Jarbas commited on
Commit
81eb88c
·
verified ·
1 Parent(s): 32f32f0

ArkTTS ONNX for the phoonnx arktts engine

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,176 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - yue
4
+ - zh
5
+ - nl
6
+ - en
7
+ - fr
8
+ - de
9
+ - it
10
+ - ja
11
+ - ko
12
+ - pl
13
+ - es
14
+ license: apache-2.0
15
+ base_model:
16
+ - Audio8/Audio8-TTS-Preview-0.6b
17
+ tags:
18
+ - text-to-speech
19
+ - tts
20
+ - multilingual
21
+ - onnx
22
+ - phoonnx
23
+ - arktts
24
+ - voice-cloning
25
+ pipeline_tag: text-to-speech
26
+ library_name: phoonnx
27
+ ---
28
+
29
+ # phoonnx-audio8-tts
30
+
31
+ ONNX export of **[`Audio8/Audio8-TTS-Preview-0.6b`](https://huggingface.co/Audio8/Audio8-TTS-Preview-0.6b)**
32
+ — an 11-language text-to-speech model — packaged for the
33
+ [phoonnx](https://github.com/OpenVoiceOS/phoonnx) `arktts` engine.
34
+
35
+ Upstream ships PyTorch only. These graphs were exported with
36
+ [`scripts/conversion/arktts/export_arktts_onnx.py`](https://github.com/OpenVoiceOS/phoonnx/tree/dev/scripts/conversion/arktts),
37
+ using the same contract and the same tensor names as the official export of the model's
38
+ Basque fine-tune, [`itzune/zortzi-tts-onnx`](https://huggingface.co/itzune/zortzi-tts-onnx),
39
+ so graphs from either source are interchangeable.
40
+
41
+ Languages: Cantonese, Chinese, Dutch, English, French, German, Italian, Japanese, Korean,
42
+ Polish, Spanish.
43
+
44
+ ## Usage
45
+
46
+ ```python
47
+ from phoonnx.model_manager import TTSModelManager
48
+ from phoonnx.voice import TTSVoice
49
+
50
+ voice = TTSVoice.load(TTSModelManager().get_voice("arktts/audio8-maider/en"))
51
+ audio = voice.synthesize("The quick brown fox jumps over the lazy dog.")
52
+ ```
53
+
54
+ Voice ids are `arktts/audio8-{maider,antton}/{yue,zh,nl,en,fr,de,it,ja,ko,pl,es}`.
55
+
56
+ ## Files
57
+
58
+ ```
59
+ slow_ar_fp16.onnx 24-layer backbone, KV-cached (1.35 GB)
60
+ fast_ar_fp16.onnx 4-layer depth transformer (134 MB)
61
+ codec_decoder_fp16.onnx 10 codebooks -> 44.1 kHz waveform (278 MB)
62
+ tokenizer.json the model's own Qwen2 subword BPE
63
+ voices/maider.json reference codes + transcription
64
+ voices/antton.json reference codes + transcription
65
+ runtime_manifest.json precision and architecture metadata
66
+ ```
67
+
68
+ ## Voices, and where they come from
69
+
70
+ Upstream ships **no reference voices** — Audio8 is a cloning model, and a voice is the codec
71
+ codes of a short clip rather than an entry in a speaker table. To make it usable as a
72
+ preset-voice engine, this mirror carries two voices encoded from the reference clips
73
+ published in [`itzune/zortzi-tts`](https://huggingface.co/itzune/zortzi-tts):
74
+ `voices/maider.wav` and `voices/antton.wav`.
75
+
76
+ Those clips are from the **HiTZ-Aholab Basque TTS dataset** (CC BY 4.0). They condition
77
+ **timbre, not language** — the model infers the language from the text — so both voices are
78
+ offered for all eleven languages, and the WER table below is what they actually sound like
79
+ in each. The clips are Basque, which is not one of Audio8's languages; that is fine for
80
+ conditioning and is stated here so nobody has to work it out.
81
+
82
+ The two checkpoints carry a byte-identical `codec.pth`, so re-encoding those clips with
83
+ Audio8's own codec reproduces the codes upstream published for Zortzi: codebook 0 matches
84
+ exactly, and all ten codebooks agree on 99.2 to 99.6 % of frames. The published Zortzi codes
85
+ are shipped here for that reason — they are the reproducible artifact.
86
+
87
+ To add your own voice, encode a clip offline with
88
+ `scripts/conversion/arktts/mint_voice.py`. This mirror ships the codec **decoder** only;
89
+ cloning at synthesis time would need the encoder, which is not exported.
