Hugging Face's logo

myshell-ai
/
MeloTTS-English-v2

Text-to-Speech
Transformers
Korean
Inference Endpoints
Model card Files Community
2
ylacombe HF staff commited on
Commit
1d0d08a
•
1 Parent(s): 41a5932

Upload 2 files

Browse files
Files changed (2) hide show
  1. checkpoint.pth +3 -0
  2. config.json +189 -0
checkpoint.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:794226eb7c1745f3ca281b290613d5f39aa5b0d3b16a117009966f4aaf184757
3
+ size 207769356
config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "segment_size": 16384
4
+ },
5
+ "data": {
6
+ "sampling_rate": 44100,
7
+ "filter_length": 2048,
8
+ "hop_length": 512,
9
+ "add_blank": true,
10
+ "n_speakers": 256,
11
+ "spk2id": {
12
+ "EN-US": 0,
13
+ "EN-BR": 1,
14
+ "EN-INDIA": 2,
15
+ "EN-AU": 4
16
+ }
17
+ },
18
+ "model": {
19
+ "use_spk_conditioned_encoder": true,
20
+ "use_noise_scaled_mas": true,
21
+ "use_mel_posterior_encoder": false,
22
+ "use_duration_discriminator": true,
23
+ "inter_channels": 192,
24
+ "hidden_channels": 192,
25
+ "filter_channels": 768,
26
+ "n_heads": 2,
27
+ "n_layers": 6,
28
+ "n_layers_trans_flow": 3,
29
+ "kernel_size": 3,
30
+ "p_dropout": 0.1,
31
+ "resblock": "1",
32
+ "resblock_kernel_sizes": [
33
+ 3,
34
+ 7,
35
+ 11
36
+ ],
37
+ "resblock_dilation_sizes": [
38
+ [
39
+ 1,
40
+ 3,
41
+ 5
42
+ ],
43
+ [
44
+ 1,
45
+ 3,
46
+ 5
47
+ ],
48
+ [
49
+ 1,
50
+ 3,
51
+ 5
52
+ ]
53
+ ],
54
+ "upsample_rates": [
55
+ 8,
56
+ 8,
57
+ 2,
58
+ 2,
59
+ 2
60
+ ],
61
+ "upsample_initial_channel": 512,
62
+ "upsample_kernel_sizes": [
63
+ 16,
64
+ 16,
65
+ 8,
66
+ 2,
67
+ 2
68
+ ],
69
+ "n_layers_q": 3,
70
+ "use_spectral_norm": false,
71
+ "gin_channels": 256
72
+ },
73
+ "symbols": [
74
+ "_",
75
+ "AA",
76
+ "E",
77
+ "EE",
78
+ "En",
79
+ "N",
80
+ "OO",
81
+ "V",
82
+ "a",
83
+ "a:",
84
+ "aa",
85
+ "ae",
86
+ "ah",
87
+ "ai",
88
+ "an",
89
+ "ang",
90
+ "ao",
91
+ "aw",
92
+ "ay",
93
+ "b",
94
+ "by",
95
+ "c",
96
+ "ch",
97
+ "d",
98
+ "dh",
99
+ "dy",
100
+ "e",
101
+ "e:",
102
+ "eh",
103
+ "ei",
104
+ "en",
105
+ "eng",
106
+ "er",
107
+ "ey",
108
+ "f",
109
+ "g",
110
+ "gy",
111
+ "h",
112
+ "hh",
113
+ "hy",
114
+ "i",
115
+ "i0",
116
+ "i:",
117
+ "ia",
118
+ "ian",
119
+ "iang",
120
+ "iao",
121
+ "ie",
122
+ "ih",
123
+ "in",
124
+ "ing",
125
+ "iong",
126
+ "ir",
127
+ "iu",
128
+ "iy",
129
+ "j",
130
+ "jh",
131
+ "k",
132
+ "ky",
133
+ "l",
134
+ "m",
135
+ "my",
136
+ "n",
137
+ "ng",
138
+ "ny",
139
+ "o",
140
+ "o:",
141
+ "ong",
142
+ "ou",
143
+ "ow",
144
+ "oy",
145
+ "p",
146
+ "py",
147
+ "q",
148
+ "r",
149
+ "ry",
150
+ "s",
151
+ "sh",
152
+ "t",
153
+ "th",
154
+ "ts",
155
+ "ty",
156
+ "u",
157
+ "u:",
158
+ "ua",
159
+ "uai",
160
+ "uan",
161
+ "uang",
162
+ "uh",
163
+ "ui",
164
+ "un",
165
+ "uo",
166
+ "uw",
167
+ "v",
168
+ "van",
169
+ "ve",
170
+ "vn",
171
+ "w",
172
+ "x",
173
+ "y",
174
+ "z",
175
+ "zh",
176
+ "zy",
177
+ "!",
178
+ "?",
179
+ "…",
180
+ ",",
181
+ ".",
182
+ "'",
183
+ "-",
184
+ "SP",
185
+ "UNK"
186
+ ],
187
+ "num_tones": 11,
188
+ "num_languages": 3
189
+ }