390915
573
+ ],
574
+ "tgt_wav": "assets/wavs/en_US_vctk_p316.wav",
575
+ "affect_wav": "assets/wavs/en_US_vctk_p316_affect.wav",
576
+ "fig_file": "assets/en_US_vctk_p316.png",
577
+ "str_voice": "en_US_vctk_p316"
578
+ },
579
+ "48": {
580
+ "voice": "en_US/vctk_low#p244",
581
+ "emotion": [
582
+ 0.0,
583
+ 0.17397373169660568,
584
+ 0.2858740836381912
585
+ ],
586
+ "tgt_wav": "assets/wavs/en_US_vctk_p244.wav",
587
+ "affect_wav": "assets/wavs/en_US_vctk_p244_affect.wav",
588
+ "fig_file": "assets/en_US_vctk_p244.png",
589
+ "str_voice": "en_US_vctk_p244"
590
+ },
591
+ "49": {
592
+ "voice": "en_US/vctk_low#p307",
593
+ "emotion": [
594
+ 0.002823776798322797,
595
+ 0.19307326525449753,
596
+ 0.31698666512966156
597
+ ],
598
+ "tgt_wav": "assets/wavs/en_US_vctk_p307.wav",
599
+ "affect_wav": "assets/wavs/en_US_vctk_p307_affect.wav",
600
+ "fig_file": "assets/en_US_vctk_p307.png",
601
+ "str_voice": "en_US_vctk_p307"
602
+ },
603
+ "50": {
604
+ "voice": "en_US/vctk_low#p363",
605
+ "emotion": [
606
+ 0.07410074770450592,
607
+ 0.23522694408893585,
608
+ 0.28694698959589005
609
+ ],
610
+ "tgt_wav": "assets/wavs/en_US_vctk_p363.wav",
611
+ "affect_wav": "assets/wavs/en_US_vctk_p363_affect.wav",
612
+ "fig_file": "assets/en_US_vctk_p363.png",
613
+ "str_voice": "en_US_vctk_p363"
614
+ },
615
+ "51": {
616
+ "voice": "en_US/vctk_low#p336",
617
+ "emotion": [
618
+ 0.0,
619
+ 0.18255245685577393,
620
+ 0.3172556608915329
621
+ ],
622
+ "tgt_wav": "assets/wavs/en_US_vctk_p336.wav",
623
+ "affect_wav": "assets/wavs/en_US_vctk_p336_affect.wav",
624
+ "fig_file": "assets/en_US_vctk_p336.png",
625
+ "str_voice": "en_US_vctk_p336"
626
+ },
627
+ "52": {
628
+ "voice": "en_US/vctk_low#p312",
629
+ "emotion": [
630
+ 0.0,
631
+ 0.1412380039691925,
632
+ 0.31754522025585175
633
+ ],
634
+ "tgt_wav": "assets/wavs/en_US_vctk_p312.wav",
635
+ "affect_wav": "assets/wavs/en_US_vctk_p312_affect.wav",
636
+ "fig_file": "assets/en_US_vctk_p312.png",
637
+ "str_voice": "en_US_vctk_p312"
638
+ },
639
+ "53": {
640
+ "voice": "en_US/vctk_low#p267",
641
+ "emotion": [
642
+ 0.0,
643
+ 0.16513923555612564,
644
+ 0.2956678122282028
645
+ ],
646
+ "tgt_wav": "assets/wavs/en_US_vctk_p267.wav",
647
+ "affect_wav": "assets/wavs/en_US_vctk_p267_affect.wav",
648
+ "fig_file": "assets/en_US_vctk_p267.png",
649
+ "str_voice": "en_US_vctk_p267"
650
+ },
651
+ "54": {
652
+ "voice": "en_US/vctk_low#p297",
653
+ "emotion": [
654
+ 0.0,
655
+ 0.16045919805765152,
656
+ 0.29962848126888275
657
+ ],
658
+ "tgt_wav": "assets/wavs/en_US_vctk_p297.wav",
659
+ "affect_wav": "assets/wavs/en_US_vctk_p297_affect.wav",
660
+ "fig_file": "assets/en_US_vctk_p297.png",
661
+ "str_voice": "en_US_vctk_p297"
662
+ },
663
+ "55": {
664
+ "voice": "en_US/vctk_low#p275",
665
+ "emotion": [
666
+ 0.1402822844684124,
667
+ 0.30610391497612,
668
+ 0.3047277480363846
669
+ ],
670
+ "tgt_wav": "assets/wavs/en_US_vctk_p275.wav",
671
+ "affect_wav": "assets/wavs/en_US_vctk_p275_affect.wav",
672
+ "fig_file": "assets/en_US_vctk_p275.png",
673
+ "str_voice": "en_US_vctk_p275"
674
+ },
675
+ "56": {
676
+ "voice": "en_US/vctk_low#p295",
677
+ "emotion": [
678
+ 0.0,
679
+ 0.14374630898237228,
680
+ 0.28910958766937256
681
+ ],
682
+ "tgt_wav": "assets/wavs/en_US_vctk_p295.wav",
683
+ "affect_wav": "assets/wavs/en_US_vctk_p295_affect.wav",
684
+ "fig_file": "assets/en_US_vctk_p295.png",
685
+ "str_voice": "en_US_vctk_p295"
686
+ },
687
+ "57": {
688
+ "voice": "en_US/vctk_low#p288",
689
+ "emotion": [
690
+ 0.03262748196721077,
691
+ 0.21881252527236938,
692
+ 0.35440801084041595
693
+ ],
694
+ "tgt_wav": "assets/wavs/en_US_vctk_p288.wav",
695
+ "affect_wav": "assets/wavs/en_US_vctk_p288_affect.wav",
696
+ "fig_file": "assets/en_US_vctk_p288.png",
697
+ "str_voice": "en_US_vctk_p288"
698
+ },
699
+ "58": {
700
+ "voice": "en_US/vctk_low#p258",
701
+ "emotion": [
702
+ 0.1531713381409645,
703
+ 0.3358590602874756,
704
+ 0.3292926698923111
705
+ ],
706
+ "tgt_wav": "assets/wavs/en_US_vctk_p258.wav",
707
+ "affect_wav": "assets/wavs/en_US_vctk_p258_affect.wav",
708
+ "fig_file": "assets/en_US_vctk_p258.png",
709