90
+
91
+ ## What was verified
92
+
93
+ Checked against the PyTorch checkpoint (float32) with
94
+ `scripts/conversion/arktts/verify_parity.py`, over 24 lockstep greedy decode steps:
95
+
96
+ | Tensor | max abs diff | greedy agreement | note |
97
+ |---|---|---|---|
98
+ | prompt `[1, 11, T]` | — | exact | identical to upstream's own processor output |
99
+ | slow AR logits | 0.118 | 24/24 | |
100
+ | slow AR hidden | 0.039 | — | |
101
+ | fast AR logits | 1.36 | 208/216 | every miss is a tie; worst margin 0.225 |
102
+ | codec decoder | 6.3e-4 | — | correlation 0.999999 |
103
+
104
+ The fast AR is noisier here than in the official Zortzi export because the half-precision
105
+ graph accumulates its RMS norms in half precision rather than single — the explicit
106
+ float32 round-trip upstream writes cannot survive the fp16 converter. It changes no greedy
107
+ decision that was not already a tie.
108
+
109
+ Intelligibility, one sentence per language per voice, transcribed with
110
+ `onnx-community/whisper-large-v3-turbo` on CPU:
111
+
112
+ | Language | WER | Note |
113
+ |---|---|---|
114
+ | en | 0.000 | |
115
+ | de | 0.000 | |
116
+ | es | 0.000 | |
117
+ | it | 0.000 | |
118
+ | nl | 0.125 | one clip, one recogniser slip |
119
+ | pl | 0.143 | the recogniser wrote "10" for "dziesięć" |
120
+ | fr | 0.333 | the recogniser heard a plural subject in both clips |
121
+
122
+ Cantonese, Chinese, Japanese and Korean are **not** covered by this run. Whisper is the only
123
+ model in `onnx-asr` that claims `yue` at all, and a Whisper-only Cantonese number would say
124
+ more about the recogniser than about the voice; when those are measured they should be
125
+ scored by character (CER), not by word.
126
+
127
+ CPU real-time factor on twelve cores: **13.1x** (Maider), **14.3x** (Antton) — slower than
128
+ the official Zortzi graphs at 8x, because this export carries the tied embedding matrix
129
+ twice and is not otherwise optimised. This is not a real-time model on CPU.
130
+
131
+ ## Licence and attribution
132
+
133
+ Apache-2.0, from the base model. The model is by **Audio8**; this repository only exports
134
+ and repackages it.
135
+
136
+ The reference clips that carry the two voices are from the **HiTZ-Aholab Basque TTS
137
+ dataset** (CC BY 4.0, [10.5281/zenodo.17952596](https://doi.org/10.5281/zenodo.17952596)),
138
+ by way of [`itzune/zortzi-tts`](https://huggingface.co/itzune/zortzi-tts). That licence
139
+ conditions the voices, so redistributing them *or audio generated with them* carries the
140
+ attribution below.
141
+
142
+ > The HiTZ-Aholab speech synthesis dataset was developed with funding from the Ministerio
143
+ > para la Transformación Digital y de la Función Pública and Plan de Recuperación,
144
+ > Transformación y Resiliencia — Funded by EU — NextGenerationEU within the framework of the
145
+ > project ILENIA (ref. 2022/TL22/00215335), and by a grant from the Department of Culture
146
+ > and Language Policy of the Basque Government (IKER-GAITU project).
147
+
148
+ ```bibtex
149
+ @dataset{navas_hernaez_2025_17952596,
150
+ author = {Navas, Eva and Hernaez Rioja, Inmaculada and Saratxaga, Ibon and
151
+ Sanchez, Jon and García Romillo, Víctor and Flores Ríos, Mariana and
152
+ Bellanco, Aitor},
153
+ title = {{HiTZ-Aholab speech synthesis dataset in Basque}},
154
+ month = dec, year = 2025, publisher = {Zenodo}, version = {1.0},
155
+ doi = {10.5281/zenodo.17952596},
156
+ url = {https://doi.org/10.5281/zenodo.17952596}
157
+ }
158
+ ```
159
+
160
+ ## Known limitations
161
+
162
+ Upstream's, carried over unchanged:
163
+
164
+ - **Numbers are mispronounced** — spell them out in the text.
165
+ - **No text normalization** — expand acronyms yourself ("TTS" as "te te ese").
166
+ - **Sampling is required** — greedy decoding never reaches end-of-speech. The engine's
167
+ defaults (temperature 0.8, top-p 0.95) are upstream's.
168
+
169
+ Specific to this export:
170
+
171
+ - **Half precision only.** The codec decoder is the fp16 graph from
172
+ `itzune/zortzi-tts-onnx`, which is valid here because both checkpoints carry the same
173
+ `codec.pth`; it is verified above against *this* model's PyTorch codec.
