Hugging Face's logo

Muhammadidrees
/
jaybark-small

Safetensors
bark
Model card Files
xet
Community
Muhammadidrees commited on
Commit
ab4e2cc
·
verified ·
1 Parent(s): 4372285

Upload Bark-small model duplicated from suno/bark-small

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +2 -0
  2. .locks/models--suno--bark-small/3c076663001cdfcc4b7805ac7d17aa74fcb84bf0.lock +0 -0
  3. .locks/models--suno--bark-small/4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74.lock +0 -0
  4. .locks/models--suno--bark-small/82654c5189e5c83867e61b1186c2fe364e3c9bd0.lock +0 -0
  5. .locks/models--suno--bark-small/e602b1fa2372be3ada89135f477e300a57b3f21a.lock +0 -0
  6. .locks/models--suno--bark-small/f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9.lock +0 -0
  7. config.json +104 -0
  8. generation_config.json +173 -0
  9. model.safetensors +3 -0
  10. models--suno--bark-small/.no_exist/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/adapter_config.json +0 -0
  11. models--suno--bark-small/.no_exist/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/model.safetensors +0 -0
  12. models--suno--bark-small/.no_exist/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/model.safetensors.index.json +0 -0
  13. models--suno--bark-small/blobs/3c076663001cdfcc4b7805ac7d17aa74fcb84bf0 +340 -0
  14. models--suno--bark-small/blobs/4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74 +3 -0
  15. models--suno--bark-small/blobs/82654c5189e5c83867e61b1186c2fe364e3c9bd0 +1 -0
  16. models--suno--bark-small/blobs/e602b1fa2372be3ada89135f477e300a57b3f21a +172 -0
  17. models--suno--bark-small/blobs/f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9 +3 -0
  18. models--suno--bark-small/refs/main +1 -0
  19. models--suno--bark-small/refs/refs/pr/13 +1 -0
  20. models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/config.json +340 -0
  21. models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/generation_config.json +172 -0
  22. models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/pytorch_model.bin +3 -0
  23. models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/speaker_embeddings_path.json +1 -0
  24. models--suno--bark-small/snapshots/fb35f1a1c6fbf84382a6be730793da74d38f4821/model.safetensors +3 -0
  25. processor_config.json +1311 -0
  26. speaker_embeddings/announcer_coarse_prompt.npy +3 -0
  27. speaker_embeddings/announcer_fine_prompt.npy +3 -0
  28. speaker_embeddings/announcer_semantic_prompt.npy +3 -0
  29. speaker_embeddings/de_speaker_0_coarse_prompt.npy +3 -0
  30. speaker_embeddings/de_speaker_0_fine_prompt.npy +3 -0
  31. speaker_embeddings/de_speaker_0_semantic_prompt.npy +3 -0
  32. speaker_embeddings/de_speaker_1_coarse_prompt.npy +3 -0
  33. speaker_embeddings/de_speaker_1_fine_prompt.npy +3 -0
  34. speaker_embeddings/de_speaker_1_semantic_prompt.npy +3 -0
  35. speaker_embeddings/de_speaker_2_coarse_prompt.npy +3 -0
  36. speaker_embeddings/de_speaker_2_fine_prompt.npy +3 -0
  37. speaker_embeddings/de_speaker_2_semantic_prompt.npy +3 -0
  38. speaker_embeddings/de_speaker_3_coarse_prompt.npy +3 -0
  39. speaker_embeddings/de_speaker_3_fine_prompt.npy +3 -0
  40. speaker_embeddings/de_speaker_3_semantic_prompt.npy +3 -0
  41. speaker_embeddings/de_speaker_4_coarse_prompt.npy +3 -0
  42. speaker_embeddings/de_speaker_4_fine_prompt.npy +3 -0
  43. speaker_embeddings/de_speaker_4_semantic_prompt.npy +3 -0
  44. speaker_embeddings/de_speaker_5_coarse_prompt.npy +3 -0
  45. speaker_embeddings/de_speaker_5_fine_prompt.npy +3 -0
  46. speaker_embeddings/de_speaker_5_semantic_prompt.npy +3 -0
  47. speaker_embeddings/de_speaker_6_coarse_prompt.npy +3 -0
  48. speaker_embeddings/de_speaker_6_fine_prompt.npy +3 -0
  49. speaker_embeddings/de_speaker_6_semantic_prompt.npy +3 -0
  50. speaker_embeddings/de_speaker_7_coarse_prompt.npy +3 -0
.gitattributes CHANGED
@@ -33,3 +33,5 @@ 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
+ models--suno--bark-small/blobs/4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74 filter=lfs diff=lfs merge=lfs -text
37
+ models--suno--bark-small/blobs/f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9 filter=lfs diff=lfs merge=lfs -text
.locks/models--suno--bark-small/3c076663001cdfcc4b7805ac7d17aa74fcb84bf0.lock ADDED
File without changes
.locks/models--suno--bark-small/4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74.lock ADDED
File without changes
.locks/models--suno--bark-small/82654c5189e5c83867e61b1186c2fe364e3c9bd0.lock ADDED
File without changes
.locks/models--suno--bark-small/e602b1fa2372be3ada89135f477e300a57b3f21a.lock ADDED
File without changes
.locks/models--suno--bark-small/f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9.lock ADDED
File without changes
config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "BarkModel"
4
+ ],
5
+ "coarse_acoustics_config": {
6
+ "architectures": [
7
+ "BarkCoarseModel"
8
+ ],
9
+ "bias": false,
10
+ "block_size": 1024,
11
+ "dropout": 0.0,
12
+ "dtype": "float32",
13
+ "hidden_size": 768,
14
+ "initializer_range": 0.02,
15
+ "input_vocab_size": 12096,
16
+ "model_type": "coarse_acoustics",
17
+ "num_heads": 12,
18
+ "num_layers": 12,
19
+ "output_vocab_size": 12096,
20
+ "use_cache": true
21
+ },
22
+ "codec_config": {
23
+ "_name_or_path": "facebook/encodec_24khz",
24
+ "architectures": [
25
+ "EncodecModel"
26
+ ],
27
+ "audio_channels": 1,
28
+ "chunk_length_s": null,
29
+ "codebook_dim": 128,
30
+ "codebook_size": 1024,
31
+ "compress": 2,
32
+ "dilation_growth_rate": 2,
33
+ "dtype": "float32",
34
+ "hidden_size": 128,
35
+ "kernel_size": 7,
36
+ "last_kernel_size": 7,
37
+ "model_type": "encodec",
38
+ "norm_type": "weight_norm",
39
+ "normalize": false,
40
+ "num_filters": 32,
41
+ "num_lstm_layers": 2,
42
+ "num_residual_layers": 1,
43
+ "overlap": null,
44
+ "pad_mode": "reflect",
45
+ "residual_kernel_size": 3,
46
+ "sampling_rate": 24000,
47
+ "target_bandwidths": [
48
+ 1.5,
49
+ 3.0,
50
+ 6.0,
51
+ 12.0,
52
+ 24.0
53
+ ],
54
+ "trim_right_ratio": 1.0,
55
+ "upsampling_ratios": [
56
+ 8,
57
+ 5,
58
+ 4,
59
+ 2
60
+ ],
61
+ "use_causal_conv": true,
62
+ "use_conv_shortcut": true
63
+ },
64
+ "dtype": "float32",
65
+ "fine_acoustics_config": {
66
+ "architectures": [
67
+ "BarkFineModel"
68
+ ],
69
+ "bias": false,
70
+ "block_size": 1024,
71
+ "dropout": 0.0,
72
+ "dtype": "float32",
73
+ "hidden_size": 768,
74
+ "initializer_range": 0.02,
75
+ "input_vocab_size": 1056,
76
+ "model_type": "fine_acoustics",
77
+ "n_codes_given": 1,
78
+ "n_codes_total": 8,
79
+ "num_heads": 12,
80
+ "num_layers": 12,
81
+ "output_vocab_size": 1056,
82
+ "use_cache": true
83
+ },
84
+ "initializer_range": 0.02,
85
+ "model_type": "bark",
86
+ "semantic_config": {
87
+ "architectures": [
88
+ "BarkSemanticModel"
89
+ ],
90
+ "bias": false,
91
+ "block_size": 1024,
92
+ "dropout": 0.0,
93
+ "dtype": "float32",
94
+ "hidden_size": 768,
95
+ "initializer_range": 0.02,
96
+ "input_vocab_size": 129600,
97
+ "model_type": "semantic",
98
+ "num_heads": 12,
99
+ "num_layers": 12,
100
+ "output_vocab_size": 10048,
101
+ "use_cache": true
102
+ },
103
+ "transformers_version": "4.57.0"
104
+ }
generation_config.json ADDED
@@ -0,0 +1,173 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "coarse_acoustics_config": {
3
+ "_from_model_config": false,
4
+ "bad_words_ids": null,
5
+ "begin_suppress_tokens": null,
6
+ "bos_token_id": null,
7
+ "coarse_infer_token": 12050,
8
+ "coarse_rate_hz": 75,
9
+ "coarse_semantic_pad_token": 12048,
10
+ "constraints": null,
11
+ "decoder_start_token_id": null,
12
+ "diversity_penalty": 0.0,
13
+ "do_sample": true,
14
+ "early_stopping": false,
15
+ "encoder_no_repeat_ngram_size": 0,
16
+ "encoder_repetition_penalty": 1.0,
17
+ "eos_token_id": null,
18
+ "epsilon_cutoff": 0.0,
19
+ "eta_cutoff": 0.0,
20
+ "exponential_decay_length_penalty": null,
21
+ "force_words_ids": null,
22
+ "forced_bos_token_id": null,
23
+ "forced_decoder_ids": null,
24
+ "forced_eos_token_id": null,
25
+ "generation_kwargs": {},
26
+ "guidance_scale": null,
27
+ "length_penalty": 1.0,
28
+ "max_coarse_history": 630,
29
+ "max_coarse_input_length": 256,
30
+ "max_length": 20,
31
+ "max_new_tokens": null,
32
+ "max_time": null,
33
+ "min_length": 0,
34
+ "min_new_tokens": null,
35
+ "n_coarse_codebooks": 2,
36
+ "no_repeat_ngram_size": 0,
37
+ "num_beam_groups": 1,
38
+ "num_beams": 1,
39
+ "num_return_sequences": 1,
40
+ "output_attentions": false,
41
+ "output_hidden_states": false,
42
+ "output_scores": false,
43
+ "pad_token_id": null,
44
+ "penalty_alpha": null,
45
+ "remove_invalid_values": false,
46
+ "renormalize_logits": true,
47
+ "repetition_penalty": 1.0,
48
+ "return_dict_in_generate": false,
49
+ "sequence_bias": null,
50
+ "sliding_window_len": 60,
51
+ "suppress_tokens": null,
52
+ "temperature": 0.7,
53
+ "top_k": 50,
54
+ "top_p": 1.0,
55
+ "transformers_version": "4.31.0.dev0",
56
+ "typical_p": 1.0,
57
+ "use_cache": true
58
+ },
59
+ "codebook_size": 1024,
60
+ "fine_acoustics_config": {
61
+ "_from_model_config": false,
62
+ "bad_words_ids": null,
63
+ "begin_suppress_tokens": null,
64
+ "bos_token_id": null,
65
+ "constraints": null,
66
+ "decoder_start_token_id": null,
67
+ "diversity_penalty": 0.0,
68
+ "do_sample": false,
69
+ "early_stopping": false,
70
+ "encoder_no_repeat_ngram_size": 0,
71
+ "encoder_repetition_penalty": 1.0,
72
+ "eos_token_id": null,
73
+ "epsilon_cutoff": 0.0,
74
+ "eta_cutoff": 0.0,
75
+ "exponential_decay_length_penalty": null,
76
+ "force_words_ids": null,
77
+ "forced_bos_token_id": null,
78
+ "forced_decoder_ids": null,
79
+ "forced_eos_token_id": null,
80
+ "generation_kwargs": {},
81
+ "guidance_scale": null,
82
+ "length_penalty": 1.0,
83
+ "max_fine_history_length": 512,
84
+ "max_fine_input_length": 1024,
85
+ "max_length": 20,
86
+ "max_new_tokens": null,
87
+ "max_time": null,
88
+ "min_length": 0,
89
+ "min_new_tokens": null,
90
+ "n_fine_codebooks": 8,
91
+ "no_repeat_ngram_size": 0,
92
+ "num_beam_groups": 1,
93
+ "num_beams": 1,
94
+ "num_return_sequences": 1,
95
+ "output_attentions": false,
96
+ "output_hidden_states": false,
97
+ "output_scores": false,
98
+ "pad_token_id": null,
99
+ "penalty_alpha": null,
100
+ "remove_invalid_values": false,
101
+ "renormalize_logits": false,
102
+ "repetition_penalty": 1.0,
103
+ "return_dict_in_generate": false,
104
+ "sequence_bias": null,
105
+ "suppress_tokens": null,
106
+ "temperature": 0.5,
107
+ "top_k": 50,
108
+ "top_p": 1.0,
109
+ "transformers_version": "4.31.0.dev0",
110
+ "typical_p": 1.0,
111
+ "use_cache": true
112
+ },
113
+ "model_type": "bark",
114
+ "sample_rate": 24000,
115
+ "semantic_config": {
116
+ "_from_model_config": false,
117
+ "bad_words_ids": null,
118
+ "begin_suppress_tokens": null,
119
+ "bos_token_id": null,
120
+ "constraints": null,
121
+ "decoder_start_token_id": null,
122
+ "diversity_penalty": 0.0,
123
+ "do_sample": true,
124
+ "early_stopping": false,
125
+ "encoder_no_repeat_ngram_size": 0,
126
+ "encoder_repetition_penalty": 1.0,
127
+ "eos_token_id": 10000,
128
+ "epsilon_cutoff": 0.0,
129
+ "eta_cutoff": 0.0,
130
+ "exponential_decay_length_penalty": null,
131
+ "force_words_ids": null,
132
+ "forced_bos_token_id": null,
133
+ "forced_decoder_ids": null,
134
+ "forced_eos_token_id": null,
135
+ "generation_kwargs": {},
136
+ "guidance_scale": null,
137
+ "length_penalty": 1.0,
138
+ "max_input_semantic_length": 256,
139
+ "max_length": 20,
140
+ "max_new_tokens": 768,
141
+ "max_time": null,
142
+ "min_length": 0,
143
+ "min_new_tokens": null,
144
+ "no_repeat_ngram_size": 0,
145
+ "num_beam_groups": 1,
146
+ "num_beams": 1,
147
+ "num_return_sequences": 1,
148
+ "output_attentions": false,
149
+ "output_hidden_states": false,
150
+ "output_scores": false,
151
+ "pad_token_id": null,
152
+ "penalty_alpha": null,
153
+ "remove_invalid_values": false,
154
+ "renormalize_logits": true,
155
+ "repetition_penalty": 1.0,
156
+ "return_dict_in_generate": false,
157
+ "semantic_infer_token": 129599,
158
+ "semantic_pad_token": 10000,
159
+ "semantic_rate_hz": 49.9,
160
+ "semantic_vocab_size": 10000,
161
+ "sequence_bias": null,
162
+ "suppress_tokens": null,
163
+ "temperature": 0.7,
164
+ "text_encoding_offset": 10048,
165
+ "text_pad_token": 129595,
166
+ "top_k": 50,
167
+ "top_p": 1.0,
168
+ "transformers_version": "4.31.0.dev0",
169
+ "typical_p": 1.0,
170
+ "use_cache": true
171
+ },
172
+ "transformers_version": "4.57.0"
173
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e4aefb48082e1430a2b5a313d8ddb5e37386f4b0257218afef52f41b521116b
3
+ size 1676552240
models--suno--bark-small/.no_exist/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/adapter_config.json ADDED
File without changes
models--suno--bark-small/.no_exist/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/model.safetensors ADDED
File without changes
models--suno--bark-small/.no_exist/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/model.safetensors.index.json ADDED
File without changes
models--suno--bark-small/blobs/3c076663001cdfcc4b7805ac7d17aa74fcb84bf0 ADDED
@@ -0,0 +1,340 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_commit_hash": null,
3
+ "architectures": [
4
+ "BarkModel"
5
+ ],
6
+ "coarse_acoustics_config": {
7
+ "_name_or_path": "",
8
+ "add_cross_attention": false,
9
+ "architectures": [
10
+ "BarkCoarseModel"
11
+ ],
12
+ "bad_words_ids": null,
13
+ "begin_suppress_tokens": null,
14
+ "bias": false,
15
+ "block_size": 1024,
16
+ "bos_token_id": null,
17
+ "chunk_size_feed_forward": 0,
18
+ "cross_attention_hidden_size": null,
19
+ "decoder_start_token_id": null,
20
+ "diversity_penalty": 0.0,
21
+ "do_sample": false,
22
+ "dropout": 0.0,
23
+ "early_stopping": false,
24
+ "encoder_no_repeat_ngram_size": 0,
25
+ "eos_token_id": null,
26
+ "exponential_decay_length_penalty": null,
27
+ "finetuning_task": null,
28
+ "forced_bos_token_id": null,
29
+ "forced_eos_token_id": null,
30
+ "hidden_size": 768,
31
+ "id2label": {
32
+ "0": "LABEL_0",
33
+ "1": "LABEL_1"
34
+ },
35
+ "initializer_range": 0.02,
36
+ "input_vocab_size": 12096,
37
+ "is_decoder": false,
38
+ "is_encoder_decoder": false,
39
+ "label2id": {
40
+ "LABEL_0": 0,
41
+ "LABEL_1": 1
42
+ },
43
+ "length_penalty": 1.0,
44
+ "max_length": 20,
45
+ "min_length": 0,
46
+ "model_type": "coarse_acoustics",
47
+ "no_repeat_ngram_size": 0,
48
+ "num_beam_groups": 1,
49
+ "num_beams": 1,
50
+ "num_heads": 12,
51
+ "num_layers": 12,
52
+ "num_return_sequences": 1,
53
+ "output_attentions": false,
54
+ "output_hidden_states": false,
55
+ "output_scores": false,
56
+ "output_vocab_size": 12096,
57
+ "pad_token_id": null,
58
+ "prefix": null,
59
+ "problem_type": null,
60
+ "pruned_heads": {},
61
+ "remove_invalid_values": false,
62
+ "repetition_penalty": 1.0,
63
+ "return_dict": true,
64
+ "return_dict_in_generate": false,
65
+ "sep_token_id": null,
66
+ "suppress_tokens": null,
67
+ "task_specific_params": null,
68
+ "temperature": 1.0,
69
+ "tf_legacy_loss": false,
70
+ "tie_encoder_decoder": false,
71
+ "tie_word_embeddings": true,
72
+ "tokenizer_class": null,
73
+ "top_k": 50,
74
+ "top_p": 1.0,
75
+ "torch_dtype": "float32",
76
+ "torchscript": false,
77
+ "transformers_version": "4.31.0.dev0",
78
+ "typical_p": 1.0,
79
+ "use_bfloat16": false,
80
+ "use_cache": true
81
+ },
82
+ "codec_config": {
83
+ "_name_or_path": "facebook/encodec_24khz",
84
+ "add_cross_attention": false,
85
+ "architectures": [
86
+ "EncodecModel"
87
+ ],
88
+ "audio_channels": 1,
89
+ "bad_words_ids": null,
90
+ "begin_suppress_tokens": null,
91
+ "bos_token_id": null,
92
+ "chunk_length_s": null,
93
+ "chunk_size_feed_forward": 0,
94
+ "codebook_dim": 128,
95
+ "codebook_size": 1024,
96
+ "compress": 2,
97
+ "cross_attention_hidden_size": null,
98
+ "decoder_start_token_id": null,
99
+ "dilation_growth_rate": 2,
100
+ "diversity_penalty": 0.0,
101
+ "do_sample": false,
102
+ "early_stopping": false,
103
+ "encoder_no_repeat_ngram_size": 0,
104
+ "eos_token_id": null,
105
+ "exponential_decay_length_penalty": null,
106
+ "finetuning_task": null,
107
+ "forced_bos_token_id": null,
108
+ "forced_eos_token_id": null,
109
+ "hidden_size": 128,
110
+ "id2label": {
111
+ "0": "LABEL_0",
112
+ "1": "LABEL_1"
113
+ },
114
+ "is_decoder": false,
115
+ "is_encoder_decoder": false,
116
+ "kernel_size": 7,
117
+ "label2id": {
118
+ "LABEL_0": 0,
119
+ "LABEL_1": 1
120
+ },
121
+ "last_kernel_size": 7,
122
+ "length_penalty": 1.0,
123
+ "max_length": 20,
124
+ "min_length": 0,
125
+ "model_type": "encodec",
126
+ "no_repeat_ngram_size": 0,
127
+ "norm_type": "weight_norm",
128
+ "normalize": false,
129
+ "num_beam_groups": 1,
130
+ "num_beams": 1,
131
+ "num_filters": 32,
132
+ "num_lstm_layers": 2,
133
+ "num_residual_layers": 1,
134
+ "num_return_sequences": 1,
135
+ "output_attentions": false,
136
+ "output_hidden_states": false,
137
+ "output_scores": false,
138
+ "overlap": null,
139
+ "pad_mode": "reflect",
140
+ "pad_token_id": null,
141
+ "prefix": null,
142
+ "problem_type": null,
143
+ "pruned_heads": {},
144
+ "remove_invalid_values": false,
145
+ "repetition_penalty": 1.0,
146
+ "residual_kernel_size": 3,
147
+ "return_dict": true,
148
+ "return_dict_in_generate": false,
149
+ "sampling_rate": 24000,
150
+ "sep_token_id": null,
151
+ "suppress_tokens": null,
152
+ "target_bandwidths": [
153
+ 1.5,
154
+ 3.0,
155
+ 6.0,
156
+ 12.0,
157
+ 24.0
158
+ ],
159
+ "task_specific_params": null,
160
+ "temperature": 1.0,
161
+ "tf_legacy_loss": false,
162
+ "tie_encoder_decoder": false,
163
+ "tie_word_embeddings": true,
164
+ "tokenizer_class": null,
165
+ "top_k": 50,
166
+ "top_p": 1.0,
167
+ "torch_dtype": "float32",
168
+ "torchscript": false,
169
+ "transformers_version": "4.31.0.dev0",
170
+ "trim_right_ratio": 1.0,
171
+ "typical_p": 1.0,
172
+ "upsampling_ratios": [
173
+ 8,
174
+ 5,
175
+ 4,
176
+ 2
177
+ ],
178
+ "use_bfloat16": false,
179
+ "use_causal_conv": true,
180
+ "use_conv_shortcut": true
181
+ },
182
+ "fine_acoustics_config": {
183
+ "_name_or_path": "",
184
+ "add_cross_attention": false,
185
+ "architectures": [
186
+ "BarkFineModel"
187
+ ],
188
+ "bad_words_ids": null,
189
+ "begin_suppress_tokens": null,
190
+ "bias": false,
191
+ "block_size": 1024,
192
+ "bos_token_id": null,
193
+ "chunk_size_feed_forward": 0,
194
+ "cross_attention_hidden_size": null,
195
+ "decoder_start_token_id": null,
196
+ "diversity_penalty": 0.0,
197
+ "do_sample": false,
198
+ "dropout": 0.0,
199
+ "early_stopping": false,
200
+ "encoder_no_repeat_ngram_size": 0,
201
+ "eos_token_id": null,
202
+ "exponential_decay_length_penalty": null,
203
+ "finetuning_task": null,
204
+ "forced_bos_token_id": null,
205
+ "forced_eos_token_id": null,
206
+ "hidden_size": 768,
207
+ "id2label": {
208
+ "0": "LABEL_0",
209
+ "1": "LABEL_1"
210
+ },
211
+ "initializer_range": 0.02,
212
+ "input_vocab_size": 1056,
213
+ "is_decoder": false,
214
+ "is_encoder_decoder": false,
215
+ "label2id": {
216
+ "LABEL_0": 0,
217
+ "LABEL_1": 1
218
+ },
219
+ "length_penalty": 1.0,
220
+ "max_length": 20,
221
+ "min_length": 0,
222
+ "model_type": "fine_acoustics",
223
+ "n_codes_given": 1,
224
+ "n_codes_total": 8,
225
+ "no_repeat_ngram_size": 0,
226
+ "num_beam_groups": 1,
227
+ "num_beams": 1,
228
+ "num_heads": 12,
229
+ "num_layers": 12,
230
+ "num_return_sequences": 1,
231
+ "output_attentions": false,
232
+ "output_hidden_states": false,
233
+ "output_scores": false,
234
+ "output_vocab_size": 1056,
235
+ "pad_token_id": null,
236
+ "prefix": null,
237
+ "problem_type": null,
238
+ "pruned_heads": {},
239
+ "remove_invalid_values": false,
240
+ "repetition_penalty": 1.0,
241
+ "return_dict": true,
242
+ "return_dict_in_generate": false,
243
+ "sep_token_id": null,
244
+ "suppress_tokens": null,
245
+ "task_specific_params": null,
246
+ "temperature": 1.0,
247
+ "tf_legacy_loss": false,
248
+ "tie_encoder_decoder": false,
249
+ "tie_word_embeddings": true,
250
+ "tokenizer_class": null,
251
+ "top_k": 50,
252
+ "top_p": 1.0,
253
+ "torch_dtype": "float32",
254
+ "torchscript": false,
255
+ "transformers_version": "4.31.0.dev0",
256
+ "typical_p": 1.0,
257
+ "use_bfloat16": false,
258
+ "use_cache": true
259
+ },
260
+ "initializer_range": 0.02,
261
+ "model_type": "bark",
262
+ "semantic_config": {
263
+ "_name_or_path": "",
264
+ "add_cross_attention": false,
265
+ "architectures": [
266
+ "BarkSemanticModel"
267
+ ],
268
+ "bad_words_ids": null,
269
+ "begin_suppress_tokens": null,
270
+ "bias": false,
271
+ "block_size": 1024,
272
+ "bos_token_id": null,
273
+ "chunk_size_feed_forward": 0,
274
+ "cross_attention_hidden_size": null,
275
+ "decoder_start_token_id": null,
276
+ "diversity_penalty": 0.0,
277
+ "do_sample": false,
278
+ "dropout": 0.0,
279