+ "str_voice": "en_US_vctk_p258"
710
+ },
711
+ "59": {
712
+ "voice": "en_US/vctk_low#p301",
713
+ "emotion": [
714
+ 0.06317921355366707,
715
+ 0.2511940523982048,
716
+ 0.38544002175331116
717
+ ],
718
+ "tgt_wav": "assets/wavs/en_US_vctk_p301.wav",
719
+ "affect_wav": "assets/wavs/en_US_vctk_p301_affect.wav",
720
+ "fig_file": "assets/en_US_vctk_p301.png",
721
+ "str_voice": "en_US_vctk_p301"
722
+ },
723
+ "60": {
724
+ "voice": "en_US/vctk_low#p232",
725
+ "emotion": [
726
+ 0.13394508510828018,
727
+ 0.3066969960927963,
728
+ 0.32338330149650574
729
+ ],
730
+ "tgt_wav": "assets/wavs/en_US_vctk_p232.wav",
731
+ "affect_wav": "assets/wavs/en_US_vctk_p232_affect.wav",
732
+ "fig_file": "assets/en_US_vctk_p232.png",
733
+ "str_voice": "en_US_vctk_p232"
734
+ },
735
+ "61": {
736
+ "voice": "en_US/vctk_low#p292",
737
+ "emotion": [
738
+ 0.11305796727538109,
739
+ 0.26847032457590103,
740
+ 0.3294481784105301
741
+ ],
742
+ "tgt_wav": "assets/wavs/en_US_vctk_p292.wav",
743
+ "affect_wav": "assets/wavs/en_US_vctk_p292_affect.wav",
744
+ "fig_file": "assets/en_US_vctk_p292.png",
745
+ "str_voice": "en_US_vctk_p292"
746
+ },
747
+ "62": {
748
+ "voice": "en_US/vctk_low#p272",
749
+ "emotion": [
750
+ 0.07942825742065907,
751
+ 0.2677610442042351,
752
+ 0.2989608943462372
753
+ ],
754
+ "tgt_wav": "assets/wavs/en_US_vctk_p272.wav",
755
+ "affect_wav": "assets/wavs/en_US_vctk_p272_affect.wav",
756
+ "fig_file": "assets/en_US_vctk_p272.png",
757
+ "str_voice": "en_US_vctk_p272"
758
+ },
759
+ "63": {
760
+ "voice": "en_US/vctk_low#p278",
761
+ "emotion": [
762
+ 0.16840693354606628,
763
+ 0.3352172076702118,
764
+ 0.4037800282239914
765
+ ],
766
+ "tgt_wav": "assets/wavs/en_US_vctk_p278.wav",
767
+ "affect_wav": "assets/wavs/en_US_vctk_p278_affect.wav",
768
+ "fig_file": "assets/en_US_vctk_p278.png",
769
+ "str_voice": "en_US_vctk_p278"
770
+ },
771
+ "64": {
772
+ "voice": "en_US/vctk_low#p280",
773
+ "emotion": [
774
+ 0.0,
775
+ 0.15486077219247818,
776
+ 0.2730677127838135
777
+ ],
778
+ "tgt_wav": "assets/wavs/en_US_vctk_p280.wav",
779
+ "affect_wav": "assets/wavs/en_US_vctk_p280_affect.wav",
780
+ "fig_file": "assets/en_US_vctk_p280.png",
781
+ "str_voice": "en_US_vctk_p280"
782
+ },
783
+ "65": {
784
+ "voice": "en_US/vctk_low#p341",
785
+ "emotion": [
786
+ 0.02341086696833372,
787
+ 0.22712908685207367,
788
+ 0.37220966815948486
789
+ ],
790
+ "tgt_wav": "assets/wavs/en_US_vctk_p341.wav",
791
+ "affect_wav": "assets/wavs/en_US_vctk_p341_affect.wav",
792
+ "fig_file": "assets/en_US_vctk_p341.png",
793
+ "str_voice": "en_US_vctk_p341"
794
+ },
795
+ "66": {
796
+ "voice": "en_US/vctk_low#p268",
797
+ "emotion": [
798
+ 0.04387800768017769,
799
+ 0.22660920768976212,
800
+ 0.3172284960746765
801
+ ],
802
+ "tgt_wav": "assets/wavs/en_US_vctk_p268.wav",
803
+ "affect_wav": "assets/wavs/en_US_vctk_p268_affect.wav",
804
+ "fig_file": "assets/en_US_vctk_p268.png",
805
+ "str_voice": "en_US_vctk_p268"
806
+ },
807
+ "67": {
808
+ "voice": "en_US/vctk_low#p298",
809
+ "emotion": [
810
+ 0.12263564020395279,
811
+ 0.27540920674800873,
812
+ 0.3565220236778259
813
+ ],
814
+ "tgt_wav": "assets/wavs/en_US_vctk_p298.wav",
815
+ "affect_wav": "assets/wavs/en_US_vctk_p298_affect.wav",
816
+ "fig_file": "assets/en_US_vctk_p298.png",
817
+ "str_voice": "en_US_vctk_p298"
818
+ },
819
+ "68": {
820
+ "voice": "en_US/vctk_low#p299",
821
+ "emotion": [
822
+ 0.0,
823
+ 0.17763455212116241,
824
+ 0.31587137281894684
825
+ ],
826
+ "tgt_wav": "assets/wavs/en_US_vctk_p299.wav",
827
+ "affect_wav": "assets/wavs/en_US_vctk_p299_affect.wav",
828
+ "fig_file": "assets/en_US_vctk_p299.png",
829
+ "str_voice": "en_US_vctk_p299"
830
+ },
831
+ "69": {
832
+ "voice": "en_US/vctk_low#p279",
833
+ "emotion": [
834
+ 0.20207873731851578,
835
+ 0.34253643453121185,
836
+ 0.34679803252220154
837
+ ],
838
+ "tgt_wav": "assets/wavs/en_US_vctk_p279.wav",
839
+ "affect_wav": "assets/wavs/en_US_vctk_p279_affect.wav",
840
+ "fig_file": "assets/en_US_vctk_p279.png",
841
+ "str_voice": "en_US_vctk_p279"
842
+ },
843
+ "70": {
844
+ "voice": "en_US/vctk_low#p285",
845
+ "emotion": [
846