174
+ - **No codec encoder**, so cloning is an offline step.
175
+ - **`slow_ar_fp16.onnx` is about 270 MB larger than it needs to be** — the tracer
176
+ materialises the tied output projection separately from the embedding table.
codec_decoder_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e4ea54739b4b98188852474ce7855ff335831b09eb34f18e28c9ed482761a4e
3
+ size 278119518
fast_ar_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33dd894dc73dad4c3f74fc1a3505b88a2489684a441052bc94fc6700ea106ccd
3
+ size 134041582
runtime_manifest.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_family": "audio8_tts",
3
+ "activation_dtype": "float32",
4
+ "slow_logits_layout": "semantic_then_eos",
5
+ "slow_logits_size": 4097,
6
+ "kv_attention_layout": "valid_prefix",
7
+ "max_seq_len": 2048,
8
+ "num_layers": 24,
9
+ "num_fast_layers": 4,
10
+ "num_codebooks": 10,
11
+ "n_local_heads": 2,
12
+ "fast_n_local_heads": 2,
13
+ "head_dim": 64,
14
+ "fast_head_dim": 64,
15
+ "fast_dim": 896,
16
+ "vocab_size": 155776,
17
+ "codebook_size": 4096,
18
+ "semantic_begin_id": 151678,
19
+ "semantic_end_id": 155773,
20
+ "eos_token_id": 151645,
21
+ "pad_token_id": 151643,
22
+ "sample_rate": 44100,
23
+ "codec_sample_rate": 44100,
24
+ "codec_frame_size": 2048,
25
+ "codec_hop_length": 2048,
26
+ "ras_window_size": 10,
27
+ "ras_top_p": 0.9,
28
+ "ras_temperature": 1.0,
29
+ "default_precision": "fp16",
30
+ "available_precisions": [
31
+ "fp16"
32
+ ],
33
+ "slow_models": {
34
+ "fp16": "slow_ar_fp16.onnx"
35
+ },
36
+ "fast_models": {
37
+ "fp16": "fast_ar_fp16.onnx"
38
+ },
39
+ "default_codec_precision": "fp32",
40
+ "available_codec_precisions": [
41
+ "fp32"
42
+ ],
43
+ "codec_models": {
44
+ "fp32": "codec_decoder_fp32.onnx"
45
+ }
46
+ }
slow_ar_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b58e12eddca63b45d52a833d6f697b02d5431a8538b6cd8f1b115ecd9bded82
3
+ size 1348765672
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f24e08099d45a8adf3f52f5f0b03276e433bb9d689bb15fcbcc48ce58744588b
3
+ size 12217872
voices/antton.json ADDED
@@ -0,0 +1,739 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "antton",
3
+ "display_name": "Antton",
4
+ "reference_text": "Inguru hura gerrillarien esku izan da denbora luzean.",
5
+ "frames": 71,
6
+ "provenance": "HiTZ-Aholab Basque TTS (CC BY 4.0), clip NEU_11782.wav; codes as published by itzune/zortzi-tts-onnx",
7
+ "codes": [
8
+ [
9
+ 245,
10
+ 3560,
11
+ 1144,
12
+ 2657,
13
+ 3899,
14
+ 1079,
15
+ 3955,
16
+ 3941,
17
+ 57,
18
+ 3863,
19
+ 31,
20
+ 1620,
21
+ 965,
22
+ 747,
23
+ 4093,
24
+ 660,
25
+ 985,
26
+ 758,
27
+ 3058,
28
+ 2947,
29
+ 105,
30
+ 2180,