+ "early_stopping": false,
280
+ "encoder_no_repeat_ngram_size": 0,
281
+ "eos_token_id": null,
282
+ "exponential_decay_length_penalty": null,
283
+ "finetuning_task": null,
284
+ "forced_bos_token_id": null,
285
+ "forced_eos_token_id": null,
286
+ "hidden_size": 768,
287
+ "id2label": {
288
+ "0": "LABEL_0",
289
+ "1": "LABEL_1"
290
+ },
291
+ "initializer_range": 0.02,
292
+ "input_vocab_size": 129600,
293
+ "is_decoder": false,
294
+ "is_encoder_decoder": false,
295
+ "label2id": {
296
+ "LABEL_0": 0,
297
+ "LABEL_1": 1
298
+ },
299
+ "length_penalty": 1.0,
300
+ "max_length": 20,
301
+ "min_length": 0,
302
+ "model_type": "semantic",
303
+ "no_repeat_ngram_size": 0,
304
+ "num_beam_groups": 1,
305
+ "num_beams": 1,
306
+ "num_heads": 12,
307
+ "num_layers": 12,
308
+ "num_return_sequences": 1,
309
+ "output_attentions": false,
310
+ "output_hidden_states": false,
311
+ "output_scores": false,
312
+ "output_vocab_size": 10048,
313
+ "pad_token_id": null,
314
+ "prefix": null,
315
+ "problem_type": null,
316
+ "pruned_heads": {},
317
+ "remove_invalid_values": false,
318
+ "repetition_penalty": 1.0,
319
+ "return_dict": true,
320
+ "return_dict_in_generate": false,
321
+ "sep_token_id": null,
322
+ "suppress_tokens": null,
323
+ "task_specific_params": null,
324
+ "temperature": 1.0,
325
+ "tf_legacy_loss": false,
326
+ "tie_encoder_decoder": false,
327
+ "tie_word_embeddings": true,
328
+ "tokenizer_class": null,
329
+ "top_k": 50,
330
+ "top_p": 1.0,
331
+ "torch_dtype": "float32",
332
+ "torchscript": false,
333
+ "transformers_version": "4.31.0.dev0",
334
+ "typical_p": 1.0,
335
+ "use_bfloat16": false,
336
+ "use_cache": true
337
+ },
338
+ "torch_dtype": "float32",
339
+ "transformers_version": null
340
+ }
models--suno--bark-small/blobs/4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74
3
+ size 1676551008
models--suno--bark-small/blobs/82654c5189e5c83867e61b1186c2fe364e3c9bd0 ADDED
@@ -0,0 +1 @@
 
 
1
+ {"repo_or_path": "ylacombe/bark-small", "v2/zh_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_7_fine_prompt.npy"}, "zh_speaker_3": {"semantic_prompt": "speaker_embeddings/zh_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_3_fine_prompt.npy"}, "v2/ja_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_5_fine_prompt.npy"}, "v2/pt_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_1_fine_prompt.npy"}, "fr_speaker_1": {"semantic_prompt": "speaker_embeddings/fr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_1_fine_prompt.npy"}, "ko_speaker_4": {"semantic_prompt": "speaker_embeddings/ko_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_4_fine_prompt.npy"}, "v2/de_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_8_fine_prompt.npy"}, "ja_speaker_0": {"semantic_prompt": "speaker_embeddings/ja_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_0_fine_prompt.npy"}, "v2/tr_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_1_fine_prompt.npy"}, "de_speaker_3": {"semantic_prompt": "speaker_embeddings/de_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_3_fine_prompt.npy"}, "it_speaker_4": {"semantic_prompt": "speaker_embeddings/it_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_4_fine_prompt.npy"}, "ru_speaker_2": {"semantic_prompt": "speaker_embeddings/ru_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_2_fine_prompt.npy"}, "v2/ru_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_1_fine_prompt.npy"}, "v2/pt_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_3_fine_prompt.npy"}, "de_speaker_8": {"semantic_prompt": "speaker_embeddings/de_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_8_fine_prompt.npy"}, "pl_speaker_4": {"semantic_prompt": "speaker_embeddings/pl_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_4_fine_prompt.npy"}, "v2/de_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_1_fine_prompt.npy"}, "de_speaker_5": {"semantic_prompt": "speaker_embeddings/de_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_5_fine_prompt.npy"}, "v2/ko_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_3_fine_prompt.npy"}, "v2/tr_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_8_fine_prompt.npy"}, "pt_speaker_4": {"semantic_prompt": "speaker_embeddings/pt_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_4_fine_prompt.npy"}, "v2/ru_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_9_fine_prompt.npy"}, "hi_speaker_5": {"semantic_prompt": "speaker_embeddings/hi_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_5_fine_prompt.npy"}, "announcer": {"semantic_prompt": "speaker_embeddings/announcer_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/announcer_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/announcer_fine_prompt.npy"}, "it_speaker_3": {"semantic_prompt": "speaker_embeddings/it_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_3_fine_prompt.npy"}, "v2/ja_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_9_fine_prompt.npy"}, "fr_speaker_7": {"semantic_prompt": "speaker_embeddings/fr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_7_fine_prompt.npy"}, "v2/zh_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_5_fine_prompt.npy"}, "v2/it_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_3_fine_prompt.npy"}, "v2/fr_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_5_fine_prompt.npy"}, "v2/ja_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_2_fine_prompt.npy"}, "v2/pl_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_4_fine_prompt.npy"}, "tr_speaker_0": {"semantic_prompt": "speaker_embeddings/tr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_0_fine_prompt.npy"}, "it_speaker_2": {"semantic_prompt": "speaker_embeddings/it_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_2_fine_prompt.npy"}, "v2/en_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_7_fine_prompt.npy"}, "v2/de_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_2_fine_prompt.npy"}, "v2/fr_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_9_fine_prompt.npy"}, "v2/ru_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_4_fine_prompt.npy"}, "ru_speaker_5": {"semantic_prompt": "speaker_embeddings/ru_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_5_fine_prompt.npy"}, "es_speaker_9": {"semantic_prompt": "speaker_embeddings/es_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_9_fine_prompt.npy"}, "it_speaker_1": {"semantic_prompt": "speaker_embeddings/it_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_1_fine_prompt.npy"}, "v2/es_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_7_fine_prompt.npy"}, "ko_speaker_1": {"semantic_prompt": "speaker_embeddings/ko_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_1_fine_prompt.npy"}, "v2/ko_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_2_fine_prompt.npy"}, "v2/pt_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_7_fine_prompt.npy"}, "v2/en_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_3_fine_prompt.npy"}, "ja_speaker_9": {"semantic_prompt": "speaker_embeddings/ja_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_9_fine_prompt.npy"}, "v2/ja_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_6_fine_prompt.npy"}, "pt_speaker_0": {"semantic_prompt": "speaker_embeddings/pt_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_0_fine_prompt.npy"}, "ko_speaker_8": {"semantic_prompt": "speaker_embeddings/ko_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_8_fine_prompt.npy"}, "fr_speaker_5": {"semantic_prompt": "speaker_embeddings/fr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_5_fine_prompt.npy"}, "tr_speaker_7": {"semantic_prompt": "speaker_embeddings/tr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_7_fine_prompt.npy"}, "v2/ja_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_7_fine_prompt.npy"}, "v2/ko_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_1_fine_prompt.npy"}, "hi_speaker_0": {"semantic_prompt": "speaker_embeddings/hi_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_0_fine_prompt.npy"}, "v2/es_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_0_fine_prompt.npy"}, "v2/en_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_0_fine_prompt.npy"}, "es_speaker_6": {"semantic_prompt": "speaker_embeddings/es_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_6_fine_prompt.npy"}, "v2/ko_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_6_fine_prompt.npy"}, "hi_speaker_1": {"semantic_prompt": "speaker_embeddings/hi_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_1_fine_prompt.npy"}, "v2/ja_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_1_fine_prompt.npy"}, "pl_speaker_1": {"semantic_prompt": "speaker_embeddings/pl_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_1_fine_prompt.npy"}, "en_speaker_7": {"semantic_prompt": "speaker_embeddings/en_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_7_fine_prompt.npy"}, "v2/hi_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_3_fine_prompt.npy"}, "ko_speaker_3": {"semantic_prompt": "speaker_embeddings/ko_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_3_fine_prompt.npy"}, "v2/fr_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_0_fine_prompt.npy"}, "de_speaker_4": {"semantic_prompt": "speaker_embeddings/de_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_4_fine_prompt.npy"}, "v2/de_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_7_fine_prompt.npy"}, "v2/pl_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_5_fine_prompt.npy"}, "ko_speaker_6": {"semantic_prompt": "speaker_embeddings/ko_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_6_fine_prompt.npy"}, "en_speaker_2": {"semantic_prompt": "speaker_embeddings/en_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_2_fine_prompt.npy"}, "ru_speaker_3": {"semantic_prompt": "speaker_embeddings/ru_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_3_fine_prompt.npy"}, "v2/es_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_2_fine_prompt.npy"}, "v2/de_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_5_fine_prompt.npy"}, "ja_speaker_1": {"semantic_prompt": "speaker_embeddings/ja_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_1_fine_prompt.npy"}, "v2/it_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_4_fine_prompt.npy"}, "fr_speaker_0": {"semantic_prompt": "speaker_embeddings/fr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_0_fine_prompt.npy"}, "ko_speaker_5": {"semantic_prompt": "speaker_embeddings/ko_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_5_fine_prompt.npy"}, "v2/ko_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_9_fine_prompt.npy"}, "v2/zh_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_6_fine_prompt.npy"}, "v2/fr_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_7_fine_prompt.npy"}, "v2/fr_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_3_fine_prompt.npy"}, "zh_speaker_8": {"semantic_prompt": "speaker_embeddings/zh_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_8_fine_prompt.npy"}, "v2/en_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_2_fine_prompt.npy"}, "ja_speaker_5": {"semantic_prompt": "speaker_embeddings/ja_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_5_fine_prompt.npy"}, "fr_speaker_9": {"semantic_prompt": "speaker_embeddings/fr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_9_fine_prompt.npy"}, "v2/tr_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_6_fine_prompt.npy"}, "pl_speaker_9": {"semantic_prompt": "speaker_embeddings/pl_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_9_fine_prompt.npy"}, "hi_speaker_8": {"semantic_prompt": "speaker_embeddings/hi_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_8_fine_prompt.npy"}, "en_speaker_8": {"semantic_prompt": "speaker_embeddings/en_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_8_fine_prompt.npy"}, "de_speaker_2": {"semantic_prompt": "speaker_embeddings/de_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_2_fine_prompt.npy"}, "v2/pl_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_7_fine_prompt.npy"}, "v2/it_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_8_fine_prompt.npy"}, "pt_speaker_8": {"semantic_prompt": "speaker_embeddings/pt_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_8_fine_prompt.npy"}, "fr_speaker_3": {"semantic_prompt": "speaker_embeddings/fr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_3_fine_prompt.npy"}, "it_speaker_7": {"semantic_prompt": "speaker_embeddings/it_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_7_fine_prompt.npy"}, "v2/ja_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_3_fine_prompt.npy"}, "v2/pt_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_2_fine_prompt.npy"}, "v2/ru_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_5_fine_prompt.npy"}, "ru_speaker_8": {"semantic_prompt": "speaker_embeddings/ru_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_8_fine_prompt.npy"}, "v2/it_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_1_fine_prompt.npy"}, "de_speaker_6": {"semantic_prompt": "speaker_embeddings/de_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_6_fine_prompt.npy"}, "pl_speaker_2": {"semantic_prompt": "speaker_embeddings/pl_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_2_fine_prompt.npy"}, "v2/ja_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_0_fine_prompt.npy"}, "zh_speaker_9": {"semantic_prompt": "speaker_embeddings/zh_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_9_fine_prompt.npy"}, "v2/en_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_4_fine_prompt.npy"}, "v2/hi_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_2_fine_prompt.npy"}, "v2/zh_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_0_fine_prompt.npy"}, "es_speaker_0": {"semantic_prompt": "speaker_embeddings/es_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_0_fine_prompt.npy"}, "pt_speaker_3": {"semantic_prompt": "speaker_embeddings/pt_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_3_fine_prompt.npy"}, "de_speaker_7": {"semantic_prompt": "speaker_embeddings/de_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_7_fine_prompt.npy"}, "v2/ja_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_8_fine_prompt.npy"}, "ru_speaker_6": {"semantic_prompt": "speaker_embeddings/ru_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_6_fine_prompt.npy"}, "tr_speaker_1": {"semantic_prompt": "speaker_embeddings/tr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_1_fine_prompt.npy"}, "v2/en_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_1_fine_prompt.npy"}, "v2/it_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_6_fine_prompt.npy"}, "ja_speaker_6": {"semantic_prompt": "speaker_embeddings/ja_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_6_fine_prompt.npy"}, "pl_speaker_8": {"semantic_prompt": "speaker_embeddings/pl_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_8_fine_prompt.npy"}, "ru_speaker_7": {"semantic_prompt": "speaker_embeddings/ru_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_7_fine_prompt.npy"}, "v2/pl_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_0_fine_prompt.npy"}, "v2/ru_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_3_fine_prompt.npy"}, "v2/ru_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_0_fine_prompt.npy"}, "v2/fr_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_2_fine_prompt.npy"}, "hi_speaker_4": {"semantic_prompt": "speaker_embeddings/hi_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_4_fine_prompt.npy"}, "en_speaker_4": {"semantic_prompt": "speaker_embeddings/en_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_4_fine_prompt.npy"}, "tr_speaker_9": {"semantic_prompt": "speaker_embeddings/tr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_9_fine_prompt.npy"}, "v2/it_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_9_fine_prompt.npy"}, "v2/es_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_3_fine_prompt.npy"}, "v2/ko_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_7_fine_prompt.npy"}, "v2/it_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_7_fine_prompt.npy"}, "zh_speaker_5": {"semantic_prompt": "speaker_embeddings/zh_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_5_fine_prompt.npy"}, "v2/ru_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_8_fine_prompt.npy"}, "v2/en_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_5_fine_prompt.npy"}, "v2/hi_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_7_fine_prompt.npy"}, "v2/de_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_4_fine_prompt.npy"}, "pl_speaker_6": {"semantic_prompt": "speaker_embeddings/pl_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_6_fine_prompt.npy"}, "en_speaker_1": {"semantic_prompt": "speaker_embeddings/en_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_1_fine_prompt.npy"}, "v2/es_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_8_fine_prompt.npy"}, "v2/hi_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_8_fine_prompt.npy"}, "it_speaker_0": {"semantic_prompt": "speaker_embeddings/it_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_0_fine_prompt.npy"}, "v2/ko_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_8_fine_prompt.npy"}, "pt_speaker_5": {"semantic_prompt": "speaker_embeddings/pt_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_5_fine_prompt.npy"}, "it_speaker_5": {"semantic_prompt": "speaker_embeddings/it_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_5_fine_prompt.npy"}, "zh_speaker_2": {"semantic_prompt": "speaker_embeddings/zh_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_2_fine_prompt.npy"}, "v2/fr_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_1_fine_prompt.npy"}, "v2/pl_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_6_fine_prompt.npy"}, "fr_speaker_4": {"semantic_prompt": "speaker_embeddings/fr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_4_fine_prompt.npy"}, "v2/ru_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_6_fine_prompt.npy"}, "v2/tr_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_0_fine_prompt.npy"}, "ko_speaker_7": {"semantic_prompt": "speaker_embeddings/ko_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_7_fine_prompt.npy"}, "tr_speaker_4": {"semantic_prompt": "speaker_embeddings/tr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_3_fine_prompt.npy"}, "fr_speaker_2": {"semantic_prompt": "speaker_embeddings/fr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_2_fine_prompt.npy"}, "v2/en_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_6_fine_prompt.npy"}, "v2/hi_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_6_fine_prompt.npy"}, "pl_speaker_0": {"semantic_prompt": "speaker_embeddings/pl_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_0_fine_prompt.npy"}, "v2/pl_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_3_fine_prompt.npy"}, "pl_speaker_3": {"semantic_prompt": "speaker_embeddings/pl_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_3_fine_prompt.npy"}, "de_speaker_9": {"semantic_prompt": "speaker_embeddings/de_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_9_fine_prompt.npy"}, "es_speaker_1": {"semantic_prompt": "speaker_embeddings/es_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_1_fine_prompt.npy"}, "es_speaker_5": {"semantic_prompt": "speaker_embeddings/es_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_5_fine_prompt.npy"}, "de_speaker_0": {"semantic_prompt": "speaker_embeddings/de_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_0_fine_prompt.npy"}, "v2/pl_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_1_fine_prompt.npy"}, "v2/tr_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_4_fine_prompt.npy"}, "v2/zh_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_9_fine_prompt.npy"}, "v2/hi_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_7_fine_prompt.npy"}, "pt_speaker_6": {"semantic_prompt": "speaker_embeddings/pt_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_6_fine_prompt.npy"}, "v2/hi_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_5_fine_prompt.npy"}, "v2/de_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_9_fine_prompt.npy"}, "hi_speaker_7": {"semantic_prompt": "speaker_embeddings/hi_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_7_fine_prompt.npy"}, "fr_speaker_6": {"semantic_prompt": "speaker_embeddings/fr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_6_fine_prompt.npy"}, "v2/it_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_5_fine_prompt.npy"}, "v2/pt_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_9_fine_prompt.npy"}, "v2/es_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_4_fine_prompt.npy"}, "hi_speaker_2": {"semantic_prompt": "speaker_embeddings/hi_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_2_fine_prompt.npy"}, "v2/it_