+ 0.08038273826241493,
847
+ 0.26657819747924805,
848
+ 0.3702620416879654
849
+ ],
850
+ "tgt_wav": "assets/wavs/en_US_vctk_p285.wav",
851
+ "affect_wav": "assets/wavs/en_US_vctk_p285_affect.wav",
852
+ "fig_file": "assets/en_US_vctk_p285.png",
853
+ "str_voice": "en_US_vctk_p285"
854
+ },
855
+ "71": {
856
+ "voice": "en_US/vctk_low#p326",
857
+ "emotion": [
858
+ 0.1157725378870964,
859
+ 0.27149294316768646,
860
+ 0.27582621574401855
861
+ ],
862
+ "tgt_wav": "assets/wavs/en_US_vctk_p326.wav",
863
+ "affect_wav": "assets/wavs/en_US_vctk_p326_affect.wav",
864
+ "fig_file": "assets/en_US_vctk_p326.png",
865
+ "str_voice": "en_US_vctk_p326"
866
+ },
867
+ "72": {
868
+ "voice": "en_US/vctk_low#p300",
869
+ "emotion": [
870
+ 0.011198820546269417,
871
+ 0.19999412447214127,
872
+ 0.3253006041049957
873
+ ],
874
+ "tgt_wav": "assets/wavs/en_US_vctk_p300.wav",
875
+ "affect_wav": "assets/wavs/en_US_vctk_p300_affect.wav",
876
+ "fig_file": "assets/en_US_vctk_p300.png",
877
+ "str_voice": "en_US_vctk_p300"
878
+ },
879
+ "73": {
880
+ "voice": "en_US/vctk_low#s5",
881
+ "emotion": [
882
+ 0.09500165283679962,
883
+ 0.26117922365665436,
884
+ 0.3937820643186569
885
+ ],
886
+ "tgt_wav": "assets/wavs/en_US_vctk_s5.wav",
887
+ "affect_wav": "assets/wavs/en_US_vctk_s5_affect.wav",
888
+ "fig_file": "assets/en_US_vctk_s5.png",
889
+ "str_voice": "en_US_vctk_s5"
890
+ },
891
+ "74": {
892
+ "voice": "en_US/vctk_low#p230",
893
+ "emotion": [
894
+ 0.0,
895
+ 0.1322605162858963,
896
+ 0.3056934028863907
897
+ ],
898
+ "tgt_wav": "assets/wavs/en_US_vctk_p230.wav",
899
+ "affect_wav": "assets/wavs/en_US_vctk_p230_affect.wav",
900
+ "fig_file": "assets/en_US_vctk_p230.png",
901
+ "str_voice": "en_US_vctk_p230"
902
+ },
903
+ "75": {
904
+ "voice": "en_US/vctk_low#p254",
905
+ "emotion": [
906
+ 0.17835237830877304,
907
+ 0.332744300365448,
908
+ 0.32943306863307953
909
+ ],
910
+ "tgt_wav": "assets/wavs/en_US_vctk_p254.wav",
911
+ "affect_wav": "assets/wavs/en_US_vctk_p254_affect.wav",
912
+ "fig_file": "assets/en_US_vctk_p254.png",
913
+ "str_voice": "en_US_vctk_p254"
914
+ },
915
+ "76": {
916
+ "voice": "en_US/vctk_low#p269",
917
+ "emotion": [
918
+ 0.009964749217033386,
919
+ 0.17517337948083878,
920
+ 0.313457652926445
921
+ ],
922
+ "tgt_wav": "assets/wavs/en_US_vctk_p269.wav",
923
+ "affect_wav": "assets/wavs/en_US_vctk_p269_affect.wav",
924
+ "fig_file": "assets/en_US_vctk_p269.png",
925
+ "str_voice": "en_US_vctk_p269"
926
+ },
927
+ "77": {
928
+ "voice": "en_US/vctk_low#p293",
929
+ "emotion": [
930
+ 0.004189671948552132,
931
+ 0.19550783187150955,
932
+ 0.3696018159389496
933
+ ],
934
+ "tgt_wav": "assets/wavs/en_US_vctk_p293.wav",
935
+ "affect_wav": "assets/wavs/en_US_vctk_p293_affect.wav",
936
+ "fig_file": "assets/en_US_vctk_p293.png",
937
+ "str_voice": "en_US_vctk_p293"
938
+ },
939
+ "78": {
940
+ "voice": "en_US/vctk_low#p252",
941
+ "emotion": [
942
+ 0.11099078506231308,
943
+ 0.2923336923122406,
944
+ 0.25622545182704926
945
+ ],
946
+ "tgt_wav": "assets/wavs/en_US_vctk_p252.wav",
947
+ "affect_wav": "assets/wavs/en_US_vctk_p252_affect.wav",
948
+ "fig_file": "assets/en_US_vctk_p252.png",
949
+ "str_voice": "en_US_vctk_p252"
950
+ },
951
+ "79": {
952
+ "voice": "en_US/vctk_low#p345",
953
+ "emotion": [
954
+ 0.10635789856314659,
955
+ 0.30381371080875397,
956
+ 0.35318444669246674
957
+ ],
958
+ "tgt_wav": "assets/wavs/en_US_vctk_p345.wav",
959
+ "affect_wav": "assets/wavs/en_US_vctk_p345_affect.wav",
960
+ "fig_file": "assets/en_US_vctk_p345.png",
961
+ "str_voice": "en_US_vctk_p345"
962
+ },
963
+ "80": {
964
+ "voice": "en_US/vctk_low#p262",
965
+ "emotion": [
966
+ 0.0,
967
+ 0.17038797587156296,
968
+ 0.2913622707128525
969
+ ],
970
+ "tgt_wav": "assets/wavs/en_US_vctk_p262.wav",
971
+ "affect_wav": "assets/wavs/en_US_vctk_p262_affect.wav",
972
+ "fig_file": "assets/en_US_vctk_p262.png",
973
+ "str_voice": "en_US_vctk_p262"
974
+ },
975
+ "81": {
976
+ "voice": "en_US/vctk_low#p243",
977
+ "emotion": [
978
+ 0.10955308005213737,
979
+ 0.3093649595975876,
980
+ 0.34245653450489044
981
+ ],
982
+ "tgt_wav": "assets/wavs/en_US_vctk_p243.wav",