31
+ 3747,
32
+ 3610,
33
+ 3197,
34
+ 1169,
35
+ 3753,
36
+ 2915,
37
+ 202,
38
+ 1257,
39
+ 1984,
40
+ 3685,
41
+ 160,
42
+ 403,
43
+ 3072,
44
+ 3750,
45
+ 2536,
46
+ 491,
47
+ 2554,
48
+ 4091,
49
+ 1212,
50
+ 247,
51
+ 3871,
52
+ 2125,
53
+ 1330,
54
+ 609,
55
+ 2245,
56
+ 107,
57
+ 3543,
58
+ 741,
59
+ 1035,
60
+ 3688,
61
+ 182,
62
+ 607,
63
+ 3359,
64
+ 3019,
65
+ 2247,
66
+ 3131,
67
+ 272,
68
+ 3812,
69
+ 974,
70
+ 776,
71
+ 317,
72
+ 1353,
73
+ 3486,
74
+ 1428,
75
+ 1478,
76
+ 2421,
77
+ 3382,
78
+ 1909,
79
+ 3308
80
+ ],
81
+ [
82
+ 325,
83
+ 827,
84
+ 866,
85
+ 528,
86
+ 202,
87
+ 451,
88
+ 75,
89
+ 274,
90
+ 195,
91
+ 733,
92
+ 916,
93
+ 910,
94
+ 816,
95
+ 982,
96
+ 421,
97
+ 208,
98
+ 69,
99
+ 518,
100
+ 177,
101
+ 996,
102
+ 177,
103
+ 434,
104
+ 869,
105
+ 693,
106
+ 991,
107
+ 114,
108
+ 869,
109
+ 927,
110
+ 979,
111
+ 133,
112
+ 780,
113
+ 177,
114
+ 192,
115
+ 424,
116
+ 292,
117
+ 504,
118
+ 774,
119
+ 973,
120
+ 513,
121
+ 961,
122
+ 1000,
123
+ 222,
124
+ 502,
125
+ 733,
126
+ 519,
127
+ 869,
128
+ 421,
129
+ 607,
130
+ 996,
131
+ 884,
132
+ 32,
133
+ 88,
134
+ 774,
135
+ 9,
136
+ 89,
137
+ 985,
138
+ 1017,
139
+ 305,
140
+ 876,
141
+ 889,
142
+ 794,
143
+ 766,
144
+ 79,
145
+ 504,
146
+ 173,
147
+ 5,
148
+ 367,
149
+ 159,
150
+ 123,
151
+ 609,
152
+ 816
153
+ ],
154
+ [
155
+ 648,
156
+ 134,
157
+ 369,
158
+ 870,
159
+ 517,
160
+ 875,
161
+ 756,
162
+ 98,
163
+ 552,
164
+ 13,
165
+ 751,
166
+ 725,
167
+ 857,
168
+ 366,
169
+ 143,
170
+ 228,
171
+ 902,
172
+ 271,
173
+ 112,
174
+ 119,
175
+ 770,
176
+ 437,
177
+ 380,
178
+ 986,
179
+ 901,
180
+ 116,
181
+ 977,
182
+ 294,
183
+ 730,
184
+ 646,
185
+ 850,
186
+ 223,
187
+ 692,
188
+ 306,
189
+ 874,
190
+ 338,
191
+ 229,
192
+ 199,
193
+ 372,
194
+ 60,
195
+ 627,
196
+ 930,
197
+ 946,
198
+ 566,
199
+ 697,
200
+ 830,
201
+ 566,
202
+ 758,
203
+ 948,
204
+ 579,
205
+ 364,
206
+ 705,
207
+ 1011,
208
+ 422,
209
+ 254,
210
+ 606,
211
+ 693,
212
+ 271,
213
+ 887,
214
+ 463,
215
+ 787,
216
+ 720,
217
+ 754,
218
+ 651,
219
+ 90,
220
+ 270,
221
+ 528,
222
+ 709,
223
+ 510,
224
+ 905,
225
+ 181
226
+ ],
227
+ [
228
+ 263,
229
+ 276,
230
+ 1002,
231
+ 740,
232
+ 493,
233
+ 596,
234
+ 402,
235
+ 435,
236
+ 1023,
237
+ 369,
238
+ 462,
239
+ 228,
240
+ 777,
241
+ 536,
242
+ 96,
243
+ 344,
244
+ 438,
245
+ 442,
246
+ 236,
247
+ 248,
248
+ 286,
249
+ 162,
250
+ 1020,
251
+ 609,
252
+ 443,
253
+ 974,
254
+ 926,
255
+ 580,
256
+ 531,
257
+ 154,
258
+ 855,
259
+ 183,
260
+ 229,
261
+ 681,
262
+ 329,
263
+ 1015,
264
+ 517,
265
+ 269,
266
+ 270,
267
+ 897,
268
+ 312,
269
+ 480,
270
+ 311,
271
+ 345,
272
+ 156,
273
+ 166,
274
+ 354,
275
+ 269,
276
+ 710,