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_0_fine_prompt.npy"}, "ja_speaker_8": {"semantic_prompt": "speaker_embeddings/ja_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_8_fine_prompt.npy"}, "v2/fr_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_6_fine_prompt.npy"}, "v2/pl_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_2_fine_prompt.npy"}, "pl_speaker_5": {"semantic_prompt": "speaker_embeddings/pl_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_5_fine_prompt.npy"}, "v2/ru_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_7_fine_prompt.npy"}, "ru_speaker_0": {"semantic_prompt": "speaker_embeddings/ru_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_0_fine_prompt.npy"}, "v2/pt_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_8_fine_prompt.npy"}, "zh_speaker_1": {"semantic_prompt": "speaker_embeddings/zh_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_1_fine_prompt.npy"}, "v2/zh_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_3_fine_prompt.npy"}, "v2/pl_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_8_fine_prompt.npy"}, "es_speaker_8": {"semantic_prompt": "speaker_embeddings/es_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_8_fine_prompt.npy"}, "v2/en_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_8_fine_prompt.npy"}, "v2/es_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_9_fine_prompt.npy"}, "fr_speaker_8": {"semantic_prompt": "speaker_embeddings/fr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_8_fine_prompt.npy"}, "ja_speaker_2": {"semantic_prompt": "speaker_embeddings/ja_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_2_fine_prompt.npy"}, "ko_speaker_2": {"semantic_prompt": "speaker_embeddings/ko_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_2_fine_prompt.npy"}, "pl_speaker_7": {"semantic_prompt": "speaker_embeddings/pl_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_7_fine_prompt.npy"}, "v2/pt_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_4_fine_prompt.npy"}, "tr_speaker_2": {"semantic_prompt": "speaker_embeddings/tr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_2_fine_prompt.npy"}, "pt_speaker_9": {"semantic_prompt": "speaker_embeddings/pt_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_9_fine_prompt.npy"}, "zh_speaker_7": {"semantic_prompt": "speaker_embeddings/zh_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_7_fine_prompt.npy"}, "ru_speaker_4": {"semantic_prompt": "speaker_embeddings/ru_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_4_fine_prompt.npy"}, "ja_speaker_3": {"semantic_prompt": "speaker_embeddings/ja_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_3_fine_prompt.npy"}, "en_speaker_3": {"semantic_prompt": "speaker_embeddings/en_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_3_fine_prompt.npy"}, "v2/de_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_6_fine_prompt.npy"}, "v2/ja_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_5_fine_prompt.npy"}, "tr_speaker_3": {"semantic_prompt": "speaker_embeddings/tr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_3_fine_prompt.npy"}, "v2/it_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_2_fine_prompt.npy"}, "v2/ru_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_2_fine_prompt.npy"}, "v2/es_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_1_fine_prompt.npy"}, "v2/ko_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_0_fine_prompt.npy"}, "v2/zh_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_4_fine_prompt.npy"}, "ko_speaker_9": {"semantic_prompt": "speaker_embeddings/ko_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_9_fine_prompt.npy"}, "ru_speaker_1": {"semantic_prompt": "speaker_embeddings/ru_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_1_fine_prompt.npy"}, "v2/tr_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_9_fine_prompt.npy"}, "pt_speaker_2": {"semantic_prompt": "speaker_embeddings/pt_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_2_fine_prompt.npy"}, "v2/de_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_3_fine_prompt.npy"}, "es_speaker_4": {"semantic_prompt": "speaker_embeddings/es_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_4_fine_prompt.npy"}, "v2/ko_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_5_fine_prompt.npy"}, "pt_speaker_1": {"semantic_prompt": "speaker_embeddings/pt_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_1_fine_prompt.npy"}, "zh_speaker_6": {"semantic_prompt": "speaker_embeddings/zh_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_6_fine_prompt.npy"}, "zh_speaker_4": {"semantic_prompt": "speaker_embeddings/zh_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_4_fine_prompt.npy"}, "tr_speaker_6": {"semantic_prompt": "speaker_embeddings/tr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_6_fine_prompt.npy"}, "v2/en_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_9_fine_prompt.npy"}, "v2/de_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_0_fine_prompt.npy"}, "v2/es_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_6_fine_prompt.npy"}, "pt_speaker_7": {"semantic_prompt": "speaker_embeddings/pt_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_7_fine_prompt.npy"}, "hi_speaker_3": {"semantic_prompt": "speaker_embeddings/hi_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_3_fine_prompt.npy"}, "es_speaker_7": {"semantic_prompt": "speaker_embeddings/es_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_7_fine_prompt.npy"}, "v2/zh_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_2_fine_prompt.npy"}, "ja_speaker_7": {"semantic_prompt": "speaker_embeddings/ja_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_7_fine_prompt.npy"}, "de_speaker_1": {"semantic_prompt": "speaker_embeddings/de_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_1_fine_prompt.npy"}, "en_speaker_5": {"semantic_prompt": "speaker_embeddings/en_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_5_fine_prompt.npy"}, "v2/ko_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_4_fine_prompt.npy"}, "v2/pt_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_5_fine_prompt.npy"}, "tr_speaker_8": {"semantic_prompt": "speaker_embeddings/tr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_8_fine_prompt.npy"}, "tr_speaker_5": {"semantic_prompt": "speaker_embeddings/tr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_5_fine_prompt.npy"}, "v2/zh_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_8_fine_prompt.npy"}, "en_speaker_0": {"semantic_prompt": "speaker_embeddings/en_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_0_fine_prompt.npy"}, "hi_speaker_6": {"semantic_prompt": "speaker_embeddings/hi_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_6_fine_prompt.npy"}, "v2/fr_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_8_fine_prompt.npy"}, "es_speaker_3": {"semantic_prompt": "speaker_embeddings/es_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_3_fine_prompt.npy"}, "ko_speaker_0": {"semantic_prompt": "speaker_embeddings/ko_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_0_fine_prompt.npy"}, "v2/pt_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_0_fine_prompt.npy"}, "ja_speaker_4": {"semantic_prompt": "speaker_embeddings/ja_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_2_fine_prompt.npy"}, "es_speaker_2": {"semantic_prompt": "speaker_embeddings/es_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_2_fine_prompt.npy"}, "v2/hi_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_9_fine_prompt.npy"}, "zh_speaker_0": {"semantic_prompt": "speaker_embeddings/zh_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_0_fine_prompt.npy"}, "hi_speaker_9": {"semantic_prompt": "speaker_embeddings/hi_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_9_fine_prompt.npy"}, "v2/pt_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_6_fine_prompt.npy"}, "v2/hi_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_0_fine_prompt.npy"}, "it_speaker_9": {"semantic_prompt": "speaker_embeddings/it_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_9_fine_prompt.npy"}, "it_speaker_6": {"semantic_prompt": "speaker_embeddings/it_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_6_fine_prompt.npy"}, "it_speaker_8": {"semantic_prompt": "speaker_embeddings/it_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_8_fine_prompt.npy"}, "v2/fr_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_4_fine_prompt.npy"}, "v2/pl_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_9_fine_prompt.npy"}, "v2/zh_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_1_fine_prompt.npy"}, "ru_speaker_9": {"semantic_prompt": "speaker_embeddings/ru_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_9_fine_prompt.npy"}, "v2/hi_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_1_fine_prompt.npy"}, "en_speaker_6": {"semantic_prompt": "speaker_embeddings/en_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_6_fine_prompt.npy"}, "en_speaker_9": {"semantic_prompt": "speaker_embeddings/en_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_9_fine_prompt.npy"}, "v2/es_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_5_fine_prompt.npy"}}
models--suno--bark-small/blobs/e602b1fa2372be3ada89135f477e300a57b3f21a ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "coarse_acoustics_config": {
3
+ "_from_model_config": false,
4
+ "bad_words_ids": null,
5
+ "begin_suppress_tokens": null,
6
+ "bos_token_id": null,
7
+ "coarse_infer_token": 12050,
8
+ "coarse_rate_hz": 75,
9
+ "coarse_semantic_pad_token": 12048,
10
+ "constraints": null,
11
+ "decoder_start_token_id": null,
12
+ "diversity_penalty": 0.0,
13
+ "do_sample": true,
14
+ "early_stopping": false,
15
+ "encoder_no_repeat_ngram_size": 0,
16
+ "encoder_repetition_penalty": 1.0,
17
+ "eos_token_id": null,
18
+ "epsilon_cutoff": 0.0,
19
+ "eta_cutoff": 0.0,
20
+ "exponential_decay_length_penalty": null,
21
+ "force_words_ids": null,
22
+ "forced_bos_token_id": null,
23
+ "forced_decoder_ids": null,
24
+ "forced_eos_token_id": null,
25
+ "generation_kwargs": {},
26
+ "guidance_scale": null,
27
+ "length_penalty": 1.0,
28
+ "max_coarse_history": 630,
29
+ "max_coarse_input_length": 256,
30
+ "max_length": 20,
31
+ "max_new_tokens": null,
32
+ "max_time": null,
33
+ "min_length": 0,
34
+ "min_new_tokens": null,
35
+ "n_coarse_codebooks": 2,
36
+ "no_repeat_ngram_size": 0,
37
+ "num_beam_groups": 1,
38
+ "num_beams": 1,
39
+ "num_return_sequences": 1,
40
+ "output_attentions": false,
41
+ "output_hidden_states": false,
42
+ "output_scores": false,
43
+ "pad_token_id": null,
44
+ "penalty_alpha": null,
45
+ "remove_invalid_values": false,
46
+ "renormalize_logits": true,
47
+ "repetition_penalty": 1.0,
48
+ "return_dict_in_generate": false,
49
+ "sequence_bias": null,
50
+ "sliding_window_len": 60,
51
+ "suppress_tokens": null,
52
+ "temperature": 0.7,
53
+ "top_k": 50,
54
+ "top_p": 1.0,
55
+ "transformers_version": "4.31.0.dev0",
56
+ "typical_p": 1.0,
57
+ "use_cache": true
58
+ },
59
+ "codebook_size": 1024,
60
+ "fine_acoustics_config": {
61
+ "_from_model_config": false,
62
+ "bad_words_ids": null,
63
+ "begin_suppress_tokens": null,
64
+ "bos_token_id": null,
65
+ "constraints": null,
66
+ "decoder_start_token_id": null,
67
+ "diversity_penalty": 0.0,
68
+ "do_sample": false,
69
+ "early_stopping": false,
70
+ "encoder_no_repeat_ngram_size": 0,
71
+ "encoder_repetition_penalty": 1.0,
72
+ "eos_token_id": null,
73
+ "epsilon_cutoff": 0.0,
74
+ "eta_cutoff": 0.0,
75
+ "exponential_decay_length_penalty": null,
76
+ "force_words_ids": null,
77
+ "forced_bos_token_id": null,
78
+ "forced_decoder_ids": null,
79
+ "forced_eos_token_id": null,
80
+ "generation_kwargs": {},
81
+ "guidance_scale": null,
82
+ "length_penalty": 1.0,
83
+ "max_fine_history_length": 512,
84
+ "max_fine_input_length": 1024,
85
+ "max_length": 20,
86
+ "max_new_tokens": null,
87
+ "max_time": null,
88
+ "min_length": 0,
89
+ "min_new_tokens": null,
90
+ "n_fine_codebooks": 8,
91
+ "no_repeat_ngram_size": 0,
92
+ "num_beam_groups": 1,
93
+ "num_beams": 1,
94
+ "num_return_sequences": 1,
95
+ "output_attentions": false,
96
+ "output_hidden_states": false,
97
+ "output_scores": false,
98
+ "pad_token_id": null,
99
+ "penalty_alpha": null,
100
+ "remove_invalid_values": false,
101
+ "renormalize_logits": false,
102
+ "repetition_penalty": 1.0,
103
+ "return_dict_in_generate": false,
104
+ "sequence_bias": null,
105
+ "suppress_tokens": null,
106
+ "temperature": 0.5,
107
+ "top_k": 50,
108
+ "top_p": 1.0,
109
+ "transformers_version": "4.31.0.dev0",
110
+ "typical_p": 1.0,
111
+ "use_cache": true
112
+ },
113
+ "model_type": "bark",
114
+ "sample_rate": 24000,
115
+ "semantic_config": {
116
+ "_from_model_config": false,
117
+ "bad_words_ids": null,
118
+ "begin_suppress_tokens": null,
119
+ "bos_token_id": null,
120
+ "constraints": null,
121
+ "decoder_start_token_id": null,
122
+ "diversity_penalty": 0.0,
123
+ "do_sample": true,
124
+ "early_stopping": false,
125
+ "encoder_no_repeat_ngram_size": 0,
126
+ "encoder_repetition_penalty": 1.0,
127
+ "eos_token_id": 10000,
128
+ "epsilon_cutoff": 0.0,
129
+ "eta_cutoff": 0.0,
130
+ "exponential_decay_length_penalty": null,
131
+ "force_words_ids": null,
132
+ "forced_bos_token_id": null,
133
+ "forced_decoder_ids": null,
134
+ "forced_eos_token_id": null,
135
+ "generation_kwargs": {},
136
+ "guidance_scale": null,
137
+ "length_penalty": 1.0,
138
+ "max_input_semantic_length": 256,
139
+ "max_length": 20,
140
+ "max_new_tokens": 768,
141
+ "max_time": null,
142
+ "min_length": 0,
143
+ "min_new_tokens": null,
144
+ "no_repeat_ngram_size": 0,
145
+ "num_beam_groups": 1,
146
+ "num_beams": 1,
147
+ "num_return_sequences": 1,
148
+ "output_attentions": false,
149
+ "output_hidden_states": false,
150
+ "output_scores": false,
151
+ "pad_token_id": null,
152
+ "penalty_alpha": null,
153
+ "remove_invalid_values": false,
154
+ "renormalize_logits": true,
155
+ "repetition_penalty": 1.0,
156
+ "return_dict_in_generate": false,
157
+ "semantic_infer_token": 129599,
158
+ "semantic_pad_token": 10000,
159
+ "semantic_rate_hz": 49.9,
160
+ "semantic_vocab_size": 10000,
161
+ "sequence_bias": null,
162
+ "suppress_tokens": null,
163
+ "temperature": 0.7,
164
+ "text_encoding_offset": 10048,
165
+ "text_pad_token": 129595,
166
+ "top_k": 50,
167
+ "top_p": 1.0,
168
+ "transformers_version": "4.31.0.dev0",
169
+ "typical_p": 1.0,
170
+ "use_cache": true
171
+ }
172
+ }
models--suno--bark-small/blobs/f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9
3
+ size 1676663913
models--suno--bark-small/refs/main ADDED
@@ -0,0 +1 @@
 
 
1
+ 1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd
models--suno--bark-small/refs/refs/pr/13 ADDED
@@ -0,0 +1 @@
 
 
1
+ fb35f1a1c6fbf84382a6be730793da74d38f4821
models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/config.json ADDED
@@ -0,0 +1,340 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_commit_hash": null,
3
+ "architectures": [
4
+ "BarkModel"
5
+ ],
6
+ "coarse_acoustics_config": {
7
+ "_name_or_path": "",
8
+ "add_cross_attention": false,
9
+ "architectures": [
10
+ "BarkCoarseModel"
11
+ ],
12
+ "bad_words_ids": null,
13
+ "begin_suppress_tokens": null,
14
+ "bias": false,
15
+ "block_size": 1024,
16
+ "bos_token_id": null,
17
+ "chunk_size_feed_forward": 0,
18
+ "cross_attention_hidden_size": null,
19
+ "decoder_start_token_id": null,
20
+ "diversity_penalty": 0.0,
21
+ "do_sample": false,
22
+ "dropout": 0.0,
23
+ "early_stopping": false,
24
+ "encoder_no_repeat_ngram_size": 0,
25
+ "eos_token_id": null,
26
+ "exponential_decay_length_penalty": null,
27
+ "finetuning_task": null,
28
+ "forced_bos_token_id": null,
29
+ "forced_eos_token_id": null,
30
+ "hidden_size": 768,
31
+ "id2label": {
32
+ "0": "LABEL_0",
33
+ "1": "LABEL_1"
34
+ },
35
+ "initializer_range": 0.02,
36
+ "input_vocab_size": 12096,
37
+ "is_decoder": false,
38
+ "is_encoder_decoder": false,
39
+ "label2id": {
40
+ "LABEL_0": 0,
41
+ "LABEL_1": 1
42
+ },
43
+ "length_penalty": 1.0,
44
+ "max_length": 20,
45
+ "min_length": 0,
46
+ "model_type": "coarse_acoustics",
47
+ "no_repeat_ngram_size": 0,
48
+ "num_beam_groups": 1,
49
+ "num_beams": 1,
50
+ "num_heads": 12,
51
+ "num_layers": 12,
52
+ "num_return_sequences": 1,
53
+ "output_attentions": false,
54
+ "output_hidden_states": false,
55
+ "output_scores": false,
56
+ "output_vocab_size": 12096,
57
+ "pad_token_id": null,
58
+ "prefix": null,
59
+ "problem_type": null,
60
+ "pruned_heads": {},
61
+ "remove_invalid_values": false,
62
+ "repetition_penalty": 1.0,
63
+ "return_dict": true,
64
+ "return_dict_in_generate": false,
65
+ "sep_token_id": null,
66
+ "suppress_tokens": null,
67
+ "task_specific_params": null,
68
+ "temperature": 1.0,
69
+ "tf_legacy_loss": false,
70
+ "tie_encoder_decoder": false,
71
+ "tie_word_embeddings": true,
72
+ "tokenizer_class": null,
73
+ "top_k": 50,
74
+ "top_p": 1.0,
75
+ "torch_dtype": "float32",
76
+ "torchscript": false,
77
+ "transformers_version": "4.31.0.dev0",
78
+ "typical_p": 1.0,
79
+ "use_bfloat16": false,
80
+ "use_cache": true
81
+ },
82
+ "codec_config": {
83
+ "_name_or_path": "facebook/encodec_24khz",
84
+ "add_cross_attention": false,
85
+ "architectures": [
86
+ "EncodecModel"
87
+ ],
88
+ "audio_channels": 1,
89
+ "bad_words_ids": null,
90
+ "begin_suppress_tokens": null,
91
+ "bos_token_id": null,
92
+ "chunk_length_s": null,
93
+ "chunk_size_feed_forward": 0,
94
+ "codebook_dim": 128,
95
+ "codebook_size": 1024,
96
+ "compress": 2,
97
+ "cross_attention_hidden_size": null,
98
+ "decoder_start_token_id": null,
99
+ "dilation_growth_rate": 2,
100
+ "diversity_penalty": 0.0,
101
+ "do_sample": false,
102
+ "early_stopping": false,
103
+ "encoder_no_repeat_ngram_size": 0,
104
+ "eos_token_id": null,
105
+ "exponential_decay_length_penalty": null,
106
+ "finetuning_task": null,
107
+ "forced_bos_token_id": null,
108
+ "forced_eos_token_id": null,
109
+ "hidden_size": 128,
110
+ "id2label": {
111
+ "0": "LABEL_0",
112
+ "1": "LABEL_1"
113
+ },
114
+ "is_decoder": false,
115
+ "is_encoder_decoder": false,
116
+ "kernel_size": 7,
117
+ "label2id": {
118
+ "LABEL_0": 0,
119
+ "LABEL_1": 1
120
+ },
121
+ "last_kernel_size": 7,
122
+ "length_penalty": 1.0,
123
+ "max_length": 20,
124
+ "min_length": 0,
125
+ "model_type": "encodec",
126
+ "no_repeat_ngram_size": 0,
127
+ "norm_type": "weight_norm",
128
+ "normalize": false,
129
+ "num_beam_groups": 1,
130
+ "num_beams": 1,
131
+ "num_filters": 32,
132
+ "num_lstm_layers": 2,
133
+ "num_residual_layers": 1,
134
+ "num_return_sequences": 1,
135
+ "output_attentions": false,
136
+ "output_hidden_states": false,
137
+ "output_scores": false,
138
+ "overlap": null,
139
+ "pad_mode": "reflect",
140
+ "pad_token_id": null,
141
+ "prefix": null,
142
+ "problem_type": null,
143
+ "pruned_heads": {},
144
+ "remove_invalid_values": false,
145
+ "repetition_penalty": 1.0,
146
+ "residual_kernel_size": 3,
147
+ "return_dict": true,
148
+ "return_dict_in_generate": false,
149
+ "sampling_rate": 24000,
150
+ "sep_token_id": null,
151
+ "suppress_tokens": null,
152
+ "target_bandwidths": [
153
+ 1.5,
154
+ 3.0,
155
+ 6.0,
156
+ 12.0,
157
+ 24.0
158
+ ],
159
+ "task_specific_params": null,
160
+ "temperature": 1.0,
161
+ "tf_legacy_loss": false,
162
+ "tie_encoder_decoder": false,
163
+ "tie_word_embeddings": true,
164
+ "tokenizer_class": null,
165
+ "top_k": 50,
166
+ "top_p": 1.0,
167
+ "torch_dtype": "float32",
168
+ "torchscript": false,
169
+ "transformers_version": "4.31.0.dev0",
170
+ "trim_right_ratio": 1.0,
171
+ "typical_p": 1.0,
172
+ "upsampling_ratios": [
173
+ 8,
174
+ 5,
175
+ 4,
176
+ 2
177
+ ],
178
+ "use_bfloat16": false,
179
+ "use_causal_conv": true,
180
+ "use_conv_shortcut": true
181
+ },
182
+ "fine_acoustics_config": {
183
+ "_name_or_path": "",
184
+ "add_cross_attention": false,
185
+ "architectures": [
186
+ "BarkFineModel"
187
+ ],
188
+ "bad_words_ids": null,
189
+ "begin_suppress_tokens": null,
190
+ "bias": false,
191
+ "block_size": 1024,
192
+ "bos_token_id": null,
193
+ "chunk_size_feed_forward": 0,
194
+ "cross_attention_hidden_size": null,
195
+ "decoder_start_token_id": null,
196
+ "diversity_penalty": 0.0,
197
+ "do_sample": false,
198
+ "dropout": 0.0,
199
+ "early_stopping": false,
200
+ "encoder_no_repeat_ngram_size": 0,
201
+ "eos_token_id": null,
202
+ "exponential_decay_length_penalty": null,
203
+ "finetuning_task": null,
204
+ "forced_bos_token_id": null,
205
+ "forced_eos_token_id": null,
206
+ "hidden_size": 768,
207
+ "id2label": {
208
+ "0": "LABEL_0",
209
+ "1": "LABEL_1"
210
+ },
211
+ "initializer_range": 0.02,
212
+ "input_vocab_size": 1056,
213
+ "is_decoder": false,
214
+ "is_encoder_decoder": false,
215
+ "label2id": {
216
+ "LABEL_0": 0,