983
+ "affect_wav": "assets/wavs/en_US_vctk_p243_affect.wav",
984
+ "fig_file": "assets/en_US_vctk_p243.png",
985
+ "str_voice": "en_US_vctk_p243"
986
+ },
987
+ "82": {
988
+ "voice": "en_US/vctk_low#p227",
989
+ "emotion": [
990
+ 0.12342802435159683,
991
+ 0.3099231421947479,
992
+ 0.29126007854938507
993
+ ],
994
+ "tgt_wav": "assets/wavs/en_US_vctk_p227.wav",
995
+ "affect_wav": "assets/wavs/en_US_vctk_p227_affect.wav",
996
+ "fig_file": "assets/en_US_vctk_p227.png",
997
+ "str_voice": "en_US_vctk_p227"
998
+ },
999
+ "83": {
1000
+ "voice": "en_US/vctk_low#p343",
1001
+ "emotion": [
1002
+ 0.0678184274584055,
1003
+ 0.2594190612435341,
1004
+ 0.3394549489021301
1005
+ ],
1006
+ "tgt_wav": "assets/wavs/en_US_vctk_p343.wav",
1007
+ "affect_wav": "assets/wavs/en_US_vctk_p343_affect.wav",
1008
+ "fig_file": "assets/en_US_vctk_p343.png",
1009
+ "str_voice": "en_US_vctk_p343"
1010
+ },
1011
+ "84": {
1012
+ "voice": "en_US/vctk_low#p255",
1013
+ "emotion": [
1014
+ 0.05333407036960125,
1015
+ 0.2598796635866165,
1016
+ 0.34575600922107697
1017
+ ],
1018
+ "tgt_wav": "assets/wavs/en_US_vctk_p255.wav",
1019
+ "affect_wav": "assets/wavs/en_US_vctk_p255_affect.wav",
1020
+ "fig_file": "assets/en_US_vctk_p255.png",
1021
+ "str_voice": "en_US_vctk_p255"
1022
+ },
1023
+ "85": {
1024
+ "voice": "en_US/vctk_low#p229",
1025
+ "emotion": [
1026
+ 0.0,
1027
+ 0.15859586000442505,
1028
+ 0.32699747383594513
1029
+ ],
1030
+ "tgt_wav": "assets/wavs/en_US_vctk_p229.wav",
1031
+ "affect_wav": "assets/wavs/en_US_vctk_p229_affect.wav",
1032
+ "fig_file": "assets/en_US_vctk_p229.png",
1033
+ "str_voice": "en_US_vctk_p229"
1034
+ },
1035
+ "86": {
1036
+ "voice": "en_US/vctk_low#p240",
1037
+ "emotion": [
1038
+ 0.012523185461759567,
1039
+ 0.1857067495584488,
1040
+ 0.37508755922317505
1041
+ ],
1042
+ "tgt_wav": "assets/wavs/en_US_vctk_p240.wav",
1043
+ "affect_wav": "assets/wavs/en_US_vctk_p240_affect.wav",
1044
+ "fig_file": "assets/en_US_vctk_p240.png",
1045
+ "str_voice": "en_US_vctk_p240"
1046
+ },
1047
+ "87": {
1048
+ "voice": "en_US/vctk_low#p248",
1049
+ "emotion": [
1050
+ 0.08036761730909348,
1051
+ 0.259002685546875,
1052
+ 0.354027196764946
1053
+ ],
1054
+ "tgt_wav": "assets/wavs/en_US_vctk_p248.wav",
1055
+ "affect_wav": "assets/wavs/en_US_vctk_p248_affect.wav",
1056
+ "fig_file": "assets/en_US_vctk_p248.png",
1057
+ "str_voice": "en_US_vctk_p248"
1058
+ },
1059
+ "88": {
1060
+ "voice": "en_US/vctk_low#p253",
1061
+ "emotion": [
1062
+ 0.0348813533782959,
1063
+ 0.2233794927597046,
1064
+ 0.3074892535805702
1065
+ ],
1066
+ "tgt_wav": "assets/wavs/en_US_vctk_p253.wav",
1067
+ "affect_wav": "assets/wavs/en_US_vctk_p253_affect.wav",
1068
+ "fig_file": "assets/en_US_vctk_p253.png",
1069
+ "str_voice": "en_US_vctk_p253"
1070
+ },
1071
+ "89": {
1072
+ "voice": "en_US/vctk_low#p233",
1073
+ "emotion": [
1074
+ 0.0,
1075
+ 0.10816441476345062,
1076
+ 0.2894563674926758
1077
+ ],
1078
+ "tgt_wav": "assets/wavs/en_US_vctk_p233.wav",
1079
+ "affect_wav": "assets/wavs/en_US_vctk_p233_affect.wav",
1080
+ "fig_file": "assets/en_US_vctk_p233.png",
1081
+ "str_voice": "en_US_vctk_p233"
1082
+ },
1083
+ "90": {
1084
+ "voice": "en_US/vctk_low#p228",
1085
+ "emotion": [
1086
+ 0.0,
1087
+ 0.1351209357380867,
1088
+ 0.29203541576862335
1089
+ ],
1090
+ "tgt_wav": "assets/wavs/en_US_vctk_p228.wav",
1091
+ "affect_wav": "assets/wavs/en_US_vctk_p228_affect.wav",
1092
+ "fig_file": "assets/en_US_vctk_p228.png",
1093
+ "str_voice": "en_US_vctk_p228"
1094
+ },
1095
+ "91": {
1096
+ "voice": "en_US/vctk_low#p251",
1097
+ "emotion": [
1098
+ 0.10159945487976074,
1099
+ 0.29825548827648163,
1100
+ 0.29027314484119415
1101
+ ],
1102
+ "tgt_wav": "assets/wavs/en_US_vctk_p251.wav",
1103
+ "affect_wav": "assets/wavs/en_US_vctk_p251_affect.wav",
1104
+ "fig_file": "assets/en_US_vctk_p251.png",
1105
+ "str_voice": "en_US_vctk_p251"
1106
+ },
1107
+ "92": {
1108
+ "voice": "en_US/vctk_low#p282",
1109
+ "emotion": [
1110
+ 0.006283540278673172,
1111
+ 0.20235596597194672,
1112
+ 0.3499891757965088
1113
+ ],
1114
+ "tgt_wav": "assets/wavs/en_US_vctk_p282.wav",
1115