277
+ 889,
278
+ 76,
279
+ 266,
280
+ 883,
281
+ 377,
282
+ 138,
283
+ 739,
284
+ 936,
285
+ 744,
286
+ 576,
287
+ 469,
288
+ 60,
289
+ 559,
290
+ 253,
291
+ 316,
292
+ 574,
293
+ 312,
294
+ 534,
295
+ 824,
296
+ 36,
297
+ 151,
298
+ 214
299
+ ],
300
+ [
301
+ 784,
302
+ 932,
303
+ 548,
304
+ 377,
305
+ 198,
306
+ 948,
307
+ 143,
308
+ 270,
309
+ 83,
310
+ 362,
311
+ 123,
312
+ 1012,
313
+ 965,
314
+ 838,
315
+ 765,
316
+ 777,
317
+ 367,
318
+ 145,
319
+ 51,
320
+ 83,
321
+ 80,
322
+ 692,
323
+ 134,
324
+ 152,
325
+ 759,
326
+ 279,
327
+ 87,
328
+ 80,
329
+ 77,
330
+ 115,
331
+ 827,
332
+ 551,
333
+ 339,
334
+ 57,
335
+ 997,
336
+ 309,
337
+ 945,
338
+ 85,
339
+ 1007,
340
+ 562,
341
+ 989,
342
+ 207,
343
+ 850,
344
+ 95,
345
+ 69,
346
+ 1016,
347
+ 565,
348
+ 229,
349
+ 170,
350
+ 199,
351
+ 648,
352
+ 751,
353
+ 655,
354
+ 63,
355
+ 24,
356
+ 332,
357
+ 657,
358
+ 148,
359
+ 839,
360
+ 128,
361
+ 643,
362
+ 567,
363
+ 265,
364
+ 403,
365
+ 438,
366
+ 433,
367
+ 796,
368
+ 437,
369
+ 957,
370
+ 44,
371
+ 571
372
+ ],
373
+ [
374
+ 383,
375
+ 50,
376
+ 307,
377
+ 1021,
378
+ 492,
379
+ 81,
380
+ 37,
381
+ 367,
382
+ 941,
383
+ 214,
384
+ 134,
385
+ 134,
386
+ 618,
387
+ 868,
388
+ 621,
389
+ 114,
390
+ 992,
391
+ 150,
392
+ 96,
393
+ 489,
394
+ 569,
395
+ 1005,
396
+ 148,
397
+ 379,
398
+ 167,
399
+ 969,
400
+ 425,
401
+ 215,
402
+ 499,
403
+ 92,
404
+ 80,
405
+ 715,
406
+ 327,
407
+ 23,
408
+ 45,
409
+ 967,
410
+ 542,
411
+ 859,
412
+ 339,
413
+ 868,
414
+ 187,
415
+ 95,
416
+ 349,
417
+ 492,
418
+ 81,
419
+ 950,
420
+ 727,
421
+ 219,
422
+ 145,
423
+ 720,
424
+ 656,
425
+ 106,
426
+ 993,
427
+ 174,
428
+ 348,
429
+ 462,
430
+ 738,
431
+ 841,
432
+ 878,
433
+ 627,
434
+ 773,
435
+ 898,
436
+ 681,
437
+ 368,
438
+ 938,
439
+ 450,
440
+ 679,
441
+ 759,
442
+ 45,
443
+ 864,
444
+ 57
445
+ ],
446
+ [
447
+ 223,
448
+ 411,
449
+ 75,
450
+ 578,
451
+ 116,
452
+ 155,
453
+ 49,
454
+ 51,
455
+ 257,
456
+ 923,
457
+ 588,
458
+ 603,
459
+ 849,
460
+ 40,
461
+ 13,
462
+ 374,
463
+ 830,
464
+ 77,
465
+ 430,
466
+ 177,
467
+ 741,
468
+ 591,
469
+ 954,
470
+ 333,
471
+ 671,
472
+ 446,
473
+ 74,
474
+ 277,
475
+ 615,
476
+ 60,
477
+ 291,
478
+ 383,
479
+ 130,
480
+ 459,
481
+ 99,
482
+ 955,
483
+ 545,
484
+ 593,
485
+ 90,
486
+ 199,
487
+ 875,
488
+ 668,
489
+ 974,
490
+ 595,
491
+ 936,
492
+ 413,
493
+ 1007,
494
+ 610,
495
+ 529,
496
+ 631,
497
+ 85,
498
+ 503,
499
+ 422,
500
+ 750,
501
+ 11,
502
+ 220,
503
+ 99,
504
+ 348,
505
+ 591,
506
+ 534,
507
+ 329,
508
+ 271,
509
+ 718,
510
+ 847,
511
+ 632,
512
+ 469,
513
+ 192,
514
+ 367,
515
+ 946,
516
+ 755,
517
+ 186
518
+ ],
519
+ [
520
+ 603,
521
+ 492,
522
+ 223,
523