217
+ "LABEL_1": 1
218
+ },
219
+ "length_penalty": 1.0,
220
+ "max_length": 20,
221
+ "min_length": 0,
222
+ "model_type": "fine_acoustics",
223
+ "n_codes_given": 1,
224
+ "n_codes_total": 8,
225
+ "no_repeat_ngram_size": 0,
226
+ "num_beam_groups": 1,
227
+ "num_beams": 1,
228
+ "num_heads": 12,
229
+ "num_layers": 12,
230
+ "num_return_sequences": 1,
231
+ "output_attentions": false,
232
+ "output_hidden_states": false,
233
+ "output_scores": false,
234
+ "output_vocab_size": 1056,
235
+ "pad_token_id": null,
236
+ "prefix": null,
237
+ "problem_type": null,
238
+ "pruned_heads": {},
239
+ "remove_invalid_values": false,
240
+ "repetition_penalty": 1.0,
241
+ "return_dict": true,
242
+ "return_dict_in_generate": false,
243
+ "sep_token_id": null,
244
+ "suppress_tokens": null,
245
+ "task_specific_params": null,
246
+ "temperature": 1.0,
247
+ "tf_legacy_loss": false,
248
+ "tie_encoder_decoder": false,
249
+ "tie_word_embeddings": true,
250
+ "tokenizer_class": null,
251
+ "top_k": 50,
252
+ "top_p": 1.0,
253
+ "torch_dtype": "float32",
254
+ "torchscript": false,
255
+ "transformers_version": "4.31.0.dev0",
256
+ "typical_p": 1.0,
257
+ "use_bfloat16": false,
258
+ "use_cache": true
259
+ },
260
+ "initializer_range": 0.02,
261
+ "model_type": "bark",
262
+ "semantic_config": {
263
+ "_name_or_path": "",
264
+ "add_cross_attention": false,
265
+ "architectures": [
266
+ "BarkSemanticModel"
267
+ ],
268
+ "bad_words_ids": null,
269
+ "begin_suppress_tokens": null,
270
+ "bias": false,
271
+ "block_size": 1024,
272
+ "bos_token_id": null,
273
+ "chunk_size_feed_forward": 0,
274
+ "cross_attention_hidden_size": null,
275
+ "decoder_start_token_id": null,
276
+ "diversity_penalty": 0.0,
277
+ "do_sample": false,
278
+ "dropout": 0.0,
279
+ "early_stopping": false,
280
+ "encoder_no_repeat_ngram_size": 0,
281
+ "eos_token_id": null,
282
+ "exponential_decay_length_penalty": null,
283
+ "finetuning_task": null,
284
+ "forced_bos_token_id": null,
285
+ "forced_eos_token_id": null,
286
+ "hidden_size": 768,
287
+ "id2label": {
288
+ "0": "LABEL_0",
289
+ "1": "LABEL_1"
290
+ },
291
+ "initializer_range": 0.02,
292
+ "input_vocab_size": 129600,
293
+ "is_decoder": false,
294
+ "is_encoder_decoder": false,
295
+ "label2id": {
296
+ "LABEL_0": 0,
297
+ "LABEL_1": 1
298
+ },
299
+ "length_penalty": 1.0,
300
+ "max_length": 20,
301
+ "min_length": 0,
302
+ "model_type": "semantic",
303
+ "no_repeat_ngram_size": 0,
304
+ "num_beam_groups": 1,
305
+ "num_beams": 1,
306
+ "num_heads": 12,
307
+ "num_layers": 12,
308
+ "num_return_sequences": 1,
309
+ "output_attentions": false,
310
+ "output_hidden_states": false,
311
+ "output_scores": false,
312
+ "output_vocab_size": 10048,
313
+ "pad_token_id": null,
314
+ "prefix": null,
315
+ "problem_type": null,
316
+ "pruned_heads": {},
317
+ "remove_invalid_values": false,
318
+ "repetition_penalty": 1.0,
319
+ "return_dict": true,
320
+ "return_dict_in_generate": false,
321
+ "sep_token_id": null,
322
+ "suppress_tokens": null,
323
+ "task_specific_params": null,
324
+ "temperature": 1.0,
325
+ "tf_legacy_loss": false,
326
+ "tie_encoder_decoder": false,
327
+ "tie_word_embeddings": true,
328
+ "tokenizer_class": null,
329
+ "top_k": 50,
330
+ "top_p": 1.0,
331
+ "torch_dtype": "float32",
332
+ "torchscript": false,
333
+ "transformers_version": "4.31.0.dev0",
334
+ "typical_p": 1.0,
335
+ "use_bfloat16": false,
336
+ "use_cache": true
337
+ },
338
+ "torch_dtype": "float32",
339
+ "transformers_version": null
340
+ }
models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/generation_config.json ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "coarse_acoustics_config": {
3
+ "_from_model_config": false,
4
+ "bad_words_ids": null,
5
+ "begin_suppress_tokens": null,
6
+ "bos_token_id": null,
7
+ "coarse_infer_token": 12050,
8
+ "coarse_rate_hz": 75,
9
+ "coarse_semantic_pad_token": 12048,
10
+ "constraints": null,
11
+ "decoder_start_token_id": null,
12
+ "diversity_penalty": 0.0,
13
+ "do_sample": true,
14
+ "early_stopping": false,
15
+ "encoder_no_repeat_ngram_size": 0,
16
+ "encoder_repetition_penalty": 1.0,
17
+ "eos_token_id": null,
18
+ "epsilon_cutoff": 0.0,
19
+ "eta_cutoff": 0.0,
20
+ "exponential_decay_length_penalty": null,
21
+ "force_words_ids": null,
22
+ "forced_bos_token_id": null,
23
+ "forced_decoder_ids": null,
24
+ "forced_eos_token_id": null,
25
+ "generation_kwargs": {},
26
+ "guidance_scale": null,
27
+ "length_penalty": 1.0,
28
+ "max_coarse_history": 630,
29
+ "max_coarse_input_length": 256,
30
+ "max_length": 20,
31
+ "max_new_tokens": null,
32
+ "max_time": null,
33
+ "min_length": 0,
34
+ "min_new_tokens": null,
35
+ "n_coarse_codebooks": 2,
36
+ "no_repeat_ngram_size": 0,
37
+ "num_beam_groups": 1,
38
+ "num_beams": 1,
39
+ "num_return_sequences": 1,
40
+ "output_attentions": false,
41
+ "output_hidden_states": false,
42
+ "output_scores": false,
43
+ "pad_token_id": null,
44
+ "penalty_alpha": null,
45
+ "remove_invalid_values": false,
46
+ "renormalize_logits": true,
47
+ "repetition_penalty": 1.0,
48
+ "return_dict_in_generate": false,
49
+ "sequence_bias": null,
50
+ "sliding_window_len": 60,
51
+ "suppress_tokens": null,
52
+ "temperature": 0.7,
53
+ "top_k": 50,
54
+ "top_p": 1.0,
55
+ "transformers_version": "4.31.0.dev0",
56
+ "typical_p": 1.0,
57
+ "use_cache": true
58
+ },
59
+ "codebook_size": 1024,
60
+ "fine_acoustics_config": {
61
+ "_from_model_config": false,
62
+ "bad_words_ids": null,
63
+ "begin_suppress_tokens": null,
64
+ "bos_token_id": null,
65
+ "constraints": null,
66
+ "decoder_start_token_id": null,
67
+ "diversity_penalty": 0.0,
68
+ "do_sample": false,
69
+ "early_stopping": false,
70
+ "encoder_no_repeat_ngram_size": 0,
71
+ "encoder_repetition_penalty": 1.0,
72
+ "eos_token_id": null,
73
+ "epsilon_cutoff": 0.0,
74
+ "eta_cutoff": 0.0,
75
+ "exponential_decay_length_penalty": null,
76
+ "force_words_ids": null,
77
+ "forced_bos_token_id": null,
78
+ "forced_decoder_ids": null,
79
+ "forced_eos_token_id": null,
80
+ "generation_kwargs": {},
81
+ "guidance_scale": null,
82
+ "length_penalty": 1.0,
83
+ "max_fine_history_length": 512,
84
+ "max_fine_input_length": 1024,
85
+ "max_length": 20,
86
+ "max_new_tokens": null,
87
+ "max_time": null,
88
+ "min_length": 0,
89
+ "min_new_tokens": null,
90
+ "n_fine_codebooks": 8,
91
+ "no_repeat_ngram_size": 0,
92
+ "num_beam_groups": 1,
93
+ "num_beams": 1,
94
+ "num_return_sequences": 1,
95
+ "output_attentions": false,
96
+ "output_hidden_states": false,
97
+ "output_scores": false,
98
+ "pad_token_id": null,
99
+ "penalty_alpha": null,
100
+ "remove_invalid_values": false,
101
+ "renormalize_logits": false,
102
+ "repetition_penalty": 1.0,
103
+ "return_dict_in_generate": false,
104
+ "sequence_bias": null,
105
+ "suppress_tokens": null,
106
+ "temperature": 0.5,
107
+ "top_k": 50,
108
+ "top_p": 1.0,
109
+ "transformers_version": "4.31.0.dev0",
110
+ "typical_p": 1.0,
111
+ "use_cache": true
112
+ },
113
+ "model_type": "bark",
114
+ "sample_rate": 24000,
115
+ "semantic_config": {
116
+ "_from_model_config": false,
117
+ "bad_words_ids": null,
118
+ "begin_suppress_tokens": null,
119
+ "bos_token_id": null,
120
+ "constraints": null,
121
+ "decoder_start_token_id": null,
122
+ "diversity_penalty": 0.0,
123
+ "do_sample": true,
124
+ "early_stopping": false,
125
+ "encoder_no_repeat_ngram_size": 0,
126
+ "encoder_repetition_penalty": 1.0,
127
+ "eos_token_id": 10000,
128
+ "epsilon_cutoff": 0.0,
129
+ "eta_cutoff": 0.0,
130
+ "exponential_decay_length_penalty": null,
131
+ "force_words_ids": null,
132
+ "forced_bos_token_id": null,
133
+ "forced_decoder_ids": null,
134
+ "forced_eos_token_id": null,
135
+ "generation_kwargs": {},
136
+ "guidance_scale": null,
137
+ "length_penalty": 1.0,
138
+ "max_input_semantic_length": 256,
139
+ "max_length": 20,
140
+ "max_new_tokens": 768,
141
+ "max_time": null,
142
+ "min_length": 0,
143
+ "min_new_tokens": null,
144
+ "no_repeat_ngram_size": 0,
145
+ "num_beam_groups": 1,
146
+ "num_beams": 1,
147
+ "num_return_sequences": 1,
148
+ "output_attentions": false,
149
+ "output_hidden_states": false,
150
+ "output_scores": false,
151
+ "pad_token_id": null,
152
+ "penalty_alpha": null,
153
+ "remove_invalid_values": false,
154
+ "renormalize_logits": true,
155
+ "repetition_penalty": 1.0,
156
+ "return_dict_in_generate": false,
157
+ "semantic_infer_token": 129599,
158
+ "semantic_pad_token": 10000,
159
+ "semantic_rate_hz": 49.9,
160
+ "semantic_vocab_size": 10000,
161
+ "sequence_bias": null,
162
+ "suppress_tokens": null,
163
+ "temperature": 0.7,
164
+ "text_encoding_offset": 10048,
165
+ "text_pad_token": 129595,
166
+ "top_k": 50,
167
+ "top_p": 1.0,
168
+ "transformers_version": "4.31.0.dev0",
169
+ "typical_p": 1.0,
170
+ "use_cache": true
171
+ }
172
+ }
models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0f7f16b24f65789ce42b3c491aa6a1cdf219f7ef425066fcd194485245e65d9
3
+ size 1676663913
models--suno--bark-small/snapshots/1dbd7a128513b8ae4a4e2130fed57b7ac9da5bcd/speaker_embeddings_path.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"repo_or_path": "ylacombe/bark-small", "v2/zh_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_7_fine_prompt.npy"}, "zh_speaker_3": {"semantic_prompt": "speaker_embeddings/zh_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_3_fine_prompt.npy"}, "v2/ja_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_5_fine_prompt.npy"}, "v2/pt_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_1_fine_prompt.npy"}, "fr_speaker_1": {"semantic_prompt": "speaker_embeddings/fr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_1_fine_prompt.npy"}, "ko_speaker_4": {"semantic_prompt": "speaker_embeddings/ko_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_4_fine_prompt.npy"}, "v2/de_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_8_fine_prompt.npy"}, "ja_speaker_0": {"semantic_prompt": "speaker_embeddings/ja_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_0_fine_prompt.npy"}, "v2/tr_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_1_fine_prompt.npy"}, "de_speaker_3": {"semantic_prompt": "speaker_embeddings/de_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_3_fine_prompt.npy"}, "it_speaker_4": {"semantic_prompt": "speaker_embeddings/it_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_4_fine_prompt.npy"}, "ru_speaker_2": {"semantic_prompt": "speaker_embeddings/ru_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_2_fine_prompt.npy"}, "v2/ru_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_1_fine_prompt.npy"}, "v2/pt_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_3_fine_prompt.npy"}, "de_speaker_8": {"semantic_prompt": "speaker_embeddings/de_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_8_fine_prompt.npy"}, "pl_speaker_4": {"semantic_prompt": "speaker_embeddings/pl_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_4_fine_prompt.npy"}, "v2/de_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_1_fine_prompt.npy"}, "de_speaker_5": {"semantic_prompt": "speaker_embeddings/de_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_5_fine_prompt.npy"}, "v2/ko_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_3_fine_prompt.npy"}, "v2/tr_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_8_fine_prompt.npy"}, "pt_speaker_4": {"semantic_prompt": "speaker_embeddings/pt_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_4_fine_prompt.npy"}, "v2/ru_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_9_fine_prompt.npy"}, "hi_speaker_5": {"semantic_prompt": "speaker_embeddings/hi_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_5_fine_prompt.npy"}, "announcer": {"semantic_prompt": "speaker_embeddings/announcer_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/announcer_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/announcer_fine_prompt.npy"}, "it_speaker_3": {"semantic_prompt": "speaker_embeddings/it_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_3_fine_prompt.npy"}, "v2/ja_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_9_fine_prompt.npy"}, "fr_speaker_7": {"semantic_prompt": "speaker_embeddings/fr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_7_fine_prompt.npy"}, "v2/zh_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_5_fine_prompt.npy"}, "v2/it_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_3_fine_prompt.npy"}, "v2/fr_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_5_fine_prompt.npy"}, "v2/ja_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_2_fine_prompt.npy"}, "v2/pl_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_4_fine_prompt.npy"}, "tr_speaker_0": {"semantic_prompt": "speaker_embeddings/tr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_0_fine_prompt.npy"}, "it_speaker_2": {"semantic_prompt": "speaker_embeddings/it_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_2_fine_prompt.npy"}, "v2/en_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_7_fine_prompt.npy"}, "v2/de_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_2_fine_prompt.npy"}, "v2/fr_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_9_fine_prompt.npy"}, "v2/ru_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_4_fine_prompt.npy"}, "ru_speaker_5": {"semantic_prompt": "speaker_embeddings/ru_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_5_fine_prompt.npy"}, "es_speaker_9": {"semantic_prompt": "speaker_embeddings/es_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_9_fine_prompt.npy"}, "it_speaker_1": {"semantic_prompt": "speaker_embeddings/it_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_1_fine_prompt.npy"}, "v2/es_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_7_fine_prompt.npy"}, "ko_speaker_1": {"semantic_prompt": "speaker_embeddings/ko_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_1_fine_prompt.npy"}, "v2/ko_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_2_fine_prompt.npy"}, "v2/pt_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_7_fine_prompt.npy"}, "v2/en_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_3_fine_prompt.npy"}, "ja_speaker_9": {"semantic_prompt": "speaker_embeddings/ja_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_9_fine_prompt.npy"}, "v2/ja_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_6_fine_prompt.npy"}, "pt_speaker_0": {"semantic_prompt": "speaker_embeddings/pt_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_0_fine_prompt.npy"}, "ko_speaker_8": {"semantic_prompt": "speaker_embeddings/ko_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_8_fine_prompt.npy"}, "fr_speaker_5": {"semantic_prompt": "speaker_embeddings/fr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_5_fine_prompt.npy"}, "tr_speaker_7": {"semantic_prompt": "speaker_embeddings/tr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_7_fine_prompt.npy"}, "v2/ja_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_7_fine_prompt.npy"}, "v2/ko_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_1_fine_prompt.npy"}, "hi_speaker_0": {"semantic_prompt": "speaker_embeddings/hi_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_0_fine_prompt.npy"}, "v2/es_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_0_fine_prompt.npy"}, "v2/en_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_0_fine_prompt.npy"}, "es_speaker_6": {"semantic_prompt": "speaker_embeddings/es_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_6_fine_prompt.npy"}, "v2/ko_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_6_fine_prompt.npy"}, "hi_speaker_1": {"semantic_prompt": "speaker_embeddings/hi_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_1_fine_prompt.npy"}, "v2/ja_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_1_fine_prompt.npy"}, "pl_speaker_1": {"semantic_prompt": "speaker_embeddings/pl_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_1_fine_prompt.npy"}, "en_speaker_7": {"semantic_prompt": "speaker_embeddings/en_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_7_fine_prompt.npy"}, "v2/hi_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_3_fine_prompt.npy"}, "ko_speaker_3": {"semantic_prompt": "speaker_embeddings/ko_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_3_fine_prompt.npy"}, "v2/fr_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_0_fine_prompt.npy"}, "de_speaker_4": {"semantic_prompt": "speaker_embeddings/de_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_4_fine_prompt.npy"}, "v2/de_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_7_fine_prompt.npy"}, "v2/pl_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_5_fine_prompt.npy"}, "ko_speaker_6": {"semantic_prompt": "speaker_embeddings/ko_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_6_fine_prompt.npy"}, "en_speaker_2": {"semantic_prompt": "speaker_embeddings/en_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_2_fine_prompt.npy"}, "ru_speaker_3": {"semantic_prompt": "speaker_embeddings/ru_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_3_fine_prompt.npy"}, "v2/es_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_2_fine_prompt.npy"}, "v2/de_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_5_fine_prompt.npy"}, "ja_speaker_1": {"semantic_prompt": "speaker_embeddings/ja_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_1_fine_prompt.npy"}, "v2/it_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_4_fine_prompt.npy"}, "fr_speaker_0": {"semantic_prompt": "speaker_embeddings/fr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_0_fine_prompt.npy"}, "ko_speaker_5": {"semantic_prompt": "speaker_embeddings/ko_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_5_fine_prompt.npy"}, "v2/ko_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_9_fine_prompt.npy"}, "v2/zh_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_6_fine_prompt.npy"}, "v2/fr_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_7_fine_prompt.npy"}, "v2/fr_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_3_fine_prompt.npy"}, "zh_speaker_8": {"semantic_prompt": "speaker_embeddings/zh_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_8_fine_prompt.npy"}, "v2/en_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_2_fine_prompt.npy"}, "ja_speaker_5": {"semantic_prompt": "speaker_embeddings/ja_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_5_fine_prompt.npy"}, "fr_speaker_9": {"semantic_prompt": "speaker_embeddings/fr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_9_fine_prompt.npy"}, "v2/tr_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_6_fine_prompt.npy"}, "pl_speaker_9": {"semantic_prompt": "speaker_embeddings/pl_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_9_fine_prompt.npy"}, "hi_speaker_8": {"semantic_prompt": "speaker_embeddings/hi_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_8_fine_prompt.npy"}, "en_speaker_8": {"semantic_prompt": "speaker_embeddings/en_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_8_fine_prompt.npy"}, "de_speaker_2": {"semantic_prompt": "speaker_embeddings/de_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_2_fine_prompt.npy"}, "v2/pl_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_7_fine_prompt.npy"}, "v2/it_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_8_fine_prompt.npy"}, "pt_speaker_8": {"semantic_prompt": "speaker_embeddings/pt_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_8_fine_prompt.npy"}, "fr_speaker_3": {"semantic_prompt": "speaker_embeddings/fr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_3_fine_prompt.npy"}, "it_speaker_7": {"semantic_prompt": "speaker_embeddings/it_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_7_fine_prompt.npy"}, "v2/ja_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_3_fine_prompt.npy"}, "v2/pt_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_2_fine_prompt.npy"}, "v2/ru_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_5_fine_prompt.npy"}, "ru_speaker_8": {"semantic_prompt": "speaker_embeddings/ru_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_8_fine_prompt.npy"}, "v2/it_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_1_fine_prompt.npy"}, "de_speaker_6": {"semantic_prompt": "speaker_embeddings/de_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_6_fine_prompt.npy"}, "pl_speaker_2": {"semantic_prompt": "speaker_embeddings/pl_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_2_fine_prompt.npy"}, "v2/ja_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_0_fine_prompt.npy"}, "zh_speaker_9": {"semantic_prompt": "speaker_embeddings/zh_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_9_fine_prompt.npy"}, "v2/en_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_4_fine_prompt.npy"}, "v2/hi_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_2_fine_prompt.npy"}, "v