Hugging Face's logo

thelamapi
/
neuvoice

Text-to-Speech
Transformers
ONNX
PyTorch
text
speech
tts
neuvoice
turkish
turkey
ai voice
text to speech
turkish tts
open source ai
huggingface model
voice cloning
realistic ai voice
speech synthesis
turkish ai
local ai
offline tts
low latency tts
multilingual tts
audio generation
speech model
real time tts
hf model
synthetic voice
Model card Files
xet
 Community
2
Lamapi commited on
Commit
555672f
·
verified ·
1 Parent(s): 32f7a91

Update onnx/tts.json

Browse files
Files changed (1) hide show
  1. onnx/tts.json +50 -16
onnx/tts.json CHANGED
@@ -20,14 +20,16 @@
20
  "idim": 256,
21
  "ksz": 5,
22
  "intermediate_dim": 1024,
23
- "num_layers": 6,
24
  "dilation_lst": [
25
  1,
26
  1,
27
  2,
28
  2,
29
  4,
30
- 4
 
 
31
  ]
32
  },
33
  "attn_encoder": {
@@ -55,13 +57,15 @@
55
  "idim": 256,
56
  "ksz": 5,
57
  "intermediate_dim": 1024,
58
- "num_layers": 6,
59
  "dilation_lst": [
60
  1,
61
  1,
62
  1,
63
  1,
64
  1,
 
 
65
  1
66
  ]
67
  },
@@ -72,14 +76,14 @@
72
  "style_value_dim": 256,
73
  "prototype_dim": 256,
74
  "n_units": 256,
75
- "n_heads": 2
76
  }
77
  },
78
  "speech_prompted_text_encoder": {
79
  "text_dim": 256,
80
  "style_dim": 256,
81
  "n_units": 256,
82
- "n_heads": 2
83
  },
84
  "uncond_masker": {
85
  "prob_both_uncond": 0.04,
@@ -99,8 +103,8 @@
99
  "odim": 512
100
  },
101
  "time_encoder": {
102
- "time_dim": 64,
103
- "hdim": 256
104
  },
105
  "main_blocks": {
106
  "n_blocks": 4,
@@ -125,12 +129,20 @@
125
  "idim": 512,
126
  "ksz": 5,
127
  "intermediate_dim": 2048,
128
- "num_layers": 4,
129
  "dilation_lst": [
130
  1,
131
  2,
132
  4,
133
- 8
 
 
 
 
 
 
 
 
134
  ]
135
  },
136
  "convnext_1": {
@@ -156,8 +168,12 @@
156
  "idim": 512,
157
  "ksz": 5,
158
  "intermediate_dim": 2048,
159
- "num_layers": 4,
160
  "dilation_lst": [
 
 
 
 
161
  1,
162
  1,
163
  1,
@@ -190,7 +206,7 @@
190
  },
191
  "ksz_init": 7,
192
  "ksz": 7,
193
- "num_layers": 10,
194
  "dilation_lst": [
195
  1,
196
  1,
@@ -201,6 +217,12 @@
201
  1,
202
  1,
203
  1,
 
 
 
 
 
 
204
  1
205
  ],
206
  "intermediate_dim": 2048,
@@ -211,7 +233,7 @@
211
  "decoder": {
212
  "ksz_init": 7,
213
  "ksz": 7,
214
- "num_layers": 10,
215
  "dilation_lst": [
216
  1,
217
  2,
@@ -222,6 +244,12 @@
222
  1,
223
  1,
224
  1,
 
 
 
 
 
 
225
  1
226
  ],
227
  "intermediate_dim": 2048,
@@ -250,13 +278,15 @@
250
  "idim": 64,
251
  "ksz": 5,
252
  "intermediate_dim": 256,
253
- "num_layers": 6,
254
  "dilation_lst": [
255
  1,
256
  1,
257
  1,
258
  1,
259
  1,
 
 
260
  1
261
  ]
262
  },
@@ -282,8 +312,12 @@
282
  "idim": 64,
283
  "ksz": 5,
284
  "intermediate_dim": 256,
285
- "num_layers": 4,
286
  "dilation_lst": [
 
 
 