+ 537,
524
+ 997,
525
+ 915,
526
+ 115,
527
+ 99,
528
+ 99,
529
+ 612,
530
+ 137,
531
+ 258,
532
+ 757,
533
+ 795,
534
+ 863,
535
+ 905,
536
+ 623,
537
+ 535,
538
+ 89,
539
+ 860,
540
+ 996,
541
+ 564,
542
+ 140,
543
+ 832,
544
+ 238,
545
+ 653,
546
+ 720,
547
+ 241,
548
+ 52,
549
+ 409,
550
+ 805,
551
+ 151,
552
+ 305,
553
+ 555,
554
+ 951,
555
+ 203,
556
+ 910,
557
+ 651,
558
+ 722,
559
+ 729,
560
+ 666,
561
+ 195,
562
+ 515,
563
+ 250,
564
+ 634,
565
+ 759,
566
+ 424,
567
+ 607,
568
+ 886,
569
+ 325,
570
+ 412,
571
+ 292,
572
+ 171,
573
+ 709,
574
+ 515,
575
+ 152,
576
+ 699,
577
+ 697,
578
+ 793,
579
+ 805,
580
+ 376,
581
+ 672,
582
+ 562,
583
+ 396,
584
+ 854,
585
+ 743,
586
+ 164,
587
+ 438,
588
+ 332,
589
+ 170,
590
+ 235
591
+ ],
592
+ [
593
+ 223,
594
+ 975,
595
+ 648,
596
+ 447,
597
+ 926,
598
+ 614,
599
+ 392,
600
+ 455,
601
+ 185,
602
+ 158,
603
+ 519,
604
+ 425,
605
+ 560,
606
+ 335,
607
+ 255,
608
+ 175,
609
+ 913,
610
+ 562,
611
+ 820,
612
+ 621,
613
+ 989,
614
+ 48,
615
+ 486,
616
+ 827,
617
+ 119,
618
+ 661,
619
+ 218,
620
+ 838,
621
+ 326,
622
+ 337,
623
+ 272,
624
+ 612,
625
+ 407,
626
+ 660,
627
+ 99,
628
+ 149,
629
+ 853,
630
+ 884,
631
+ 385,
632
+ 418,
633
+ 941,
634
+ 22,
635
+ 514,
636
+ 412,
637
+ 514,
638
+ 772,
639
+ 692,
640
+ 7,
641
+ 48,
642
+ 444,
643
+ 646,
644
+ 660,
645
+ 477,
646
+ 163,
647
+ 737,
648
+ 81,
649
+ 875,
650
+ 34,
651
+ 239,
652
+ 910,
653
+ 169,
654
+ 278,
655
+ 633,
656
+ 196,
657
+ 169,
658
+ 771,
659
+ 667,
660
+ 698,
661
+ 835,
662
+ 558,
663
+ 948
664
+ ],
665
+ [
666
+ 1019,
667
+ 736,
668
+ 891,
669
+ 992,
670
+ 842,
671
+ 1017,
672
+ 763,
673
+ 731,
674
+ 521,
675
+ 275,
676
+ 809,
677
+ 191,
678
+ 879,
679
+ 701,
680
+ 608,
681
+ 815,
682
+ 154,
683
+ 642,
684
+ 282,
685
+ 778,
686
+ 986,
687
+ 135,
688
+ 507,
689
+ 825,
690
+ 36,
691
+ 401,
692
+ 227,
693
+ 111,
694
+ 164,
695
+ 401,
696
+ 866,
697
+ 962,
698
+ 801,
699
+ 336,
700
+ 207,
701
+ 966,
702
+ 719,
703
+ 492,
704
+ 826,
705
+ 974,
706
+ 521,
707
+ 391,
708
+ 136,
709
+ 562,
710
+ 252,
711
+ 521,
712
+ 443,
713
+ 549,
714
+ 162,
715
+ 633,
716
+ 793,
717
+ 343,
718
+ 749,
719
+ 469,
720
+ 860,
721
+ 54,
722
+ 900,
723
+ 496,
724
+ 963,
725
+ 374,
726
+ 990,
727
+ 463,
728
+ 249,
729
+ 676,
730
+ 55,
731
+ 759,
732
+ 959,
733
+ 236,
734
+ 903,
735
+ 233,
736
+ 213
737
+ ]
738
+ ]
739
+ }
voices/maider.json ADDED
@@ -0,0 +1,499 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "maider",
3
+ "display_name": "Maider",
4
+ "reference_text": "Aurrelaria prest dago jokatzeko.",
5
+ "frames": 47,
6
+ "provenance": "HiTZ-Aholab Basque TTS (CC BY 4.0), clip NEU_05850.wav; codes as published by itzune/zortzi-tts-onnx",