2/zh_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_0_fine_prompt.npy"}, "es_speaker_0": {"semantic_prompt": "speaker_embeddings/es_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_0_fine_prompt.npy"}, "pt_speaker_3": {"semantic_prompt": "speaker_embeddings/pt_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_3_fine_prompt.npy"}, "de_speaker_7": {"semantic_prompt": "speaker_embeddings/de_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_7_fine_prompt.npy"}, "v2/ja_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_8_fine_prompt.npy"}, "ru_speaker_6": {"semantic_prompt": "speaker_embeddings/ru_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_6_fine_prompt.npy"}, "tr_speaker_1": {"semantic_prompt": "speaker_embeddings/tr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_1_fine_prompt.npy"}, "v2/en_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_1_fine_prompt.npy"}, "v2/it_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_6_fine_prompt.npy"}, "ja_speaker_6": {"semantic_prompt": "speaker_embeddings/ja_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_6_fine_prompt.npy"}, "pl_speaker_8": {"semantic_prompt": "speaker_embeddings/pl_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_8_fine_prompt.npy"}, "ru_speaker_7": {"semantic_prompt": "speaker_embeddings/ru_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_7_fine_prompt.npy"}, "v2/pl_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_0_fine_prompt.npy"}, "v2/ru_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_3_fine_prompt.npy"}, "v2/ru_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_0_fine_prompt.npy"}, "v2/fr_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_2_fine_prompt.npy"}, "hi_speaker_4": {"semantic_prompt": "speaker_embeddings/hi_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_4_fine_prompt.npy"}, "en_speaker_4": {"semantic_prompt": "speaker_embeddings/en_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_4_fine_prompt.npy"}, "tr_speaker_9": {"semantic_prompt": "speaker_embeddings/tr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_9_fine_prompt.npy"}, "v2/it_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_9_fine_prompt.npy"}, "v2/es_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_3_fine_prompt.npy"}, "v2/ko_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_7_fine_prompt.npy"}, "v2/it_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_7_fine_prompt.npy"}, "zh_speaker_5": {"semantic_prompt": "speaker_embeddings/zh_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_5_fine_prompt.npy"}, "v2/ru_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_8_fine_prompt.npy"}, "v2/en_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_5_fine_prompt.npy"}, "v2/hi_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_7_fine_prompt.npy"}, "v2/de_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_4_fine_prompt.npy"}, "pl_speaker_6": {"semantic_prompt": "speaker_embeddings/pl_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_6_fine_prompt.npy"}, "en_speaker_1": {"semantic_prompt": "speaker_embeddings/en_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_1_fine_prompt.npy"}, "v2/es_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_8_fine_prompt.npy"}, "v2/hi_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_8_fine_prompt.npy"}, "it_speaker_0": {"semantic_prompt": "speaker_embeddings/it_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_0_fine_prompt.npy"}, "v2/ko_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_8_fine_prompt.npy"}, "pt_speaker_5": {"semantic_prompt": "speaker_embeddings/pt_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_5_fine_prompt.npy"}, "it_speaker_5": {"semantic_prompt": "speaker_embeddings/it_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_5_fine_prompt.npy"}, "zh_speaker_2": {"semantic_prompt": "speaker_embeddings/zh_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_2_fine_prompt.npy"}, "v2/fr_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_1_fine_prompt.npy"}, "v2/pl_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_6_fine_prompt.npy"}, "fr_speaker_4": {"semantic_prompt": "speaker_embeddings/fr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_4_fine_prompt.npy"}, "v2/ru_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_6_fine_prompt.npy"}, "v2/tr_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_0_fine_prompt.npy"}, "ko_speaker_7": {"semantic_prompt": "speaker_embeddings/ko_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_7_fine_prompt.npy"}, "tr_speaker_4": {"semantic_prompt": "speaker_embeddings/tr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_3_fine_prompt.npy"}, "fr_speaker_2": {"semantic_prompt": "speaker_embeddings/fr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_2_fine_prompt.npy"}, "v2/en_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_6_fine_prompt.npy"}, "v2/hi_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_6_fine_prompt.npy"}, "pl_speaker_0": {"semantic_prompt": "speaker_embeddings/pl_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_0_fine_prompt.npy"}, "v2/pl_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_3_fine_prompt.npy"}, "pl_speaker_3": {"semantic_prompt": "speaker_embeddings/pl_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_3_fine_prompt.npy"}, "de_speaker_9": {"semantic_prompt": "speaker_embeddings/de_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_9_fine_prompt.npy"}, "es_speaker_1": {"semantic_prompt": "speaker_embeddings/es_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_1_fine_prompt.npy"}, "es_speaker_5": {"semantic_prompt": "speaker_embeddings/es_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_5_fine_prompt.npy"}, "de_speaker_0": {"semantic_prompt": "speaker_embeddings/de_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_0_fine_prompt.npy"}, "v2/pl_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_1_fine_prompt.npy"}, "v2/tr_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_4_fine_prompt.npy"}, "v2/zh_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_9_fine_prompt.npy"}, "v2/hi_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_7_fine_prompt.npy"}, "pt_speaker_6": {"semantic_prompt": "speaker_embeddings/pt_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_6_fine_prompt.npy"}, "v2/hi_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_5_fine_prompt.npy"}, "v2/de_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_9_fine_prompt.npy"}, "hi_speaker_7": {"semantic_prompt": "speaker_embeddings/hi_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_7_fine_prompt.npy"}, "fr_speaker_6": {"semantic_prompt": "speaker_embeddings/fr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_6_fine_prompt.npy"}, "v2/it_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_5_fine_prompt.npy"}, "v2/pt_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_9_fine_prompt.npy"}, "v2/es_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_4_fine_prompt.npy"}, "hi_speaker_2": {"semantic_prompt": "speaker_embeddings/hi_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_2_fine_prompt.npy"}, "v2/it_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_0_fine_prompt.npy"}, "ja_speaker_8": {"semantic_prompt": "speaker_embeddings/ja_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_8_fine_prompt.npy"}, "v2/fr_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_6_fine_prompt.npy"}, "v2/pl_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_2_fine_prompt.npy"}, "pl_speaker_5": {"semantic_prompt": "speaker_embeddings/pl_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_5_fine_prompt.npy"}, "v2/ru_speaker_7": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_7_fine_prompt.npy"}, "ru_speaker_0": {"semantic_prompt": "speaker_embeddings/ru_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_0_fine_prompt.npy"}, "v2/pt_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_8_fine_prompt.npy"}, "zh_speaker_1": {"semantic_prompt": "speaker_embeddings/zh_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_1_fine_prompt.npy"}, "v2/zh_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_3_fine_prompt.npy"}, "v2/pl_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_8_fine_prompt.npy"}, "es_speaker_8": {"semantic_prompt": "speaker_embeddings/es_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_8_fine_prompt.npy"}, "v2/en_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_8_fine_prompt.npy"}, "v2/es_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_9_fine_prompt.npy"}, "fr_speaker_8": {"semantic_prompt": "speaker_embeddings/fr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/fr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/fr_speaker_8_fine_prompt.npy"}, "ja_speaker_2": {"semantic_prompt": "speaker_embeddings/ja_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_2_fine_prompt.npy"}, "ko_speaker_2": {"semantic_prompt": "speaker_embeddings/ko_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_2_fine_prompt.npy"}, "pl_speaker_7": {"semantic_prompt": "speaker_embeddings/pl_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pl_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pl_speaker_7_fine_prompt.npy"}, "v2/pt_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_4_fine_prompt.npy"}, "tr_speaker_2": {"semantic_prompt": "speaker_embeddings/tr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_2_fine_prompt.npy"}, "pt_speaker_9": {"semantic_prompt": "speaker_embeddings/pt_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_9_fine_prompt.npy"}, "zh_speaker_7": {"semantic_prompt": "speaker_embeddings/zh_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_7_fine_prompt.npy"}, "ru_speaker_4": {"semantic_prompt": "speaker_embeddings/ru_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_4_fine_prompt.npy"}, "ja_speaker_3": {"semantic_prompt": "speaker_embeddings/ja_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_3_fine_prompt.npy"}, "en_speaker_3": {"semantic_prompt": "speaker_embeddings/en_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_3_fine_prompt.npy"}, "v2/de_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_6_fine_prompt.npy"}, "v2/ja_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/ja_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ja_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ja_speaker_5_fine_prompt.npy"}, "tr_speaker_3": {"semantic_prompt": "speaker_embeddings/tr_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_3_fine_prompt.npy"}, "v2/it_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/it_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/it_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/it_speaker_2_fine_prompt.npy"}, "v2/ru_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/ru_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ru_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ru_speaker_2_fine_prompt.npy"}, "v2/es_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_1_fine_prompt.npy"}, "v2/ko_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_0_fine_prompt.npy"}, "v2/zh_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_4_fine_prompt.npy"}, "ko_speaker_9": {"semantic_prompt": "speaker_embeddings/ko_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_9_fine_prompt.npy"}, "ru_speaker_1": {"semantic_prompt": "speaker_embeddings/ru_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_1_fine_prompt.npy"}, "v2/tr_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_9_fine_prompt.npy"}, "pt_speaker_2": {"semantic_prompt": "speaker_embeddings/pt_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_2_fine_prompt.npy"}, "v2/de_speaker_3": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_3_fine_prompt.npy"}, "es_speaker_4": {"semantic_prompt": "speaker_embeddings/es_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_4_fine_prompt.npy"}, "v2/ko_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_5_fine_prompt.npy"}, "pt_speaker_1": {"semantic_prompt": "speaker_embeddings/pt_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_1_fine_prompt.npy"}, "zh_speaker_6": {"semantic_prompt": "speaker_embeddings/zh_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_6_fine_prompt.npy"}, "zh_speaker_4": {"semantic_prompt": "speaker_embeddings/zh_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_4_fine_prompt.npy"}, "tr_speaker_6": {"semantic_prompt": "speaker_embeddings/tr_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_6_fine_prompt.npy"}, "v2/en_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/en_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/en_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/en_speaker_9_fine_prompt.npy"}, "v2/de_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/de_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/de_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/de_speaker_0_fine_prompt.npy"}, "v2/es_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_6_fine_prompt.npy"}, "pt_speaker_7": {"semantic_prompt": "speaker_embeddings/pt_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/pt_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/pt_speaker_7_fine_prompt.npy"}, "hi_speaker_3": {"semantic_prompt": "speaker_embeddings/hi_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_3_fine_prompt.npy"}, "es_speaker_7": {"semantic_prompt": "speaker_embeddings/es_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_7_fine_prompt.npy"}, "v2/zh_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_2_fine_prompt.npy"}, "ja_speaker_7": {"semantic_prompt": "speaker_embeddings/ja_speaker_7_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_7_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_7_fine_prompt.npy"}, "de_speaker_1": {"semantic_prompt": "speaker_embeddings/de_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/de_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/de_speaker_1_fine_prompt.npy"}, "en_speaker_5": {"semantic_prompt": "speaker_embeddings/en_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_5_fine_prompt.npy"}, "v2/ko_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/ko_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/ko_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/ko_speaker_4_fine_prompt.npy"}, "v2/pt_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_5_fine_prompt.npy"}, "tr_speaker_8": {"semantic_prompt": "speaker_embeddings/tr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_8_fine_prompt.npy"}, "tr_speaker_5": {"semantic_prompt": "speaker_embeddings/tr_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/tr_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/tr_speaker_5_fine_prompt.npy"}, "v2/zh_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_8_fine_prompt.npy"}, "en_speaker_0": {"semantic_prompt": "speaker_embeddings/en_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_0_fine_prompt.npy"}, "hi_speaker_6": {"semantic_prompt": "speaker_embeddings/hi_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_6_fine_prompt.npy"}, "v2/fr_speaker_8": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_8_fine_prompt.npy"}, "es_speaker_3": {"semantic_prompt": "speaker_embeddings/es_speaker_3_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_3_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_3_fine_prompt.npy"}, "ko_speaker_0": {"semantic_prompt": "speaker_embeddings/ko_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ko_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ko_speaker_0_fine_prompt.npy"}, "v2/pt_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_0_fine_prompt.npy"}, "ja_speaker_4": {"semantic_prompt": "speaker_embeddings/ja_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ja_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ja_speaker_4_fine_prompt.npy"}, "v2/tr_speaker_2": {"semantic_prompt": "speaker_embeddings/v2/tr_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/tr_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/tr_speaker_2_fine_prompt.npy"}, "es_speaker_2": {"semantic_prompt": "speaker_embeddings/es_speaker_2_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/es_speaker_2_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/es_speaker_2_fine_prompt.npy"}, "v2/hi_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_9_fine_prompt.npy"}, "zh_speaker_0": {"semantic_prompt": "speaker_embeddings/zh_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/zh_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/zh_speaker_0_fine_prompt.npy"}, "hi_speaker_9": {"semantic_prompt": "speaker_embeddings/hi_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/hi_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/hi_speaker_9_fine_prompt.npy"}, "v2/pt_speaker_6": {"semantic_prompt": "speaker_embeddings/v2/pt_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pt_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pt_speaker_6_fine_prompt.npy"}, "v2/hi_speaker_0": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_0_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_0_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_0_fine_prompt.npy"}, "it_speaker_9": {"semantic_prompt": "speaker_embeddings/it_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_9_fine_prompt.npy"}, "it_speaker_6": {"semantic_prompt": "speaker_embeddings/it_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_6_fine_prompt.npy"}, "it_speaker_8": {"semantic_prompt": "speaker_embeddings/it_speaker_8_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/it_speaker_8_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/it_speaker_8_fine_prompt.npy"}, "v2/fr_speaker_4": {"semantic_prompt": "speaker_embeddings/v2/fr_speaker_4_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/fr_speaker_4_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/fr_speaker_4_fine_prompt.npy"}, "v2/pl_speaker_9": {"semantic_prompt": "speaker_embeddings/v2/pl_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/pl_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/pl_speaker_9_fine_prompt.npy"}, "v2/zh_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/zh_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/zh_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/zh_speaker_1_fine_prompt.npy"}, "ru_speaker_9": {"semantic_prompt": "speaker_embeddings/ru_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/ru_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/ru_speaker_9_fine_prompt.npy"}, "v2/hi_speaker_1": {"semantic_prompt": "speaker_embeddings/v2/hi_speaker_1_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/hi_speaker_1_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/hi_speaker_1_fine_prompt.npy"}, "en_speaker_6": {"semantic_prompt": "speaker_embeddings/en_speaker_6_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_6_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_6_fine_prompt.npy"}, "en_speaker_9": {"semantic_prompt": "speaker_embeddings/en_speaker_9_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/en_speaker_9_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/en_speaker_9_fine_prompt.npy"}, "v2/es_speaker_5": {"semantic_prompt": "speaker_embeddings/v2/es_speaker_5_semantic_prompt.npy", "coarse_prompt": "speaker_embeddings/v2/es_speaker_5_coarse_prompt.npy", "fine_prompt": "speaker_embeddings/v2/es_speaker_5_fine_prompt.npy"}}
models--suno--bark-small/snapshots/fb35f1a1c6fbf84382a6be730793da74d38f4821/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a52e32e1487adc051a57c66e18a891c6b7668231469e860e110d12d75e12c74
3
+ size 1676551008
processor_config.json ADDED
@@ -0,0 +1,1311 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "BarkProcessor",
3
+ "speaker_embeddings": {
4
+ "announcer": {
5
+ "coarse_prompt": "speaker_embeddings/announcer_coarse_prompt.npy",
6
+ "fine_prompt": "speaker_embeddings/announcer_fine_prompt.npy",
7
+ "semantic_prompt": "speaker_embeddings/announcer_semantic_prompt.npy"
8
+ },
9
+ "de_speaker_0": {
10
+ "coarse_prompt": "speaker_embeddings/de_speaker_0_coarse_prompt.npy",
11
+ "fine_prompt": "speaker_embeddings/de_speaker_0_fine_prompt.npy",
12
+ "semantic_prompt": "speaker_embeddings/de_speaker_0_semantic_prompt.npy"
13
+ },
14
+ "de_speaker_1": {
15
+ "coarse_prompt": "speaker_embeddings/de_speaker_1_coarse_prompt.npy",
16
+ "fine_prompt": "speaker_embeddings/de_speaker_1_fine_prompt.npy",
17
+ "semantic_prompt": "speaker_embeddings/de_speaker_1_semantic_prompt.npy"
18
+ },
19
+ "de_speaker_2": {
20
+ "coarse_prompt": "speaker_embeddings/de_speaker_2_coarse_prompt.npy",
21
+ "fine_prompt": "speaker_embeddings/de_speaker_2_fine_prompt.npy",
22
+ "semantic_prompt": "speaker_embeddings/de_speaker_2_semantic_prompt.npy"