 
287
  1,
288
  1,
289
  1,
@@ -297,7 +331,7 @@
297
  "style_value_dim": 16,
298
  "prototype_dim": 64,
299
  "n_units": 64,
300
- "n_heads": 2
301
  }
302
  },
303
  "predictor": {
@@ -305,7 +339,7 @@
305
  "n_style": 8,
306
  "style_dim": 16,
307
  "hdim": 128,
308
- "n_layer": 2
309
  }
310
  }
311
  }
 
20
  "idim": 256,
21
  "ksz": 5,
22
  "intermediate_dim": 1024,
23
+ "num_layers": 8,
24
  "dilation_lst": [
25
  1,
26
  1,
27
  2,
28
  2,
29
  4,
30
+ 4,
31
+ 8,
32
+ 8
33
  ]
34
  },
35
  "attn_encoder": {
 
57
  "idim": 256,
58
  "ksz": 5,
59
  "intermediate_dim": 1024,
60
+ "num_layers": 8,
61
  "dilation_lst": [
62
  1,
63
  1,
64
  1,
65
  1,
66
  1,
67
+ 1,
68
+ 1,
69
  1
70
  ]
71
  },
 
76
  "style_value_dim": 256,
77
  "prototype_dim": 256,
78
  "n_units": 256,
79
+ "n_heads": 4
80
  }
81
  },
82
  "speech_prompted_text_encoder": {
83
  "text_dim": 256,
84
  "style_dim": 256,
85
  "n_units": 256,
86
+ "n_heads": 4
87
  },
88
  "uncond_masker": {
89
  "prob_both_uncond": 0.04,
 
103
  "odim": 512
104
  },
105
  "time_encoder": {
106
+ "time_dim": 512,
107
+ "hdim": 2048
108
  },
109
  "main_blocks": {
110
  "n_blocks": 4,
 
129
  "idim": 512,
130
  "ksz": 5,
131
  "intermediate_dim": 2048,
132
+ "num_layers": 12,
133
  "dilation_lst": [
134
  1,
135
  2,
136
  4,
137
+ 8,
138
+ 16,
139
+ 32,
140
+ 64,
141
+ 128,
142
+ 256,
143
+ 512,
144
+ 1024,
145
+ 2048
146
  ]
147
  },
148
  "convnext_1": {
 
168
  "idim": 512,
169
  "ksz": 5,
170
  "intermediate_dim": 2048,
171
+ "num_layers": 8,
172
  "dilation_lst": [
173
+ 1,
174
+ 1,
175
+ 1,
176
+ 1,
177
  1,
178
  1,
179
  1,
 
206
  },
207
  "ksz_init": 7,
208
  "ksz": 7,
209
+ "num_layers": 16,
210
  "dilation_lst": [
211
  1,
212
  1,
 
217
  1,
218
  1,
219
  1,
220
+ 1,
221
+ 1,
222
+ 1,
223
+ 1,
224
+ 1,
225
+ 1,
226
  1
227
  ],
228
  "intermediate_dim": 2048,
 
233
  "decoder": {
234
  "ksz_init": 7,
235
  "ksz": 7,
236
+ "num_layers": 16,
237
  "dilation_lst": [
238
  1,
239
  2,
 
244
  1,
245
  1,
246
  1,
247
+ 1,
248
+ 1,
249
+ 1,
250
+ 1,
251
+ 1,
252
+ 1,
253
  1
254
  ],
255
  "intermediate_dim": 2048,
 
278
  "idim": 64,
279
  "ksz": 5,
280
  "intermediate_dim": 256,
281
+ "num_layers": 8,
282
  "dilation_lst": [
283
  1,
284
  1,
285
  1,
286
  1,
287
  1,
288
+ 1,
289
+ 1,
290
  1
291
  ]
292
  },
 
312
  "idim": 64,
313
  "ksz": 5,
314
  "intermediate_dim": 256,
315
+ "num_layers": 8,
316
  "dilation_lst": [
317
+ 1,
318
+ 1,
319
+ 1,
320
+ 1,
321
  1,
322
  1,
323
  1,
 
331
  "style_value_dim": 16,
332
  "prototype_dim": 64,
333
  "n_units": 64,
334
+ "n_heads": 4
335
  }
336
  },
337
  "predictor": {
 
339
  "n_style": 8,
340
  "style_dim": 16,
341
  "hdim": 128,
342
+ "n_layer": 16
343
  }
344
  }
345
  }