+ "affect_wav": "assets/wavs/en_US_vctk_p282_affect.wav",
1116
+ "fig_file": "assets/en_US_vctk_p282.png",
1117
+ "str_voice": "en_US_vctk_p282"
1118
+ },
1119
+ "93": {
1120
+ "voice": "en_US/vctk_low#p246",
1121
+ "emotion": [
1122
+ 0.1335299015045166,
1123
+ 0.29863762855529785,
1124
+ 0.276714026927948
1125
+ ],
1126
+ "tgt_wav": "assets/wavs/en_US_vctk_p246.wav",
1127
+ "affect_wav": "assets/wavs/en_US_vctk_p246_affect.wav",
1128
+ "fig_file": "assets/en_US_vctk_p246.png",
1129
+ "str_voice": "en_US_vctk_p246"
1130
+ },
1131
+ "94": {
1132
+ "voice": "en_US/vctk_low#p234",
1133
+ "emotion": [
1134
+ 0.046149978414177895,
1135
+ 0.21969492733478546,
1136
+ 0.29968127608299255
1137
+ ],
1138
+ "tgt_wav": "assets/wavs/en_US_vctk_p234.wav",
1139
+ "affect_wav": "assets/wavs/en_US_vctk_p234_affect.wav",
1140
+ "fig_file": "assets/en_US_vctk_p234.png",
1141
+ "str_voice": "en_US_vctk_p234"
1142
+ },
1143
+ "95": {
1144
+ "voice": "en_US/vctk_low#p226",
1145
+ "emotion": [
1146
+ 0.10644334554672241,
1147
+ 0.2899332344532013,
1148
+ 0.2954626679420471
1149
+ ],
1150
+ "tgt_wav": "assets/wavs/en_US_vctk_p226.wav",
1151
+ "affect_wav": "assets/wavs/en_US_vctk_p226_affect.wav",
1152
+ "fig_file": "assets/en_US_vctk_p226.png",
1153
+ "str_voice": "en_US_vctk_p226"
1154
+ },
1155
+ "96": {
1156
+ "voice": "en_US/vctk_low#p260",
1157
+ "emotion": [
1158
+ 0.09629054740071297,
1159
+ 0.286439448595047,
1160
+ 0.30624865740537643
1161
+ ],
1162
+ "tgt_wav": "assets/wavs/en_US_vctk_p260.wav",
1163
+ "affect_wav": "assets/wavs/en_US_vctk_p260_affect.wav",
1164
+ "fig_file": "assets/en_US_vctk_p260.png",
1165
+ "str_voice": "en_US_vctk_p260"
1166
+ },
1167
+ "97": {
1168
+ "voice": "en_US/vctk_low#p245",
1169
+ "emotion": [
1170
+ 0.13227056711912155,
1171
+ 0.30784252285957336,
1172
+ 0.36357882618904114
1173
+ ],
1174
+ "tgt_wav": "assets/wavs/en_US_vctk_p245.wav",
1175
+ "affect_wav": "assets/wavs/en_US_vctk_p245_affect.wav",
1176
+ "fig_file": "assets/en_US_vctk_p245.png",
1177
+ "str_voice": "en_US_vctk_p245"
1178
+ },
1179
+ "98": {
1180
+ "voice": "en_US/vctk_low#p241",
1181
+ "emotion": [
1182
+ 0.03028629533946514,
1183
+ 0.2272450551390648,
1184
+ 0.2857053875923157
1185
+ ],
1186
+ "tgt_wav": "assets/wavs/en_US_vctk_p241.wav",
1187
+ "affect_wav": "assets/wavs/en_US_vctk_p241_affect.wav",
1188
+ "fig_file": "assets/en_US_vctk_p241.png",
1189
+ "str_voice": "en_US_vctk_p241"
1190
+ },
1191
+ "99": {
1192
+ "voice": "en_US/vctk_low#p303",
1193
+ "emotion": [
1194
+ 0.005720391869544983,
1195
+ 0.20771541446447372,
1196
+ 0.3387496918439865
1197
+ ],
1198
+ "tgt_wav": "assets/wavs/en_US_vctk_p303.wav",
1199
+ "affect_wav": "assets/wavs/en_US_vctk_p303_affect.wav",
1200
+ "fig_file": "assets/en_US_vctk_p303.png",
1201
+ "str_voice": "en_US_vctk_p303"
1202
+ },
1203
+ "100": {
1204
+ "voice": "en_US/vctk_low#p265",
1205
+ "emotion": [
1206
+ 0.059831541031599045,
1207
+ 0.24712997674942017,
1208
+ 0.38666336238384247
1209
+ ],
1210
+ "tgt_wav": "assets/wavs/en_US_vctk_p265.wav",
1211
+ "affect_wav": "assets/wavs/en_US_vctk_p265_affect.wav",
1212
+ "fig_file": "assets/en_US_vctk_p265.png",
1213
+ "str_voice": "en_US_vctk_p265"
1214
+ },
1215
+ "101": {
1216
+ "voice": "en_US/vctk_low#p306",
1217
+ "emotion": [
1218
+ 0.0,
1219
+ 0.13954143971204758,
1220
+ 0.30395881831645966
1221
+ ],
1222
+ "tgt_wav": "assets/wavs/en_US_vctk_p306.wav",
1223
+ "affect_wav": "assets/wavs/en_US_vctk_p306_affect.wav",
1224
+ "fig_file": "assets/en_US_vctk_p306.png",
1225
+ "str_voice": "en_US_vctk_p306"
1226
+ },
1227
+ "102": {
1228
+ "voice": "en_US/vctk_low#p237",
1229
+ "emotion": [
1230
+ 0.20525796338915825,
1231
+ 0.345702163875103,
1232
+ 0.4544626623392105
1233
+ ],
1234
+ "tgt_wav": "assets/wavs/en_US_vctk_p237.wav",
1235
+ "affect_wav": "assets/wavs/en_US_vctk_p237_affect.wav",
1236
+ "fig_file": "assets/en_US_vctk_p237.png",
1237
+ "str_voice": "en_US_vctk_p237"
1238
+ },
1239
+ "103": {
1240
+ "voice": "en_US/vctk_low#p249",
1241
+ "emotion": [
1242
+ 0.0040201363153755665,
1243
+ 0.19356297701597214,
1244
+ 0.2982936501502991
1245
+ ],
1246
+ "tgt_wav": "assets/wavs/en_US_vctk_p249.wav",