23
+ },
24
+ "de_speaker_3": {
25
+ "coarse_prompt": "speaker_embeddings/de_speaker_3_coarse_prompt.npy",
26
+ "fine_prompt": "speaker_embeddings/de_speaker_3_fine_prompt.npy",
27
+ "semantic_prompt": "speaker_embeddings/de_speaker_3_semantic_prompt.npy"
28
+ },
29
+ "de_speaker_4": {
30
+ "coarse_prompt": "speaker_embeddings/de_speaker_4_coarse_prompt.npy",
31
+ "fine_prompt": "speaker_embeddings/de_speaker_4_fine_prompt.npy",
32
+ "semantic_prompt": "speaker_embeddings/de_speaker_4_semantic_prompt.npy"
33
+ },
34
+ "de_speaker_5": {
35
+ "coarse_prompt": "speaker_embeddings/de_speaker_5_coarse_prompt.npy",
36
+ "fine_prompt": "speaker_embeddings/de_speaker_5_fine_prompt.npy",
37
+ "semantic_prompt": "speaker_embeddings/de_speaker_5_semantic_prompt.npy"
38
+ },
39
+ "de_speaker_6": {
40
+ "coarse_prompt": "speaker_embeddings/de_speaker_6_coarse_prompt.npy",
41
+ "fine_prompt": "speaker_embeddings/de_speaker_6_fine_prompt.npy",
42
+ "semantic_prompt": "speaker_embeddings/de_speaker_6_semantic_prompt.npy"
43
+ },
44
+ "de_speaker_7": {
45
+ "coarse_prompt": "speaker_embeddings/de_speaker_7_coarse_prompt.npy",
46
+ "fine_prompt": "speaker_embeddings/de_speaker_7_fine_prompt.npy",
47
+ "semantic_prompt": "speaker_embeddings/de_speaker_7_semantic_prompt.npy"
48
+ },
49
+ "de_speaker_8": {
50
+ "coarse_prompt": "speaker_embeddings/de_speaker_8_coarse_prompt.npy",
51
+ "fine_prompt": "speaker_embeddings/de_speaker_8_fine_prompt.npy",
52
+ "semantic_prompt": "speaker_embeddings/de_speaker_8_semantic_prompt.npy"
53
+ },
54
+ "de_speaker_9": {
55
+ "coarse_prompt": "speaker_embeddings/de_speaker_9_coarse_prompt.npy",
56
+ "fine_prompt": "speaker_embeddings/de_speaker_9_fine_prompt.npy",
57
+ "semantic_prompt": "speaker_embeddings/de_speaker_9_semantic_prompt.npy"
58
+ },
59
+ "en_speaker_0": {
60
+ "coarse_prompt": "speaker_embeddings/en_speaker_0_coarse_prompt.npy",
61
+ "fine_prompt": "speaker_embeddings/en_speaker_0_fine_prompt.npy",
62
+ "semantic_prompt": "speaker_embeddings/en_speaker_0_semantic_prompt.npy"
63
+ },
64
+ "en_speaker_1": {
65
+ "coarse_prompt": "speaker_embeddings/en_speaker_1_coarse_prompt.npy",
66
+ "fine_prompt": "speaker_embeddings/en_speaker_1_fine_prompt.npy",
67
+ "semantic_prompt": "speaker_embeddings/en_speaker_1_semantic_prompt.npy"
68
+ },
69
+ "en_speaker_2": {
70
+ "coarse_prompt": "speaker_embeddings/en_speaker_2_coarse_prompt.npy",
71
+ "fine_prompt": "speaker_embeddings/en_speaker_2_fine_prompt.npy",
72
+ "semantic_prompt": "speaker_embeddings/en_speaker_2_semantic_prompt.npy"
73
+ },
74
+ "en_speaker_3": {
75
+ "coarse_prompt": "speaker_embeddings/en_speaker_3_coarse_prompt.npy",
76
+ "fine_prompt": "speaker_embeddings/en_speaker_3_fine_prompt.npy",
77
+ "semantic_prompt": "speaker_embeddings/en_speaker_3_semantic_prompt.npy"
78
+ },
79
+ "en_speaker_4": {
80
+ "coarse_prompt": "speaker_embeddings/en_speaker_4_coarse_prompt.npy",
81
+ "fine_prompt": "speaker_embeddings/en_speaker_4_fine_prompt.npy",
82
+ "semantic_prompt": "speaker_embeddings/en_speaker_4_semantic_prompt.npy"
83
+ },
84
+ "en_speaker_5": {
85
+ "coarse_prompt": "speaker_embeddings/en_speaker_5_coarse_prompt.npy",
86
+ "fine_prompt": "speaker_embeddings/en_speaker_5_fine_prompt.npy",
87
+ "semantic_prompt": "speaker_embeddings/en_speaker_5_semantic_prompt.npy"
88
+ },
89
+ "en_speaker_6": {
90
+ "coarse_prompt": "speaker_embeddings/en_speaker_6_coarse_prompt.npy",
91
+ "fine_prompt": "speaker_embeddings/en_speaker_6_fine_prompt.npy",
92
+ "semantic_prompt": "speaker_embeddings/en_speaker_6_semantic_prompt.npy"
93
+ },
94
+ "en_speaker_7": {
95
+ "coarse_prompt": "speaker_embeddings/en_speaker_7_coarse_prompt.npy",
96
+ "fine_prompt": "speaker_embeddings/en_speaker_7_fine_prompt.npy",
97
+ "semantic_prompt": "speaker_embeddings/en_speaker_7_semantic_prompt.npy"
98
+ },
99
+ "en_speaker_8": {
100
+ "coarse_prompt": "speaker_embeddings/en_speaker_8_coarse_prompt.npy",
101
+ "fine_prompt": "speaker_embeddings/en_speaker_8_fine_prompt.npy",
102
+ "semantic_prompt": "speaker_embeddings/en_speaker_8_semantic_prompt.npy"
103
+ },
104
+ "en_speaker_9": {
105
+ "coarse_prompt": "speaker_embeddings/en_speaker_9_coarse_prompt.npy",
106
+ "fine_prompt": "speaker_embeddings/en_speaker_9_fine_prompt.npy",
107
+ "semantic_prompt": "speaker_embeddings/en_speaker_9_semantic_prompt.npy"
108
+ },
109
+ "es_speaker_0": {
110
+ "coarse_prompt": "speaker_embeddings/es_speaker_0_coarse_prompt.npy",
111
+ "fine_prompt": "speaker_embeddings/es_speaker_0_fine_prompt.npy",
112
+ "semantic_prompt": "speaker_embeddings/es_speaker_0_semantic_prompt.npy"
113
+ },
114
+ "es_speaker_1": {
115
+ "coarse_prompt": "speaker_embeddings/es_speaker_1_coarse_prompt.npy",
116
+ "fine_prompt": "speaker_embeddings/es_speaker_1_fine_prompt.npy",
117
+ "semantic_prompt": "speaker_embeddings/es_speaker_1_semantic_prompt.npy"
118
+ },
119
+ "es_speaker_2": {
120
+ "coarse_prompt": "speaker_embeddings/es_speaker_2_coarse_prompt.npy",
121
+ "fine_prompt": "speaker_embeddings/es_speaker_2_fine_prompt.npy",
122
+ "semantic_prompt": "speaker_embeddings/es_speaker_2_semantic_prompt.npy"
123
+ },
124
+ "es_speaker_3": {
125
+ "coarse_prompt": "speaker_embeddings/es_speaker_3_coarse_prompt.npy",
126
+ "fine_prompt": "speaker_embeddings/es_speaker_3_fine_prompt.npy",
127
+ "semantic_prompt": "speaker_embeddings/es_speaker_3_semantic_prompt.npy"
128
+ },
129
+ "es_speaker_4": {
130
+ "coarse_prompt": "speaker_embeddings/es_speaker_4_coarse_prompt.npy",
131
+ "fine_prompt": "speaker_embeddings/es_speaker_4_fine_prompt.npy",
132
+ "semantic_prompt": "speaker_embeddings/es_speaker_4_semantic_prompt.npy"
133
+ },
134
+ "es_speaker_5": {
135
+ "coarse_prompt": "speaker_embeddings/es_speaker_5_coarse_prompt.npy",
136
+ "fine_prompt": "speaker_embeddings/es_speaker_5_fine_prompt.npy",
137
+ "semantic_prompt": "speaker_embeddings/es_speaker_5_semantic_prompt.npy"
138
+ },
139
+ "es_speaker_6": {
140
+ "coarse_prompt": "speaker_embeddings/es_speaker_6_coarse_prompt.npy",
141
+ "fine_prompt": "speaker_embeddings/es_speaker_6_fine_prompt.npy",
142
+ "semantic_prompt": "speaker_embeddings/es_speaker_6_semantic_prompt.npy"
143
+ },
144
+ "es_speaker_7": {
145
+ "coarse_prompt": "speaker_embeddings/es_speaker_7_coarse_prompt.npy",
146
+ "fine_prompt": "speaker_embeddings/es_speaker_7_fine_prompt.npy",
147
+ "semantic_prompt": "speaker_embeddings/es_speaker_7_semantic_prompt.npy"
148
+ },
149
+ "es_speaker_8": {
150
+ "coarse_prompt": "speaker_embeddings/es_speaker_8_coarse_prompt.npy",
151
+ "fine_prompt": "speaker_embeddings/es_speaker_8_fine_prompt.npy",
152
+ "semantic_prompt": "speaker_embeddings/es_speaker_8_semantic_prompt.npy"
153
+ },
154
+ "es_speaker_9": {
155
+ "coarse_prompt": "speaker_embeddings/es_speaker_9_coarse_prompt.npy",
156
+ "fine_prompt": "speaker_embeddings/es_speaker_9_fine_prompt.npy",
157
+ "semantic_prompt": "speaker_embeddings/es_speaker_9_semantic_prompt.npy"
158
+ },
159
+ "fr_speaker_0": {
160
+ "coarse_prompt": "speaker_embeddings/fr_speaker_0_coarse_prompt.npy",
161
+ "fine_prompt": "speaker_embeddings/fr_speaker_0_fine_prompt.npy",
162
+ "semantic_prompt": "speaker_embeddings/fr_speaker_0_semantic_prompt.npy"
163
+ },
164
+ "fr_speaker_1": {
165
+ "coarse_prompt": "speaker_embeddings/fr_speaker_1_coarse_prompt.npy",
166
+ "fine_prompt": "speaker_embeddings/fr_speaker_1_fine_prompt.npy",
167
+ "semantic_prompt": "speaker_embeddings/fr_speaker_1_semantic_prompt.npy"
168
+ },
169
+ "fr_speaker_2": {
170
+ "coarse_prompt": "speaker_embeddings/fr_speaker_2_coarse_prompt.npy",
171
+ "fine_prompt": "speaker_embeddings/fr_speaker_2_fine_prompt.npy",
172
+ "semantic_prompt": "speaker_embeddings/fr_speaker_2_semantic_prompt.npy"
173
+ },
174
+ "fr_speaker_3": {
175
+ "coarse_prompt": "speaker_embeddings/fr_speaker_3_coarse_prompt.npy",
176
+ "fine_prompt": "speaker_embeddings/fr_speaker_3_fine_prompt.npy",
177
+ "semantic_prompt": "speaker_embeddings/fr_speaker_3_semantic_prompt.npy"
178
+ },
179
+ "fr_speaker_4": {
180
+ "coarse_prompt": "speaker_embeddings/fr_speaker_4_coarse_prompt.npy",
181
+ "fine_prompt": "speaker_embeddings/fr_speaker_4_fine_prompt.npy",
182
+ "semantic_prompt": "speaker_embeddings/fr_speaker_4_semantic_prompt.npy"
183
+ },
184
+ "fr_speaker_5": {
185
+ "coarse_prompt": "speaker_embeddings/fr_speaker_5_coarse_prompt.npy",
186
+ "fine_prompt": "speaker_embeddings/fr_speaker_5_fine_prompt.npy",
187
+ "semantic_prompt": "speaker_embeddings/fr_speaker_5_semantic_prompt.npy"
188
+ },
189
+ "fr_speaker_6": {
190
+ "coarse_prompt": "speaker_embeddings/fr_speaker_6_coarse_prompt.npy",
191
+ "fine_prompt": "speaker_embeddings/fr_speaker_6_fine_prompt.npy",
192
+ "semantic_prompt": "speaker_embeddings/fr_speaker_6_semantic_prompt.npy"
193
+ },
194
+ "fr_speaker_7": {
195
+ "coarse_prompt": "speaker_embeddings/fr_speaker_7_coarse_prompt.npy",
196
+ "fine_prompt": "speaker_embeddings/fr_speaker_7_fine_prompt.npy",
197
+ "semantic_prompt": "speaker_embeddings/fr_speaker_7_semantic_prompt.npy"
198
+ },
199
+ "fr_speaker_8": {
200
+ "coarse_prompt": "speaker_embeddings/fr_speaker_8_coarse_prompt.npy",
201
+ "fine_prompt": "speaker_embeddings/fr_speaker_8_fine_prompt.npy",
202
+ "semantic_prompt": "speaker_embeddings/fr_speaker_8_semantic_prompt.npy"
203
+ },
204
+ "fr_speaker_9": {
205
+ "coarse_prompt": "speaker_embeddings/fr_speaker_9_coarse_prompt.npy",
206
+ "fine_prompt": "speaker_embeddings/fr_speaker_9_fine_prompt.npy",
207
+ "semantic_prompt": "speaker_embeddings/fr_speaker_9_semantic_prompt.npy"
208
+ },
209
+ "hi_speaker_0": {
210
+ "coarse_prompt": "speaker_embeddings/hi_speaker_0_coarse_prompt.npy",
211
+ "fine_prompt": "speaker_embeddings/hi_speaker_0_fine_prompt.npy",
212
+ "semantic_prompt": "speaker_embeddings/hi_speaker_0_semantic_prompt.npy"
213
+ },
214
+ "hi_speaker_1": {
215
+ "coarse_prompt": "speaker_embeddings/hi_speaker_1_coarse_prompt.npy",
216
+ "fine_prompt": "speaker_embeddings/hi_speaker_1_fine_prompt.npy",
217
+ "semantic_prompt": "speaker_embeddings/hi_speaker_1_semantic_prompt.npy"
218
+ },
219
+ "hi_speaker_2": {
220
+ "coarse_prompt": "speaker_embeddings/hi_speaker_2_coarse_prompt.npy",
221
+ "fine_prompt": "speaker_embeddings/hi_speaker_2_fine_prompt.npy",
222
+ "semantic_prompt": "speaker_embeddings/hi_speaker_2_semantic_prompt.npy"
223
+ },
224
+ "hi_speaker_3": {
225
+ "coarse_prompt": "speaker_embeddings/hi_speaker_3_coarse_prompt.npy",
226
+ "fine_prompt": "speaker_embeddings/hi_speaker_3_fine_prompt.npy",
227
+ "semantic_prompt": "speaker_embeddings/hi_speaker_3_semantic_prompt.npy"
228
+ },
229
+ "hi_speaker_4": {
230
+ "coarse_prompt": "speaker_embeddings/hi_speaker_4_coarse_prompt.npy",
231
+ "fine_prompt": "speaker_embeddings/hi_speaker_4_fine_prompt.npy",
232
+ "semantic_prompt": "speaker_embeddings/hi_speaker_4_semantic_prompt.npy"
233
+ },
234
+ "hi_speaker_5": {
235
+ "coarse_prompt": "speaker_embeddings/hi_speaker_5_coarse_prompt.npy",
236
+ "fine_prompt": "speaker_embeddings/hi_speaker_5_fine_prompt.npy",
237
+ "semantic_prompt": "speaker_embeddings/hi_speaker_5_semantic_prompt.npy"
238
+ },
239
+ "hi_speaker_6": {
240
+ "coarse_prompt": "speaker_embeddings/hi_speaker_6_coarse_prompt.npy",
241
+ "fine_prompt": "speaker_embeddings/hi_speaker_6_fine_prompt.npy",
242
+ "semantic_prompt": "speaker_embeddings/hi_speaker_6_semantic_prompt.npy"
243
+ },
244
+ "hi_speaker_7": {
245
+ "coarse_prompt": "speaker_embeddings/hi_speaker_7_coarse_prompt.npy",
246
+ "fine_prompt": "speaker_embeddings/hi_speaker_7_fine_prompt.npy",
247
+ "semantic_prompt": "speaker_embeddings/hi_speaker_7_semantic_prompt.npy"
248
+ },
249
+ "hi_speaker_8": {
250
+ "coarse_prompt": "speaker_embeddings/hi_speaker_8_coarse_prompt.npy",
251
+ "fine_prompt": "speaker_embeddings/hi_speaker_8_fine_prompt.npy",
252
+ "semantic_prompt": "speaker_embeddings/hi_speaker_8_semantic_prompt.npy"
253
+ },
254
+ "hi_speaker_9": {
255
+ "coarse_prompt": "speaker_embeddings/hi_speaker_9_coarse_prompt.npy",
256
+ "fine_prompt": "speaker_embeddings/hi_speaker_9_fine_prompt.npy",
257
+ "semantic_prompt": "speaker_embeddings/hi_speaker_9_semantic_prompt.npy"
258
+ },
259
+ "it_speaker_0": {
260
+ "coarse_prompt": "speaker_embeddings/it_speaker_0_coarse_prompt.npy",
261
+ "fine_prompt": "speaker_embeddings/it_speaker_0_fine_prompt.npy",
262
+ "semantic_prompt": "speaker_embeddings/it_speaker_0_semantic_prompt.npy"
263
+ },
264
+ "it_speaker_1": {
265
+ "coarse_prompt": "speaker_embeddings/it_speaker_1_coarse_prompt.npy",
266
+ "fine_prompt": "speaker_embeddings/it_speaker_1_fine_prompt.npy",
267
+ "semantic_prompt": "speaker_embeddings/it_speaker_1_semantic_prompt.npy"
268
+ },
269
+ "it_speaker_2": {
270
+ "coarse_prompt": "speaker_embeddings/it_speaker_2_coarse_prompt.npy",
271
+ "fine_prompt": "speaker_embeddings/it_speaker_2_fine_prompt.npy",
272
+ "semantic_prompt": "speaker_embeddings/it_speaker_2_semantic_prompt.npy"
273
+ },
274
+ "it_speaker_3": {
275
+ "coarse_prompt": "speaker_embeddings/it_speaker_3_coarse_prompt.npy",
276
+ "fine_prompt": "speaker_embeddings/it_speaker_3_fine_prompt.npy",
277
+ "semantic_prompt": "speaker_embeddings/it_speaker_3_semantic_prompt.npy"
278
+ },
279
+ "it_speaker_4": {
280
+ "coarse_prompt": "speaker_embeddings/it_speaker_4_coarse_prompt.npy",
281
+ "fine_prompt": "speaker_embeddings/it_speaker_4_fine_prompt.npy",
282
+ "semantic_prompt": "speaker_embeddings/it_speaker_4_semantic_prompt.npy"
283
+ },
284
+ "it_speaker_5": {
285
+ "coarse_prompt": "speaker_embeddings/it_speaker_5_coarse_prompt.npy",
286
+ "fine_prompt": "speaker_embeddings/it_speaker_5_fine_prompt.npy",
287
+ "semantic_prompt": "speaker_embeddings/it_speaker_5_semantic_prompt.npy"
288
+ },
289
+ "it_speaker_6": {
290
+ "coarse_prompt": "speaker_embeddings/it_speaker_6_coarse_prompt.npy",
291
+ "fine_prompt": "speaker_embeddings/it_speaker_6_fine_prompt.npy",
292
+ "semantic_prompt": "speaker_embeddings/it_speaker_6_semantic_prompt.npy"
293
+ },
294
+ "it_speaker_7": {
295
+ "coarse_prompt": "speaker_embeddings/it_speaker_7_coarse_prompt.npy",
296
+ "fine_prompt": "speaker_embeddings/it_speaker_7_fine_prompt.npy",
297
+ "semantic_prompt": "speaker_embeddings/it_speaker_7_semantic_prompt.npy"
298
+ },
299
+ "it_speaker_8": {
300
+ "coarse_prompt": "speaker_embeddings/it_speaker_8_coarse_prompt.npy",
301
+ "fine_prompt": "speaker_embeddings/it_speaker_8_fine_prompt.npy",
302
+ "semantic_prompt": "speaker_embeddings/it_speaker_8_semantic_prompt.npy"
303
+ },
304
+ "it_speaker_9": {
305
+ "coarse_prompt": "speaker_embeddings/it_speaker_9_coarse_prompt.npy",
306
+ "fine_prompt": "speaker_embeddings/it_speaker_9_fine_prompt.npy",
307
+ "semantic_prompt": "speaker_embeddings/it_speaker_9_semantic_prompt.npy"
308
+ },
309
+ "ja_speaker_0": {
310
+ "coarse_prompt": "speaker_embeddings/ja_speaker_0_coarse_prompt.npy",
311
+ "fine_prompt": "speaker_embeddings/ja_speaker_0_fine_prompt.npy",
312
+ "semantic_prompt": "speaker_embeddings/ja_speaker_0_semantic_prompt.npy"
313
+ },
314
+ "ja_speaker_1": {
315
+ "coarse_prompt": "speaker_embeddings/ja_speaker_1_coarse_prompt.npy",
316
+ "fine_prompt": "speaker_embeddings/ja_speaker_1_fine_prompt.npy",
317
+ "semantic_prompt": "speaker_embeddings/ja_speaker_1_semantic_prompt.npy"
318
+ },
319
+ "ja_speaker_2": {
320
+ "coarse_prompt": "speaker_embeddings/ja_speaker_2_coarse_prompt.npy",
321
+ "fine_prompt": "speaker_embeddings/ja_speaker_2_fine_prompt.npy",
322
+ "semantic_prompt": "speaker_embeddings/ja_speaker_2_semantic_prompt.npy"
323
+ },
324
+ "ja_speaker_3": {
325
+ "coarse_prompt": "speaker_embeddings/ja_speaker_3_coarse_prompt.npy",
326
+ "fine_prompt": "speaker_embeddings/ja_speaker_3_fine_prompt.npy",
327
+ "semantic_prompt": "speaker_embeddings/ja_speaker_3_semantic_prompt.npy"
328
+ },
329
+ "ja_speaker_4": {
330
+ "coarse_prompt": "speaker_embeddings/ja_speaker_4_coarse_prompt.npy",
331
+ "fine_prompt": "speaker_embeddings/ja_speaker_4_fine_prompt.npy",
332
+ "semantic_prompt": "speaker_embeddings/ja_speaker_4_semantic_prompt.npy"
333
+ },
334
+ "ja_speaker_5": {
335
+ "coarse_prompt": "speaker_embeddings/ja_speaker_5_coarse_prompt.npy",
336
+ "fine_prompt": "speaker_embeddings/ja_speaker_5_fine_prompt.npy",
337
+ "semantic_prompt": "speaker_embeddings/ja_speaker_5_semantic_prompt.npy"
338
+ },
339
+ "ja_speaker_6": {
340
+ "coarse_prompt": "speaker_embeddings/ja_speaker_6_coarse_prompt.npy",
341
+ "fine_prompt": "speaker_embeddings/ja_speaker_6_fine_prompt.npy",
342
+ "semantic_prompt": "speaker_embeddings/ja_speaker_6_semantic_prompt.npy"
343
+ },
344
+ "ja_speaker_7": {
345
+ "coarse_prompt": "speaker_embeddings/ja_speaker_7_coarse_prompt.npy",
346
+ "fine_prompt": "speaker_embeddings/ja_speaker_7_fine_prompt.npy",
347
+ "semantic_prompt": "speaker_embeddings/ja_speaker_7_semantic_prompt.npy"
348
+ },
349
+ "ja_speaker_8": {
350
+ "coarse_prompt": "speaker_embeddings/ja_speaker_8_coarse_prompt.npy",
351
+ "fine_prompt": "speaker_embeddings/ja_speaker_8_fine_prompt.npy",
352
+ "semantic_prompt": "speaker_embeddings/ja_speaker_8_semantic_prompt.npy"
353
+ },
354
+ "ja_speaker_9": {
355
+ "coarse_prompt": "speaker_embeddings/ja_speaker_9_coarse_prompt.npy",
356
+ "fine_prompt": "speaker_embeddings/ja_speaker_9_fine_prompt.npy",
357
+ "semantic_prompt": "speaker_embeddings/ja_speaker_9_semantic_prompt.npy"
358
+ },
359
+ "ko_speaker_0": {
360
+ "coarse_prompt": "speaker_embeddings/ko_speaker_0_coarse_prompt.npy",
361
+ "fine_prompt": "speaker_embeddings/ko_speaker_0_fine_prompt.npy",
362
+ "semantic_prompt": "speaker_embeddings/ko_speaker_0_semantic_prompt.npy"
363
+ },
364
+ "ko_speaker_1": {
365
+ "coarse_prompt": "speaker_embeddings/ko_speaker_1_coarse_prompt.npy",
366
+ "fine_prompt": "speaker_embeddings/ko_speaker_1_fine_prompt.npy",
367
+ "semantic_prompt": "speaker_embeddings/ko_speaker_1_semantic_prompt.npy"
368
+ },
369
+ "ko_speaker_2": {
370
+ "coarse_prompt": "speaker_embeddings/ko_speaker_2_coarse_prompt.npy",
371
+ "fine_prompt": "speaker_embeddings/ko_speaker_2_fine_prompt.npy",
372
+ "semantic_prompt": "speaker_embeddings/ko_speaker_2_semantic_prompt.npy"
373
+ },
374
+ "ko_speaker_3": {
375
+ "coarse_prompt": "speaker_embeddings/ko_speaker_3_coarse_prompt.npy",
376
+ "fine_prompt": "speaker_embeddings/ko_speaker_3_fine_prompt.npy",
377
+ "semantic_prompt": "speaker_embeddings/ko_speaker_3_semantic_prompt.npy"
378
+ },
379
+ "ko_speaker_4": {
380
+ "coarse_prompt": "speaker_embeddings/ko_speaker_4_coarse_prompt.npy",
381
+ "fine_prompt": "speaker_embeddings/ko_speaker_4_fine_prompt.npy",
382
+ "semantic_prompt": "speaker_embeddings/ko_speaker_4_semantic_prompt.npy"
383
+ },
384
+ "ko_speaker_5": {
385
+ "coarse_prompt": "speaker_embeddings/ko_speaker_5_coarse_prompt.npy",
386
+ "fine_prompt": "speaker_embeddings/ko_speaker_5_fine_prompt.npy",
387
+ "semantic_prompt": "speaker_embeddings/ko_speaker_5_semantic_prompt.npy"
388
+ },
389
+ "ko_speaker_6": {
390
+ "coarse_prompt": "speaker_embeddings/ko_speaker_6_coarse_prompt.npy",
391
+ "fine_prompt": "speaker_embeddings/ko_speaker_6_fine_prompt.npy",
392
+ "semantic_prompt": "speaker_embeddings/ko_speaker_6_semantic_prompt.npy"
393
+ },
394
+ "ko_speaker_7": {
395
+ "coarse_prompt": "speaker_embeddings/ko_speaker_7_coarse_prompt.npy",
396
+ "fine_prompt": "speaker_embeddings/ko_speaker_7_fine_prompt.npy",
397
+ "semantic_prompt": "speaker_embeddings/ko_speaker_7_semantic_prompt.npy"
398
+ },
399
+ "ko_speaker_8": {
400
+ "coarse_prompt": "speaker_embeddings/ko_speaker_8_coarse_prompt.npy",
401
+ "fine_prompt": "speaker_embeddings/ko_speaker_8_fine_prompt.npy",
402
+ "semantic_prompt": "speaker_embeddings/ko_speaker_8_semantic_prompt.npy"
403
+ },
404
+ "ko_speaker_9": {
405
+ "coarse_prompt": "speaker_embeddings/ko_speaker_9_coarse_prompt.npy",
406
+ "fine_prompt": "speaker_embeddings/ko_speaker_9_fine_prompt.npy",
407
+ "semantic_prompt": "speaker_embeddings/ko_speaker_9_semantic_prompt.npy"
408
+ },
409
+ "pl_speaker_0": {
410
+ "coarse_prompt": "speaker_embeddings/pl_speaker_0_coarse_prompt.npy",
411
+ "fine_prompt": "speaker_embeddings/pl_speaker_0_fine_prompt.npy",
412
+ "semantic_prompt": "speaker_embeddings/pl_speaker_0_semantic_prompt.npy"
413
+ },
414
+ "pl_speaker_1": {
415
+ "coarse_prompt": "speaker_embeddings/pl_speaker_1_coarse_prompt.npy",
416
+ "fine_prompt": "speaker_embeddings/pl_speaker_1_fine_prompt.npy",
417
+ "semantic_prompt": "speaker_embeddings/pl_speaker_1_semantic_prompt.npy"
418
+ },
419
+ "pl_speaker_2": {
420
+ "coarse_prompt": "speaker_embeddings/pl_speaker_2_coarse_prompt.npy",
421
+ "fine_prompt": "speaker_embeddings/pl_speaker_2_fine_prompt.npy",
422
+ "semantic_prompt": "speaker_embeddings/pl_speaker_2_semantic_prompt.npy"
423
+ },
424
+ "pl_speaker_3": {