7
+ "codes": [
8
+ [
9
+ 2603,
10
+ 396,
11
+ 3322,
12
+ 1226,
13
+ 1001,
14
+ 247,
15
+ 3603,
16
+ 3206,
17
+ 1843,
18
+ 3333,
19
+ 1817,
20
+ 7,
21
+ 2308,
22
+ 1969,
23
+ 1456,
24
+ 3926,
25
+ 2209,
26
+ 2834,
27
+ 2381,
28
+ 3875,
29
+ 3192,
30
+ 1463,
31
+ 3740,
32
+ 1937,
33
+ 638,
34
+ 1288,
35
+ 2719,
36
+ 341,
37
+ 2261,
38
+ 844,
39
+ 1443,
40
+ 1848,
41
+ 1374,
42
+ 489,
43
+ 713,
44
+ 1212,
45
+ 3971,
46
+ 2931,
47
+ 3921,
48
+ 1088,
49
+ 1217,
50
+ 4086,
51
+ 2192,
52
+ 2834,
53
+ 3482,
54
+ 322,
55
+ 2048
56
+ ],
57
+ [
58
+ 288,
59
+ 915,
60
+ 913,
61
+ 821,
62
+ 28,
63
+ 665,
64
+ 80,
65
+ 992,
66
+ 736,
67
+ 441,
68
+ 661,
69
+ 909,
70
+ 232,
71
+ 181,
72
+ 380,
73
+ 175,
74
+ 382,
75
+ 389,
76
+ 836,
77
+ 600,
78
+ 198,
79
+ 631,
80
+ 627,
81
+ 767,
82
+ 679,
83
+ 242,
84
+ 68,
85
+ 736,
86
+ 699,
87
+ 285,
88
+ 392,
89
+ 657,
90
+ 827,
91
+ 339,
92
+ 657,
93
+ 961,
94
+ 329,
95
+ 798,
96
+ 855,
97
+ 847,
98
+ 93,
99
+ 68,
100
+ 7,
101
+ 720,
102
+ 720,
103
+ 21,
104
+ 912
105
+ ],
106
+ [
107
+ 233,
108
+ 710,
109
+ 26,
110
+ 637,
111
+ 299,
112
+ 784,
113
+ 1000,
114
+ 996,
115
+ 238,
116
+ 18,
117
+ 878,
118
+ 469,
119
+ 154,
120
+ 576,
121
+ 819,
122
+ 774,
123
+ 766,
124
+ 970,
125
+ 151,
126
+ 498,
127
+ 134,
128
+ 56,
129
+ 750,
130
+ 180,
131
+ 260,
132
+ 766,
133
+ 266,
134
+ 330,
135
+ 618,
136
+ 719,
137
+ 821,
138
+ 848,
139
+ 939,
140
+ 489,
141
+ 922,
142
+ 854,
143
+ 842,
144
+ 521,
145
+ 802,
146
+ 629,
147
+ 786,
148
+ 571,
149
+ 816,
150
+ 794,
151
+ 161,
152
+ 597,
153
+ 915
154
+ ],
155
+ [
156
+ 605,
157
+ 618,
158
+ 396,
159
+ 832,
160
+ 308,
161
+ 679,
162
+ 779,
163
+ 337,
164
+ 190,
165
+ 449,
166
+ 268,
167
+ 104,
168
+ 664,
169
+ 103,
170
+ 354,
171
+ 3,
172
+ 19,
173
+ 698,
174
+ 792,
175
+ 508,
176
+ 917,
177
+ 518,
178
+ 450,
179
+ 194,
180
+ 604,
181
+ 23,
182
+ 854,
183
+ 976,
184
+ 131,
185
+ 536,
186
+ 253,
187
+ 55,
188
+ 610,
189
+ 571,
190
+ 706,
191
+ 806,
192
+ 257,
193
+ 752,
194
+ 965,
195
+ 462,
196
+ 770,
197
+ 615,
198
+ 798,
199
+ 599,
200
+ 273,
201
+ 520,
202
+ 924
203
+ ],
204
+ [
205
+ 52,
206
+ 616,
207
+ 81,
208
+ 95,
209
+ 183,
210
+ 496,
211
+ 855,
212
+ 720,
213
+ 418,
214
+ 687,
215
+ 298,
216
+ 761,
217
+ 347,
218
+ 250,
219
+ 742,
220
+ 809,
221
+ 782,
222
+ 784,
223
+ 840,
224
+ 129,
225
+ 93,
226
+ 118,
227
+ 521,
228
+ 48,
229
+ 1014,
230
+ 716,
231
+ 467,
232
+ 938,
233
+ 109,
234
+ 63,
235
+ 446,
236
+ 181,
237
+ 285,
238
+ 351,
239
+ 686,
240
+ 888,
241
+ 241,
242
+ 625,
243
+ 118,
244
+ 532,
245
+ 1020,
246
+ 933,
247
+ 846,
248
+ 406,
249
+ 276,
250
+ 589,
251
+ 374
252
+ ],
253
+ [
254