1247
+ "affect_wav": "assets/wavs/en_US_vctk_p249_affect.wav",
1248
+ "fig_file": "assets/en_US_vctk_p249.png",
1249
+ "str_voice": "en_US_vctk_p249"
1250
+ },
1251
+ "104": {
1252
+ "voice": "en_US/vctk_low#p256",
1253
+ "emotion": [
1254
+ 0.14076407998800278,
1255
+ 0.3317306190729141,
1256
+ 0.2933240830898285
1257
+ ],
1258
+ "tgt_wav": "assets/wavs/en_US_vctk_p256.wav",
1259
+ "affect_wav": "assets/wavs/en_US_vctk_p256_affect.wav",
1260
+ "fig_file": "assets/en_US_vctk_p256.png",
1261
+ "str_voice": "en_US_vctk_p256"
1262
+ },
1263
+ "105": {
1264
+ "voice": "en_US/vctk_low#p302",
1265
+ "emotion": [
1266
+ 0.12388656847178936,
1267
+ 0.2888874188065529,
1268
+ 0.3455595523118973
1269
+ ],
1270
+ "tgt_wav": "assets/wavs/en_US_vctk_p302.wav",
1271
+ "affect_wav": "assets/wavs/en_US_vctk_p302_affect.wav",
1272
+ "fig_file": "assets/en_US_vctk_p302.png",
1273
+ "str_voice": "en_US_vctk_p302"
1274
+ },
1275
+ "106": {
1276
+ "voice": "en_US/vctk_low#p364",
1277
+ "emotion": [
1278
+ 0.09777292609214783,
1279
+ 0.25997892767190933,
1280
+ 0.3575831800699234
1281
+ ],
1282
+ "tgt_wav": "assets/wavs/en_US_vctk_p364.wav",
1283
+ "affect_wav": "assets/wavs/en_US_vctk_p364_affect.wav",
1284
+ "fig_file": "assets/en_US_vctk_p364.png",
1285
+ "str_voice": "en_US_vctk_p364"
1286
+ },
1287
+ "107": {
1288
+ "voice": "en_US/vctk_low#p225",
1289
+ "emotion": [
1290
+ 0.0,
1291
+ 0.18437553942203522,
1292
+ 0.29505328834056854
1293
+ ],
1294
+ "tgt_wav": "assets/wavs/en_US_vctk_p225.wav",
1295
+ "affect_wav": "assets/wavs/en_US_vctk_p225_affect.wav",
1296
+ "fig_file": "assets/en_US_vctk_p225.png",
1297
+ "str_voice": "en_US_vctk_p225"
1298
+ },
1299
+ "108": {
1300
+ "voice": "en_US/vctk_low#p362",
1301
+ "emotion": [
1302
+ 0.0,
1303
+ 0.13392474502325058,
1304
+ 0.2865803688764572
1305
+ ],
1306
+ "tgt_wav": "assets/wavs/en_US_vctk_p362.wav",
1307
+ "affect_wav": "assets/wavs/en_US_vctk_p362_affect.wav",
1308
+ "fig_file": "assets/en_US_vctk_p362.png",
1309
+ "str_voice": "en_US_vctk_p362"
1310
+ },
1311
+ "109": {
1312
+ "voice": "en_US/cmu-arctic_low#awbrms",
1313
+ "emotion": [
1314
+ 0.13197790831327438,
1315
+ 0.304988831281662,
1316
+ 0.29639899730682373
1317
+ ],
1318
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_awbrms.wav",
1319
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_awbrms_affect.wav",
1320
+ "fig_file": "assets/en_US_cmu-arctic_awbrms.png",
1321
+ "str_voice": "en_US_cmu-arctic_awbrms"
1322
+ },
1323
+ "110": {
1324
+ "voice": "en_US/cmu-arctic_low#slt",
1325
+ "emotion": [
1326
+ 0.0,
1327
+ 0.1312907449901104,
1328
+ 0.25008559226989746
1329
+ ],
1330
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_slt.wav",
1331
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_slt_affect.wav",
1332
+ "fig_file": "assets/en_US_cmu-arctic_slt.png",
1333
+ "str_voice": "en_US_cmu-arctic_slt"
1334
+ },
1335
+ "111": {
1336
+ "voice": "en_US/cmu-arctic_low#ksp",
1337
+ "emotion": [
1338
+ 0.12758950889110565,
1339
+ 0.3103189319372177,
1340
+ 0.31539691984653473
1341
+ ],
1342
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_ksp.wav",
1343
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_ksp_affect.wav",
1344
+ "fig_file": "assets/en_US_cmu-arctic_ksp.png",
1345
+ "str_voice": "en_US_cmu-arctic_ksp"
1346
+ },
1347
+ "112": {
1348
+ "voice": "en_US/cmu-arctic_low#clb",
1349
+ "emotion": [
1350
+ 0.0,
1351
+ 0.1583484560251236,
1352
+ 0.2977180629968643
1353
+ ],
1354
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_clb.wav",
1355
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_clb_affect.wav",
1356
+ "fig_file": "assets/en_US_cmu-arctic_clb.png",
1357
+ "str_voice": "en_US_cmu-arctic_clb"
1358
+ },
1359
+ "113": {
1360
+ "voice": "en_US/cmu-arctic_low#aew",
1361
+ "emotion": [
1362
+ 0.1128418818116188,
1363
+ 0.29078561067581177,
1364
+ 0.30981625616550446
1365
+ ],
1366
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_aew.wav",
1367
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_aew_affect.wav",
1368
+ "fig_file": "assets/en_US_cmu-arctic_aew.png",
1369
+ "str_voice": "en_US_cmu-arctic_aew"