425
+ "coarse_prompt": "speaker_embeddings/pl_speaker_3_coarse_prompt.npy",
426
+ "fine_prompt": "speaker_embeddings/pl_speaker_3_fine_prompt.npy",
427
+ "semantic_prompt": "speaker_embeddings/pl_speaker_3_semantic_prompt.npy"
428
+ },
429
+ "pl_speaker_4": {
430
+ "coarse_prompt": "speaker_embeddings/pl_speaker_4_coarse_prompt.npy",
431
+ "fine_prompt": "speaker_embeddings/pl_speaker_4_fine_prompt.npy",
432
+ "semantic_prompt": "speaker_embeddings/pl_speaker_4_semantic_prompt.npy"
433
+ },
434
+ "pl_speaker_5": {
435
+ "coarse_prompt": "speaker_embeddings/pl_speaker_5_coarse_prompt.npy",
436
+ "fine_prompt": "speaker_embeddings/pl_speaker_5_fine_prompt.npy",
437
+ "semantic_prompt": "speaker_embeddings/pl_speaker_5_semantic_prompt.npy"
438
+ },
439
+ "pl_speaker_6": {
440
+ "coarse_prompt": "speaker_embeddings/pl_speaker_6_coarse_prompt.npy",
441
+ "fine_prompt": "speaker_embeddings/pl_speaker_6_fine_prompt.npy",
442
+ "semantic_prompt": "speaker_embeddings/pl_speaker_6_semantic_prompt.npy"
443
+ },
444
+ "pl_speaker_7": {
445
+ "coarse_prompt": "speaker_embeddings/pl_speaker_7_coarse_prompt.npy",
446
+ "fine_prompt": "speaker_embeddings/pl_speaker_7_fine_prompt.npy",
447
+ "semantic_prompt": "speaker_embeddings/pl_speaker_7_semantic_prompt.npy"
448
+ },
449
+ "pl_speaker_8": {
450
+ "coarse_prompt": "speaker_embeddings/pl_speaker_8_coarse_prompt.npy",
451
+ "fine_prompt": "speaker_embeddings/pl_speaker_8_fine_prompt.npy",
452
+ "semantic_prompt": "speaker_embeddings/pl_speaker_8_semantic_prompt.npy"
453
+ },
454
+ "pl_speaker_9": {
455
+ "coarse_prompt": "speaker_embeddings/pl_speaker_9_coarse_prompt.npy",
456
+ "fine_prompt": "speaker_embeddings/pl_speaker_9_fine_prompt.npy",
457
+ "semantic_prompt": "speaker_embeddings/pl_speaker_9_semantic_prompt.npy"
458
+ },
459
+ "pt_speaker_0": {
460
+ "coarse_prompt": "speaker_embeddings/pt_speaker_0_coarse_prompt.npy",
461
+ "fine_prompt": "speaker_embeddings/pt_speaker_0_fine_prompt.npy",
462
+ "semantic_prompt": "speaker_embeddings/pt_speaker_0_semantic_prompt.npy"
463
+ },
464
+ "pt_speaker_1": {
465
+ "coarse_prompt": "speaker_embeddings/pt_speaker_1_coarse_prompt.npy",
466
+ "fine_prompt": "speaker_embeddings/pt_speaker_1_fine_prompt.npy",
467
+ "semantic_prompt": "speaker_embeddings/pt_speaker_1_semantic_prompt.npy"
468
+ },
469
+ "pt_speaker_2": {
470
+ "coarse_prompt": "speaker_embeddings/pt_speaker_2_coarse_prompt.npy",
471
+ "fine_prompt": "speaker_embeddings/pt_speaker_2_fine_prompt.npy",
472
+ "semantic_prompt": "speaker_embeddings/pt_speaker_2_semantic_prompt.npy"
473
+ },
474
+ "pt_speaker_3": {
475
+ "coarse_prompt": "speaker_embeddings/pt_speaker_3_coarse_prompt.npy",
476
+ "fine_prompt": "speaker_embeddings/pt_speaker_3_fine_prompt.npy",
477
+ "semantic_prompt": "speaker_embeddings/pt_speaker_3_semantic_prompt.npy"
478
+ },
479
+ "pt_speaker_4": {
480
+ "coarse_prompt": "speaker_embeddings/pt_speaker_4_coarse_prompt.npy",
481
+ "fine_prompt": "speaker_embeddings/pt_speaker_4_fine_prompt.npy",
482
+ "semantic_prompt": "speaker_embeddings/pt_speaker_4_semantic_prompt.npy"
483
+ },
484
+ "pt_speaker_5": {
485
+ "coarse_prompt": "speaker_embeddings/pt_speaker_5_coarse_prompt.npy",
486
+ "fine_prompt": "speaker_embeddings/pt_speaker_5_fine_prompt.npy",
487
+ "semantic_prompt": "speaker_embeddings/pt_speaker_5_semantic_prompt.npy"
488
+ },
489
+ "pt_speaker_6": {
490
+ "coarse_prompt": "speaker_embeddings/pt_speaker_6_coarse_prompt.npy",
491
+ "fine_prompt": "speaker_embeddings/pt_speaker_6_fine_prompt.npy",
492
+ "semantic_prompt": "speaker_embeddings/pt_speaker_6_semantic_prompt.npy"
493
+ },
494
+ "pt_speaker_7": {
495
+ "coarse_prompt": "speaker_embeddings/pt_speaker_7_coarse_prompt.npy",
496
+ "fine_prompt": "speaker_embeddings/pt_speaker_7_fine_prompt.npy",
497
+ "semantic_prompt": "speaker_embeddings/pt_speaker_7_semantic_prompt.npy"
498
+ },
499
+ "pt_speaker_8": {
500
+ "coarse_prompt": "speaker_embeddings/pt_speaker_8_coarse_prompt.npy",
501
+ "fine_prompt": "speaker_embeddings/pt_speaker_8_fine_prompt.npy",
502
+ "semantic_prompt": "speaker_embeddings/pt_speaker_8_semantic_prompt.npy"
503
+ },
504
+ "pt_speaker_9": {
505
+ "coarse_prompt": "speaker_embeddings/pt_speaker_9_coarse_prompt.npy",
506
+ "fine_prompt": "speaker_embeddings/pt_speaker_9_fine_prompt.npy",
507
+ "semantic_prompt": "speaker_embeddings/pt_speaker_9_semantic_prompt.npy"
508
+ },
509
+ "repo_or_path": "suno/bark-small",
510
+ "ru_speaker_0": {
511
+ "coarse_prompt": "speaker_embeddings/ru_speaker_0_coarse_prompt.npy",
512
+ "fine_prompt": "speaker_embeddings/ru_speaker_0_fine_prompt.npy",
513
+ "semantic_prompt": "speaker_embeddings/ru_speaker_0_semantic_prompt.npy"
514
+ },
515
+ "ru_speaker_1": {
516
+ "coarse_prompt": "speaker_embeddings/ru_speaker_1_coarse_prompt.npy",
517
+ "fine_prompt": "speaker_embeddings/ru_speaker_1_fine_prompt.npy",
518
+ "semantic_prompt": "speaker_embeddings/ru_speaker_1_semantic_prompt.npy"
519
+ },
520
+ "ru_speaker_2": {
521
+ "coarse_prompt": "speaker_embeddings/ru_speaker_2_coarse_prompt.npy",
522
+ "fine_prompt": "speaker_embeddings/ru_speaker_2_fine_prompt.npy",
523
+ "semantic_prompt": "speaker_embeddings/ru_speaker_2_semantic_prompt.npy"
524
+ },
525
+ "ru_speaker_3": {
526
+ "coarse_prompt": "speaker_embeddings/ru_speaker_3_coarse_prompt.npy",
527
+ "fine_prompt": "speaker_embeddings/ru_speaker_3_fine_prompt.npy",
528
+ "semantic_prompt": "speaker_embeddings/ru_speaker_3_semantic_prompt.npy"
529
+ },
530
+ "ru_speaker_4": {
531
+ "coarse_prompt": "speaker_embeddings/ru_speaker_4_coarse_prompt.npy",
532
+ "fine_prompt": "speaker_embeddings/ru_speaker_4_fine_prompt.npy",
533
+ "semantic_prompt": "speaker_embeddings/ru_speaker_4_semantic_prompt.npy"
534
+ },
535
+ "ru_speaker_5": {
536
+ "coarse_prompt": "speaker_embeddings/ru_speaker_5_coarse_prompt.npy",
537
+ "fine_prompt": "speaker_embeddings/ru_speaker_5_fine_prompt.npy",
538
+ "semantic_prompt": "speaker_embeddings/ru_speaker_5_semantic_prompt.npy"
539
+ },
540
+ "ru_speaker_6": {
541
+ "coarse_prompt": "speaker_embeddings/ru_speaker_6_coarse_prompt.npy",
542
+ "fine_prompt": "speaker_embeddings/ru_speaker_6_fine_prompt.npy",
543
+ "semantic_prompt": "speaker_embeddings/ru_speaker_6_semantic_prompt.npy"
544
+ },
545
+ "ru_speaker_7": {
546
+ "coarse_prompt": "speaker_embeddings/ru_speaker_7_coarse_prompt.npy",
547
+ "fine_prompt": "speaker_embeddings/ru_speaker_7_fine_prompt.npy",
548
+ "semantic_prompt": "speaker_embeddings/ru_speaker_7_semantic_prompt.npy"
549
+ },
550
+ "ru_speaker_8": {
551
+ "coarse_prompt": "speaker_embeddings/ru_speaker_8_coarse_prompt.npy",
552
+ "fine_prompt": "speaker_embeddings/ru_speaker_8_fine_prompt.npy",
553
+ "semantic_prompt": "speaker_embeddings/ru_speaker_8_semantic_prompt.npy"
554
+ },
555
+ "ru_speaker_9": {
556
+ "coarse_prompt": "speaker_embeddings/ru_speaker_9_coarse_prompt.npy",
557
+ "fine_prompt": "speaker_embeddings/ru_speaker_9_fine_prompt.npy",
558
+ "semantic_prompt": "speaker_embeddings/ru_speaker_9_semantic_prompt.npy"
559
+ },
560
+ "tr_speaker_0": {
561
+ "coarse_prompt": "speaker_embeddings/tr_speaker_0_coarse_prompt.npy",
562
+ "fine_prompt": "speaker_embeddings/tr_speaker_0_fine_prompt.npy",
563
+ "semantic_prompt": "speaker_embeddings/tr_speaker_0_semantic_prompt.npy"
564
+ },
565
+ "tr_speaker_1": {
566
+ "coarse_prompt": "speaker_embeddings/tr_speaker_1_coarse_prompt.npy",
567
+ "fine_prompt": "speaker_embeddings/tr_speaker_1_fine_prompt.npy",
568
+ "semantic_prompt": "speaker_embeddings/tr_speaker_1_semantic_prompt.npy"
569
+ },
570
+ "tr_speaker_2": {
571
+ "coarse_prompt": "speaker_embeddings/tr_speaker_2_coarse_prompt.npy",
572
+ "fine_prompt": "speaker_embeddings/tr_speaker_2_fine_prompt.npy",
573
+ "semantic_prompt": "speaker_embeddings/tr_speaker_2_semantic_prompt.npy"
574
+ },
575
+ "tr_speaker_3": {
576
+ "coarse_prompt": "speaker_embeddings/tr_speaker_3_coarse_prompt.npy",
577
+ "fine_prompt": "speaker_embeddings/tr_speaker_3_fine_prompt.npy",
578
+ "semantic_prompt": "speaker_embeddings/tr_speaker_3_semantic_prompt.npy"
579
+ },
580
+ "tr_speaker_4": {
581
+ "coarse_prompt": "speaker_embeddings/tr_speaker_4_coarse_prompt.npy",
582
+ "fine_prompt": "speaker_embeddings/tr_speaker_4_fine_prompt.npy",
583
+ "semantic_prompt": "speaker_embeddings/tr_speaker_4_semantic_prompt.npy"
584
+ },
585
+ "tr_speaker_5": {
586
+ "coarse_prompt": "speaker_embeddings/tr_speaker_5_coarse_prompt.npy",
587
+ "fine_prompt": "speaker_embeddings/tr_speaker_5_fine_prompt.npy",
588
+ "semantic_prompt": "speaker_embeddings/tr_speaker_5_semantic_prompt.npy"
589
+ },
590
+ "tr_speaker_6": {
591
+ "coarse_prompt": "speaker_embeddings/tr_speaker_6_coarse_prompt.npy",
592
+ "fine_prompt": "speaker_embeddings/tr_speaker_6_fine_prompt.npy",
593
+ "semantic_prompt": "speaker_embeddings/tr_speaker_6_semantic_prompt.npy"
594
+ },
595
+ "tr_speaker_7": {
596
+ "coarse_prompt": "speaker_embeddings/tr_speaker_7_coarse_prompt.npy",
597
+ "fine_prompt": "speaker_embeddings/tr_speaker_7_fine_prompt.npy",
598
+ "semantic_prompt": "speaker_embeddings/tr_speaker_7_semantic_prompt.npy"
599
+ },
600
+ "tr_speaker_8": {
601
+ "coarse_prompt": "speaker_embeddings/tr_speaker_8_coarse_prompt.npy",
602
+ "fine_prompt": "speaker_embeddings/tr_speaker_8_fine_prompt.npy",
603
+ "semantic_prompt": "speaker_embeddings/tr_speaker_8_semantic_prompt.npy"
604
+ },
605
+ "tr_speaker_9": {
606
+ "coarse_prompt": "speaker_embeddings/tr_speaker_9_coarse_prompt.npy",
607
+ "fine_prompt": "speaker_embeddings/tr_speaker_9_fine_prompt.npy",
608
+ "semantic_prompt": "speaker_embeddings/tr_speaker_9_semantic_prompt.npy"
609
+ },
610
+ "v2/de_speaker_0": {
611
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_0_coarse_prompt.npy",
612
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_0_fine_prompt.npy",
613
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_0_semantic_prompt.npy"
614
+ },
615
+ "v2/de_speaker_1": {
616
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_1_coarse_prompt.npy",
617
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_1_fine_prompt.npy",
618
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_1_semantic_prompt.npy"
619
+ },
620
+ "v2/de_speaker_2": {
621
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_2_coarse_prompt.npy",
622
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_2_fine_prompt.npy",
623
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_2_semantic_prompt.npy"
624
+ },
625
+ "v2/de_speaker_3": {
626
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_3_coarse_prompt.npy",
627
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_3_fine_prompt.npy",
628
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_3_semantic_prompt.npy"
629
+ },
630
+ "v2/de_speaker_4": {
631
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_4_coarse_prompt.npy",
632
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_4_fine_prompt.npy",
633
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_4_semantic_prompt.npy"
634
+ },
635
+ "v2/de_speaker_5": {
636
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_5_coarse_prompt.npy",
637
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_5_fine_prompt.npy",
638
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_5_semantic_prompt.npy"
639
+ },
640
+ "v2/de_speaker_6": {
641
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_6_coarse_prompt.npy",
642
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_6_fine_prompt.npy",
643
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_6_semantic_prompt.npy"
644
+ },
645
+ "v2/de_speaker_7": {
646
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_7_coarse_prompt.npy",
647
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_7_fine_prompt.npy",
648
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_7_semantic_prompt.npy"
649
+ },
650
+ "v2/de_speaker_8": {
651
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_8_coarse_prompt.npy",
652
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_8_fine_prompt.npy",
653
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_8_semantic_prompt.npy"
654
+ },
655
+ "v2/de_speaker_9": {
656
+ "coarse_prompt": "speaker_embeddings/v2/de_speaker_9_coarse_prompt.npy",
657
+ "fine_prompt": "speaker_embeddings/v2/de_speaker_9_fine_prompt.npy",
658
+ "semantic_prompt": "speaker_embeddings/v2/de_speaker_9_semantic_prompt.npy"
659
+ },
660
+ "v2/en_speaker_0": {
661
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_0_coarse_prompt.npy",
662
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_0_fine_prompt.npy",
663
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_0_semantic_prompt.npy"
664
+ },
665
+ "v2/en_speaker_1": {
666
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_1_coarse_prompt.npy",
667
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_1_fine_prompt.npy",
668
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_1_semantic_prompt.npy"
669
+ },
670
+ "v2/en_speaker_2": {
671
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_2_coarse_prompt.npy",
672
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_2_fine_prompt.npy",
673
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_2_semantic_prompt.npy"
674
+ },
675
+ "v2/en_speaker_3": {
676
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_3_coarse_prompt.npy",
677
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_3_fine_prompt.npy",
678
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_3_semantic_prompt.npy"
679
+ },
680
+ "v2/en_speaker_4": {
681
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_4_coarse_prompt.npy",
682
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_4_fine_prompt.npy",
683
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_4_semantic_prompt.npy"
684
+ },
685
+ "v2/en_speaker_5": {
686
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_5_coarse_prompt.npy",
687
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_5_fine_prompt.npy",
688
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_5_semantic_prompt.npy"
689
+ },
690
+ "v2/en_speaker_6": {
691
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_6_coarse_prompt.npy",
692
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_6_fine_prompt.npy",
693
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_6_semantic_prompt.npy"
694
+ },
695
+ "v2/en_speaker_7": {
696
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_7_coarse_prompt.npy",
697
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_7_fine_prompt.npy",
698
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_7_semantic_prompt.npy"
699
+ },
700
+ "v2/en_speaker_8": {
701
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_8_coarse_prompt.npy",
702
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_8_fine_prompt.npy",
703
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_8_semantic_prompt.npy"
704
+ },
705
+ "v2/en_speaker_9": {
706
+ "coarse_prompt": "speaker_embeddings/v2/en_speaker_9_coarse_prompt.npy",
707
+ "fine_prompt": "speaker_embeddings/v2/en_speaker_9_fine_prompt.npy",
708
+ "semantic_prompt": "speaker_embeddings/v2/en_speaker_9_semantic_prompt.npy"
709
+ },
710
+ "v2/es_speaker_0": {
711
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_0_coarse_prompt.npy",
712
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_0_fine_prompt.npy",
713
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_0_semantic_prompt.npy"
714
+ },
715
+ "v2/es_speaker_1": {
716
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_1_coarse_prompt.npy",
717
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_1_fine_prompt.npy",
718
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_1_semantic_prompt.npy"
719
+ },
720
+ "v2/es_speaker_2": {
721
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_2_coarse_prompt.npy",
722
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_2_fine_prompt.npy",
723
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_2_semantic_prompt.npy"
724
+ },
725
+ "v2/es_speaker_3": {
726
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_3_coarse_prompt.npy",
727
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_3_fine_prompt.npy",
728
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_3_semantic_prompt.npy"
729
+ },
730
+ "v2/es_speaker_4": {
731
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_4_coarse_prompt.npy",
732
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_4_fine_prompt.npy",
733
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_4_semantic_prompt.npy"
734
+ },
735
+ "v2/es_speaker_5": {
736
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_5_coarse_prompt.npy",
737
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_5_fine_prompt.npy",
738
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_5_semantic_prompt.npy"
739
+ },
740
+ "v2/es_speaker_6": {
741
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_6_coarse_prompt.npy",
742
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_6_fine_prompt.npy",
743
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_6_semantic_prompt.npy"
744
+ },
745
+ "v2/es_speaker_7": {
746
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_7_coarse_prompt.npy",
747
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_7_fine_prompt.npy",
748
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_7_semantic_prompt.npy"
749
+ },
750
+ "v2/es_speaker_8": {
751
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_8_coarse_prompt.npy",
752
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_8_fine_prompt.npy",
753
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_8_semantic_prompt.npy"
754
+ },
755
+ "v2/es_speaker_9": {
756
+ "coarse_prompt": "speaker_embeddings/v2/es_speaker_9_coarse_prompt.npy",
757
+ "fine_prompt": "speaker_embeddings/v2/es_speaker_9_fine_prompt.npy",
758
+ "semantic_prompt": "speaker_embeddings/v2/es_speaker_9_semantic_prompt.npy"
759
+ },
760
+ "v2/fr_speaker_0": {
761
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_0_coarse_prompt.npy",
762
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_0_fine_prompt.npy",
763
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_0_semantic_prompt.npy"
764
+ },
765
+ "v2/fr_speaker_1": {
766
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_1_coarse_prompt.npy",
767
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_1_fine_prompt.npy",
768
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_1_semantic_prompt.npy"
769
+ },
770
+ "v2/fr_speaker_2": {
771
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_2_coarse_prompt.npy",
772
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_2_fine_prompt.npy",
773
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_2_semantic_prompt.npy"
774
+ },
775
+ "v2/fr_speaker_3": {
776
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_3_coarse_prompt.npy",
777
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_3_fine_prompt.npy",
778
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_3_semantic_prompt.npy"
779
+ },
780
+ "v2/fr_speaker_4": {
781
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_4_coarse_prompt.npy",
782
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_4_fine_prompt.npy",
783
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_4_semantic_prompt.npy"
784
+ },
785
+ "v2/fr_speaker_5": {
786
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_5_coarse_prompt.npy",
787
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_5_fine_prompt.npy",
788
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_5_semantic_prompt.npy"
789
+ },
790
+ "v2/fr_speaker_6": {
791
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_6_coarse_prompt.npy",
792
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_6_fine_prompt.npy",
793
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_6_semantic_prompt.npy"
794
+ },
795
+ "v2/fr_speaker_7": {
796
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_7_coarse_prompt.npy",
797
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_7_fine_prompt.npy",
798
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_7_semantic_prompt.npy"
799
+ },
800
+ "v2/fr_speaker_8": {
801
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_8_coarse_prompt.npy",
802
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_8_fine_prompt.npy",
803
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_8_semantic_prompt.npy"
804
+ },
805
+ "v2/fr_speaker_9": {
806
+ "coarse_prompt": "speaker_embeddings/v2/fr_speaker_9_coarse_prompt.npy",
807
+ "fine_prompt": "speaker_embeddings/v2/fr_speaker_9_fine_prompt.npy",
808
+ "semantic_prompt": "speaker_embeddings/v2/fr_speaker_9_semantic_prompt.npy"
809
+ },
810
+ "v2/hi_speaker_0": {
811
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_0_coarse_prompt.npy",
812