+ 786,
255
+ 714,
256
+ 484,
257
+ 282,
258
+ 603,
259
+ 384,
260
+ 566,
261
+ 238,
262
+ 211,
263
+ 524,
264
+ 1007,
265
+ 0,
266
+ 967,
267
+ 96,
268
+ 1006,
269
+ 23,
270
+ 401,
271
+ 290,
272
+ 492,
273
+ 557,
274
+ 416,
275
+ 449,
276
+ 664,
277
+ 276,
278
+ 259,
279
+ 865,
280
+ 366,
281
+ 618,
282
+ 277,
283
+ 332,
284
+ 815,
285
+ 907,
286
+ 289,
287
+ 239,
288
+ 153,
289
+ 139,
290
+ 681,
291
+ 266,
292
+ 779,
293
+ 450,
294
+ 390,
295
+ 533,
296
+ 752,
297
+ 378,
298
+ 729,
299
+ 645,
300
+ 744
301
+ ],
302
+ [
303
+ 721,
304
+ 245,
305
+ 591,
306
+ 836,
307
+ 103,
308
+ 491,
309
+ 558,
310
+ 323,
311
+ 975,
312
+ 153,
313
+ 407,
314
+ 963,
315
+ 599,
316
+ 956,
317
+ 147,
318
+ 997,
319
+ 800,
320
+ 483,
321
+ 160,
322
+ 624,
323
+ 975,
324
+ 538,
325
+ 922,
326
+ 24,
327
+ 346,
328
+ 813,
329
+ 316,
330
+ 72,
331
+ 872,
332
+ 252,
333
+ 925,
334
+ 303,
335
+ 801,
336
+ 678,
337
+ 590,
338
+ 157,
339
+ 23,
340
+ 994,
341
+ 43,
342
+ 628,
343
+ 939,
344
+ 1,
345
+ 517,
346
+ 277,
347
+ 129,
348
+ 791,
349
+ 431
350
+ ],
351
+ [
352
+ 191,
353
+ 118,
354
+ 982,
355
+ 398,
356
+ 795,
357
+ 820,
358
+ 97,
359
+ 157,
360
+ 420,
361
+ 1001,
362
+ 534,
363
+ 241,
364
+ 141,
365
+ 266,
366
+ 1,
367
+ 939,
368
+ 804,
369
+ 592,
370
+ 111,
371
+ 120,
372
+ 71,
373
+ 234,
374
+ 12,
375
+ 346,
376
+ 269,
377
+ 894,
378
+ 617,
379
+ 778,
380
+ 446,
381
+ 498,
382
+ 949,
383
+ 726,
384
+ 507,
385
+ 791,
386
+ 28,
387
+ 15,
388
+ 169,
389
+ 520,
390
+ 182,
391
+ 567,
392
+ 791,
393
+ 206,
394
+ 159,
395
+ 223,
396
+ 137,
397
+ 409,
398
+ 834
399
+ ],
400
+ [
401
+ 731,
402
+ 406,
403
+ 571,
404
+ 675,
405
+ 1008,
406
+ 681,
407
+ 307,
408
+ 603,
409
+ 123,
410
+ 340,
411
+ 574,
412
+ 405,
413
+ 896,
414
+ 124,
415
+ 415,
416
+ 448,
417
+ 770,
418
+ 655,
419
+ 66,
420
+ 143,
421
+ 506,
422
+ 772,
423
+ 132,
424
+ 926,
425
+ 323,
426
+ 719,
427
+ 693,
428
+ 64,
429
+ 751,
430
+ 723,
431
+ 469,
432
+ 820,
433
+ 911,
434
+ 1012,
435
+ 460,
436
+ 530,
437
+ 196,
438
+ 524,
439
+ 277,
440
+ 590,
441
+ 512,
442
+ 719,
443
+ 206,
444
+ 41,
445
+ 700,
446
+ 464,
447
+ 355
448
+ ],
449
+ [
450
+ 881,
451
+ 154,
452
+ 742,
453
+ 437,
454
+ 111,
455
+ 58,
456
+ 504,
457
+ 400,
458
+ 225,
459
+ 518,
460
+ 787,
461
+ 772,
462
+ 110,
463
+ 535,
464
+ 249,
465
+ 634,
466
+ 575,
467
+ 620,
468
+ 932,
469
+ 247,
470
+ 290,
471
+ 419,
472
+ 367,
473
+ 946,
474
+ 250,
475
+ 45,
476
+ 232,
477
+ 586,
478
+ 54,
479
+ 956,
480
+ 891,
481
+ 570,
482
+ 799,
483
+ 209,
484
+ 165,
485
+ 353,
486
+ 772,
487
+ 544,
488
+ 857,
489
+ 560,
490
+ 196,
491
+ 290,
492
+ 308,
493
+ 597,
494
+ 402,
495
+ 799,
496
+ 180
497
+ ]
498
+ ]
499
+ }