1370
+ },
1371
+ "114": {
1372
+ "voice": "en_US/cmu-arctic_low#bdl",
1373
+ "emotion": [
1374
+ 0.106624536216259,
1375
+ 0.273576982319355,
1376
+ 0.33074524998664856
1377
+ ],
1378
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_bdl.wav",
1379
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_bdl_affect.wav",
1380
+ "fig_file": "assets/en_US_cmu-arctic_bdl.png",
1381
+ "str_voice": "en_US_cmu-arctic_bdl"
1382
+ },
1383
+ "115": {
1384
+ "voice": "en_US/cmu-arctic_low#lnh",
1385
+ "emotion": [
1386
+ 0.0669902004301548,
1387
+ 0.25475453585386276,
1388
+ 0.3285839110612869
1389
+ ],
1390
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_lnh.wav",
1391
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_lnh_affect.wav",
1392
+ "fig_file": "assets/en_US_cmu-arctic_lnh.png",
1393
+ "str_voice": "en_US_cmu-arctic_lnh"
1394
+ },
1395
+ "116": {
1396
+ "voice": "en_US/cmu-arctic_low#jmk",
1397
+ "emotion": [
1398
+ 0.1965325027704239,
1399
+ 0.36776699125766754,
1400
+ 0.33295588195323944
1401
+ ],
1402
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_jmk.wav",
1403
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_jmk_affect.wav",
1404
+ "fig_file": "assets/en_US_cmu-arctic_jmk.png",
1405
+ "str_voice": "en_US_cmu-arctic_jmk"
1406
+ },
1407
+ "117": {
1408
+ "voice": "en_US/cmu-arctic_low#rxr",
1409
+ "emotion": [
1410
+ 0.15873733907938004,
1411
+ 0.33174562454223633,
1412
+ 0.2877024859189987
1413
+ ],
1414
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_rxr.wav",
1415
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_rxr_affect.wav",
1416
+ "fig_file": "assets/en_US_cmu-arctic_rxr.png",
1417
+ "str_voice": "en_US_cmu-arctic_rxr"
1418
+ },
1419
+ "118": {
1420
+ "voice": "en_US/cmu-arctic_low#fem",
1421
+ "emotion": [
1422
+ 0.20055567473173141,
1423
+ 0.3428553193807602,
1424
+ 0.2870483994483948
1425
+ ],
1426
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_fem.wav",
1427
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_fem_affect.wav",
1428
+ "fig_file": "assets/en_US_cmu-arctic_fem.png",
1429
+ "str_voice": "en_US_cmu-arctic_fem"
1430
+ },
1431
+ "119": {
1432
+ "voice": "en_US/cmu-arctic_low#ljm",
1433
+ "emotion": [
1434
+ 0.0,
1435
+ 0.16305125504732132,
1436
+ 0.2600909247994423
1437
+ ],
1438
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_ljm.wav",
1439
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_ljm_affect.wav",
1440
+ "fig_file": "assets/en_US_cmu-arctic_ljm.png",
1441
+ "str_voice": "en_US_cmu-arctic_ljm"
1442
+ },
1443
+ "120": {
1444
+ "voice": "en_US/cmu-arctic_low#slp",
1445
+ "emotion": [
1446
+ 0.0,
1447
+ 0.17321083694696426,
1448
+ 0.28240734338760376
1449
+ ],
1450
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_slp.wav",
1451
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_slp_affect.wav",
1452
+ "fig_file": "assets/en_US_cmu-arctic_slp.png",
1453
+ "str_voice": "en_US_cmu-arctic_slp"
1454
+ },
1455
+ "121": {
1456
+ "voice": "en_US/cmu-arctic_low#ahw",
1457
+ "emotion": [
1458
+ 0.16948115825653076,
1459
+ 0.3521708846092224,
1460
+ 0.2951629012823105
1461
+ ],
1462
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_ahw.wav",
1463
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_ahw_affect.wav",
1464
+ "fig_file": "assets/en_US_cmu-arctic_ahw.png",
1465
+ "str_voice": "en_US_cmu-arctic_ahw"
1466
+ },
1467
+ "122": {
1468
+ "voice": "en_US/cmu-arctic_low#axb",
1469
+ "emotion": [
1470
+ 0.0,
1471
+ 0.1796008199453354,
1472
+ 0.2633693292737007
1473
+ ],
1474
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_axb.wav",
1475
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_axb_affect.wav",
1476
+ "fig_file": "assets/en_US_cmu-arctic_axb.png",
1477
+ "str_voice": "en_US_cmu-arctic_axb"
1478
+ },
1479
+ "123": {
1480
+ "voice": "en_US/cmu-arctic_low#aup",
1481
+ "emotion": [
1482
+ 0.09407182969152927,
1483
+ 0.2742197513580322,
1484
+ 0.3380313515663147
1485
+ ],
1486
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_aup.wav",
1487
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_aup_affect.wav",
1488
+ "fig_file": "assets/en_US_cmu-arctic_aup.png",
1489
+ "str_voice": "en_US_cmu-arctic_aup"
1490
+ },
1491
+ "124": {
1492
+ "voice": "en_US/cmu-arctic_low#eey",
1493
+ "emotion": [
1494
+ 0.0,
1495
+ 0.1493760570883751,
1496
+ 0.27280495315790176
1497
+ ],
1498
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_eey.wav",
1499
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_eey_affect.wav",
1500
+ "fig_file": "assets/en_US_cmu-arctic_eey.png",
1501
+ "str_voice": "en_US_cmu-arctic_eey"
1502
+ },
1503
+ "125": {
1504
+ "voice": "en_US/cmu-arctic_low#gka",
1505
+ "emotion": [
1506
+ 0.16550742089748383,
1507
+ 0.33900073170661926,
1508
+ 0.35270561277866364
1509
+ ],
1510
+ "tgt_wav": "assets/wavs/en_US_cmu_arctic_gka.wav",
1511
+ "affect_wav": "assets/wavs/en_US_cmu_arctic_gka_affect.wav",
1512
+ "fig_file": "assets/en_US_cmu-arctic_gka.png",
1513
+ "str_voice": "en_US_cmu-arctic_gka"
1514
+ },
1515
+ "126": {
1516
+ "voice": "en_US/hifi-tts_low#9017",
1517
+ "emotion": [
1518
+ 0.16459977999329567,
1519
+ 0.3419733792543411,
1520
+ 0.3219582810997963
1521
+ ],
1522
+ "tgt_wav": "assets/wavs/en_US_hifi-tts_9017.wav",
1523
+ "affect_wav": "assets/wavs/en_US_hifi-tts_9017_affect.wav",
1524
+ "fig_file": "assets/en_US_hifi-tts_9017.png",
1525
+ "str_voice": "en_US_hifi-tts_9017"
1526
+ },
1527
+ "127": {
1528
+ "voice": "en_US/hifi-tts_low#6097",
1529
+ "emotion": [
1530
+ 0.2978478819131851,
1531
+ 0.43715597689151764,
1532
+ 0.3822030872106552
1533
+ ],
1534
+ "tgt_wav": "assets/wavs/en_US_hifi-tts_6097.wav",
1535
+ "affect_wav": "assets/wavs/en_US_hifi-tts_6097_affect.wav",
1536
+ "fig_file": "assets/en_US_hifi-tts_6097.png",
1537
+ "str_voice": "en_US_hifi-tts_6097"
1538
+ },
1539
+ "128": {
1540
+ "voice": "en_US/hifi-tts_low#92",
1541
+ "emotion": [
1542
+ 0.025951625779271126,
1543
+ 0.2329779863357544,
1544
+ 0.3678264021873474
1545
+ ],
1546
+ "tgt_wav": "assets/wavs/en_US_hifi-tts_92.wav",
1547
+ "affect_wav": "assets/wavs/en_US_hifi-tts_92_affect.wav",
1548
+ "fig_file": "assets/en_US_hifi-tts_92.png",
1549
+ "str_voice": "en_US_hifi-tts_92"
1550
+ },
1551
+ "129": {
1552
+ "voice": "en_US/m-ailabs_low#elliot_miller",
1553
+ "emotion": [
1554
+ 0.2254963144659996,
1555
+ 0.38626913726329803,
1556
+ 0.25056935101747513
1557
+ ],
1558
+ "tgt_wav": "assets/wavs/en_US_m-ailabs_elliot_miller.wav",
1559
+ "affect_wav": "assets/wavs/en_US_m-ailabs_elliot_miller_affect.wav",
1560
+ "fig_file": "assets/en_US_m-ailabs_elliot_miller.png",
1561
+ "str_voice": "en_US_m-ailabs_elliot_miller"
1562
+ },
1563
+ "130": {
1564
+ "voice": "en_US/m-ailabs_low#judy_bieber",
1565
+ "emotion": [
1566
+ 0.0958283357322216,
1567
+ 0.2796173393726349,
1568
+ 0.3717383146286011
1569
+ ],
1570
+ "tgt_wav": "assets/wavs/en_US_m-ailabs_judy_bieber.wav",
1571
+ "affect_wav": "assets/wavs/en_US_m-ailabs_judy_bieber_affect.wav",
1572
+ "fig_file": "assets/en_US_m-ailabs_judy_bieber.png",
1573
+ "str_voice": "en_US_m-ailabs_judy_bieber"
1574
+ },
1575
+ "131": {
1576
+ "voice": "en_US/m-ailabs_low#mary_ann",
1577
+ "emotion": [
1578
+ 0.0,
1579
+ 0.15187790244817734,
1580
+ 0.3275550305843353
1581
+ ],
1582
+ "tgt_wav": "assets/wavs/en_US_m-ailabs_mary_ann.wav",
1583
+ "affect_wav": "assets/wavs/en_US_m-ailabs_mary_ann_affect.wav",
1584
+ "fig_file": "assets/en_US_m-ailabs_mary_ann.png",
1585
+ "str_voice": "en_US_m-ailabs_mary_ann"
1586
+ },
1587
+ "132": {
1588
+ "voice": "en_US/ljspeech_low",
1589
+ "emotion": [
1590
+ 0.11834734305739403,
1591
+ 0.30745357275009155,
1592
+ 0.441321462392807
1593
+ ],
1594
+ "tgt_wav": "assets/wavs/en_US_ljspeech.wav",
1595
+ "affect_wav": "assets/wavs/en_US_ljspeech_affect.wav",
1596
+ "fig_file": "assets/en_US_ljspeech.png",
1597
+ "str_voice": "en_US_ljspeech"
1598
+ },
1599
+ "133": {
1600
+ "voice": "en_UK/apope_low",
1601
+ "emotion": [
1602
+ 0.2869475930929184,
1603
+ 0.4288555979728699,
1604
+ 0.27198949456214905
1605
+ ],
1606
+ "tgt_wav": "assets/wavs/en_UK_apope.wav",
1607
+ "affect_wav": "assets/wavs/en_UK_apope_affect.wav",
1608
+ "fig_file": "assets/en_UK_apope.png",
1609
+ "str_voice": "en_UK_apope"
1610
+ }
1611
+ }
1612
+ }