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_0_fine_prompt.npy",
813
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_0_semantic_prompt.npy"
814
+ },
815
+ "v2/hi_speaker_1": {
816
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_1_coarse_prompt.npy",
817
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_1_fine_prompt.npy",
818
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_1_semantic_prompt.npy"
819
+ },
820
+ "v2/hi_speaker_2": {
821
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_2_coarse_prompt.npy",
822
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_2_fine_prompt.npy",
823
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_2_semantic_prompt.npy"
824
+ },
825
+ "v2/hi_speaker_3": {
826
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_3_coarse_prompt.npy",
827
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_3_fine_prompt.npy",
828
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_3_semantic_prompt.npy"
829
+ },
830
+ "v2/hi_speaker_4": {
831
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_4_coarse_prompt.npy",
832
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_4_fine_prompt.npy",
833
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_4_semantic_prompt.npy"
834
+ },
835
+ "v2/hi_speaker_5": {
836
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_5_coarse_prompt.npy",
837
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_5_fine_prompt.npy",
838
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_5_semantic_prompt.npy"
839
+ },
840
+ "v2/hi_speaker_6": {
841
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_6_coarse_prompt.npy",
842
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_6_fine_prompt.npy",
843
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_6_semantic_prompt.npy"
844
+ },
845
+ "v2/hi_speaker_7": {
846
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_7_coarse_prompt.npy",
847
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_7_fine_prompt.npy",
848
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_7_semantic_prompt.npy"
849
+ },
850
+ "v2/hi_speaker_8": {
851
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_8_coarse_prompt.npy",
852
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_8_fine_prompt.npy",
853
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_8_semantic_prompt.npy"
854
+ },
855
+ "v2/hi_speaker_9": {
856
+ "coarse_prompt": "speaker_embeddings/v2/hi_speaker_9_coarse_prompt.npy",
857
+ "fine_prompt": "speaker_embeddings/v2/hi_speaker_9_fine_prompt.npy",
858
+ "semantic_prompt": "speaker_embeddings/v2/hi_speaker_9_semantic_prompt.npy"
859
+ },
860
+ "v2/it_speaker_0": {
861
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_0_coarse_prompt.npy",
862
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_0_fine_prompt.npy",
863
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_0_semantic_prompt.npy"
864
+ },
865
+ "v2/it_speaker_1": {
866
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_1_coarse_prompt.npy",
867
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_1_fine_prompt.npy",
868
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_1_semantic_prompt.npy"
869
+ },
870
+ "v2/it_speaker_2": {
871
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_2_coarse_prompt.npy",
872
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_2_fine_prompt.npy",
873
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_2_semantic_prompt.npy"
874
+ },
875
+ "v2/it_speaker_3": {
876
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_3_coarse_prompt.npy",
877
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_3_fine_prompt.npy",
878
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_3_semantic_prompt.npy"
879
+ },
880
+ "v2/it_speaker_4": {
881
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_4_coarse_prompt.npy",
882
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_4_fine_prompt.npy",
883
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_4_semantic_prompt.npy"
884
+ },
885
+ "v2/it_speaker_5": {
886
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_5_coarse_prompt.npy",
887
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_5_fine_prompt.npy",
888
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_5_semantic_prompt.npy"
889
+ },
890
+ "v2/it_speaker_6": {
891
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_6_coarse_prompt.npy",
892
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_6_fine_prompt.npy",
893
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_6_semantic_prompt.npy"
894
+ },
895
+ "v2/it_speaker_7": {
896
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_7_coarse_prompt.npy",
897
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_7_fine_prompt.npy",
898
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_7_semantic_prompt.npy"
899
+ },
900
+ "v2/it_speaker_8": {
901
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_8_coarse_prompt.npy",
902
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_8_fine_prompt.npy",
903
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_8_semantic_prompt.npy"
904
+ },
905
+ "v2/it_speaker_9": {
906
+ "coarse_prompt": "speaker_embeddings/v2/it_speaker_9_coarse_prompt.npy",
907
+ "fine_prompt": "speaker_embeddings/v2/it_speaker_9_fine_prompt.npy",
908
+ "semantic_prompt": "speaker_embeddings/v2/it_speaker_9_semantic_prompt.npy"
909
+ },
910
+ "v2/ja_speaker_0": {
911
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_0_coarse_prompt.npy",
912
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_0_fine_prompt.npy",
913
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_0_semantic_prompt.npy"
914
+ },
915
+ "v2/ja_speaker_1": {
916
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_1_coarse_prompt.npy",
917
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_1_fine_prompt.npy",
918
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_1_semantic_prompt.npy"
919
+ },
920
+ "v2/ja_speaker_2": {
921
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_2_coarse_prompt.npy",
922
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_2_fine_prompt.npy",
923
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_2_semantic_prompt.npy"
924
+ },
925
+ "v2/ja_speaker_3": {
926
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_3_coarse_prompt.npy",
927
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_3_fine_prompt.npy",
928
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_3_semantic_prompt.npy"
929
+ },
930
+ "v2/ja_speaker_4": {
931
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_4_coarse_prompt.npy",
932
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_4_fine_prompt.npy",
933
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_4_semantic_prompt.npy"
934
+ },
935
+ "v2/ja_speaker_5": {
936
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_5_coarse_prompt.npy",
937
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_5_fine_prompt.npy",
938
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_5_semantic_prompt.npy"
939
+ },
940
+ "v2/ja_speaker_6": {
941
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_6_coarse_prompt.npy",
942
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_6_fine_prompt.npy",
943
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_6_semantic_prompt.npy"
944
+ },
945
+ "v2/ja_speaker_7": {
946
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_7_coarse_prompt.npy",
947
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_7_fine_prompt.npy",
948
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_7_semantic_prompt.npy"
949
+ },
950
+ "v2/ja_speaker_8": {
951
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_8_coarse_prompt.npy",
952
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_8_fine_prompt.npy",
953
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_8_semantic_prompt.npy"
954
+ },
955
+ "v2/ja_speaker_9": {
956
+ "coarse_prompt": "speaker_embeddings/v2/ja_speaker_9_coarse_prompt.npy",
957
+ "fine_prompt": "speaker_embeddings/v2/ja_speaker_9_fine_prompt.npy",
958
+ "semantic_prompt": "speaker_embeddings/v2/ja_speaker_9_semantic_prompt.npy"
959
+ },
960
+ "v2/ko_speaker_0": {
961
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_0_coarse_prompt.npy",
962
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_0_fine_prompt.npy",
963
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_0_semantic_prompt.npy"
964
+ },
965
+ "v2/ko_speaker_1": {
966
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_1_coarse_prompt.npy",
967
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_1_fine_prompt.npy",
968
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_1_semantic_prompt.npy"
969
+ },
970
+ "v2/ko_speaker_2": {
971
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_2_coarse_prompt.npy",
972
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_2_fine_prompt.npy",
973
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_2_semantic_prompt.npy"
974
+ },
975
+ "v2/ko_speaker_3": {
976
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_3_coarse_prompt.npy",
977
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_3_fine_prompt.npy",
978
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_3_semantic_prompt.npy"
979
+ },
980
+ "v2/ko_speaker_4": {
981
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_4_coarse_prompt.npy",
982
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_4_fine_prompt.npy",
983
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_4_semantic_prompt.npy"
984
+ },
985
+ "v2/ko_speaker_5": {
986
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_5_coarse_prompt.npy",
987
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_5_fine_prompt.npy",
988
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_5_semantic_prompt.npy"
989
+ },
990
+ "v2/ko_speaker_6": {
991
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_6_coarse_prompt.npy",
992
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_6_fine_prompt.npy",
993
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_6_semantic_prompt.npy"
994
+ },
995
+ "v2/ko_speaker_7": {
996
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_7_coarse_prompt.npy",
997
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_7_fine_prompt.npy",
998
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_7_semantic_prompt.npy"
999
+ },
1000
+ "v2/ko_speaker_8": {
1001
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_8_coarse_prompt.npy",
1002
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_8_fine_prompt.npy",
1003
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_8_semantic_prompt.npy"
1004
+ },
1005
+ "v2/ko_speaker_9": {
1006
+ "coarse_prompt": "speaker_embeddings/v2/ko_speaker_9_coarse_prompt.npy",
1007
+ "fine_prompt": "speaker_embeddings/v2/ko_speaker_9_fine_prompt.npy",
1008
+ "semantic_prompt": "speaker_embeddings/v2/ko_speaker_9_semantic_prompt.npy"
1009
+ },
1010
+ "v2/pl_speaker_0": {
1011
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_0_coarse_prompt.npy",
1012
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_0_fine_prompt.npy",
1013
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_0_semantic_prompt.npy"
1014
+ },
1015
+ "v2/pl_speaker_1": {
1016
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_1_coarse_prompt.npy",
1017
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_1_fine_prompt.npy",
1018
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_1_semantic_prompt.npy"
1019
+ },
1020
+ "v2/pl_speaker_2": {
1021
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_2_coarse_prompt.npy",
1022
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_2_fine_prompt.npy",
1023
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_2_semantic_prompt.npy"
1024
+ },
1025
+ "v2/pl_speaker_3": {
1026
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_3_coarse_prompt.npy",
1027
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_3_fine_prompt.npy",
1028
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_3_semantic_prompt.npy"
1029
+ },
1030
+ "v2/pl_speaker_4": {
1031
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_4_coarse_prompt.npy",
1032
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_4_fine_prompt.npy",
1033
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_4_semantic_prompt.npy"
1034
+ },
1035
+ "v2/pl_speaker_5": {
1036
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_5_coarse_prompt.npy",
1037
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_5_fine_prompt.npy",
1038
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_5_semantic_prompt.npy"
1039
+ },
1040
+ "v2/pl_speaker_6": {
1041
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_6_coarse_prompt.npy",
1042
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_6_fine_prompt.npy",
1043
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_6_semantic_prompt.npy"
1044
+ },
1045
+ "v2/pl_speaker_7": {
1046
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_7_coarse_prompt.npy",
1047
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_7_fine_prompt.npy",
1048
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_7_semantic_prompt.npy"
1049
+ },
1050
+ "v2/pl_speaker_8": {
1051
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_8_coarse_prompt.npy",
1052
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_8_fine_prompt.npy",
1053
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_8_semantic_prompt.npy"
1054
+ },
1055
+ "v2/pl_speaker_9": {
1056
+ "coarse_prompt": "speaker_embeddings/v2/pl_speaker_9_coarse_prompt.npy",
1057
+ "fine_prompt": "speaker_embeddings/v2/pl_speaker_9_fine_prompt.npy",
1058
+ "semantic_prompt": "speaker_embeddings/v2/pl_speaker_9_semantic_prompt.npy"
1059
+ },
1060
+ "v2/pt_speaker_0": {
1061
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_0_coarse_prompt.npy",
1062
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_0_fine_prompt.npy",
1063
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_0_semantic_prompt.npy"
1064
+ },
1065
+ "v2/pt_speaker_1": {
1066
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_1_coarse_prompt.npy",
1067
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_1_fine_prompt.npy",
1068
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_1_semantic_prompt.npy"
1069
+ },
1070
+ "v2/pt_speaker_2": {
1071
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_2_coarse_prompt.npy",
1072
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_2_fine_prompt.npy",
1073
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_2_semantic_prompt.npy"
1074
+ },
1075
+ "v2/pt_speaker_3": {
1076
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_3_coarse_prompt.npy",
1077
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_3_fine_prompt.npy",
1078
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_3_semantic_prompt.npy"
1079
+ },
1080
+ "v2/pt_speaker_4": {
1081
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_4_coarse_prompt.npy",
1082
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_4_fine_prompt.npy",
1083
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_4_semantic_prompt.npy"
1084
+ },
1085
+ "v2/pt_speaker_5": {
1086
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_5_coarse_prompt.npy",
1087
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_5_fine_prompt.npy",
1088
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_5_semantic_prompt.npy"
1089
+ },
1090
+ "v2/pt_speaker_6": {
1091
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_6_coarse_prompt.npy",
1092
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_6_fine_prompt.npy",
1093
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_6_semantic_prompt.npy"
1094
+ },
1095
+ "v2/pt_speaker_7": {
1096
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_7_coarse_prompt.npy",
1097
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_7_fine_prompt.npy",
1098
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_7_semantic_prompt.npy"
1099
+ },
1100
+ "v2/pt_speaker_8": {
1101
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_8_coarse_prompt.npy",
1102
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_8_fine_prompt.npy",
1103
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_8_semantic_prompt.npy"
1104
+ },
1105
+ "v2/pt_speaker_9": {
1106
+ "coarse_prompt": "speaker_embeddings/v2/pt_speaker_9_coarse_prompt.npy",
1107
+ "fine_prompt": "speaker_embeddings/v2/pt_speaker_9_fine_prompt.npy",
1108
+ "semantic_prompt": "speaker_embeddings/v2/pt_speaker_9_semantic_prompt.npy"
1109
+ },
1110
+ "v2/ru_speaker_0": {
1111
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_0_coarse_prompt.npy",
1112
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_0_fine_prompt.npy",
1113
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_0_semantic_prompt.npy"
1114
+ },
1115
+ "v2/ru_speaker_1": {
1116
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_1_coarse_prompt.npy",
1117
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_1_fine_prompt.npy",
1118
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_1_semantic_prompt.npy"
1119
+ },
1120
+ "v2/ru_speaker_2": {
1121
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_2_coarse_prompt.npy",
1122
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_2_fine_prompt.npy",
1123
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_2_semantic_prompt.npy"
1124
+ },
1125
+ "v2/ru_speaker_3": {
1126
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_3_coarse_prompt.npy",
1127
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_3_fine_prompt.npy",
1128
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_3_semantic_prompt.npy"
1129
+ },
1130
+ "v2/ru_speaker_4": {
1131
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_4_coarse_prompt.npy",
1132
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_4_fine_prompt.npy",
1133
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_4_semantic_prompt.npy"
1134
+ },
1135
+ "v2/ru_speaker_5": {
1136
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_5_coarse_prompt.npy",
1137
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_5_fine_prompt.npy",
1138
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_5_semantic_prompt.npy"
1139
+ },
1140
+ "v2/ru_speaker_6": {
1141
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_6_coarse_prompt.npy",
1142
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_6_fine_prompt.npy",
1143
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_6_semantic_prompt.npy"
1144
+ },
1145
+ "v2/ru_speaker_7": {
1146
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_7_coarse_prompt.npy",
1147
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_7_fine_prompt.npy",
1148
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_7_semantic_prompt.npy"
1149
+ },
1150
+ "v2/ru_speaker_8": {
1151
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_8_coarse_prompt.npy",
1152
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_8_fine_prompt.npy",
1153
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_8_semantic_prompt.npy"
1154
+ },
1155
+ "v2/ru_speaker_9": {
1156
+ "coarse_prompt": "speaker_embeddings/v2/ru_speaker_9_coarse_prompt.npy",
1157
+ "fine_prompt": "speaker_embeddings/v2/ru_speaker_9_fine_prompt.npy",
1158
+ "semantic_prompt": "speaker_embeddings/v2/ru_speaker_9_semantic_prompt.npy"
1159
+ },
1160
+ "v2/tr_speaker_0": {
1161
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_0_coarse_prompt.npy",
1162
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_0_fine_prompt.npy",
1163
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_0_semantic_prompt.npy"
1164
+ },
1165
+ "v2/tr_speaker_1": {
1166
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_1_coarse_prompt.npy",
1167
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_1_fine_prompt.npy",
1168
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_1_semantic_prompt.npy"
1169
+ },
1170
+ "v2/tr_speaker_2": {
1171
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_2_coarse_prompt.npy",
1172
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_2_fine_prompt.npy",
1173
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_2_semantic_prompt.npy"
1174
+ },
1175
+ "v2/tr_speaker_3": {
1176
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_3_coarse_prompt.npy",
1177
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_3_fine_prompt.npy",
1178
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_3_semantic_prompt.npy"
1179
+ },
1180
+ "v2/tr_speaker_4": {
1181
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_4_coarse_prompt.npy",
1182
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_4_fine_prompt.npy",
1183
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_4_semantic_prompt.npy"
1184
+ },
1185
+ "v2/tr_speaker_5": {
1186
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_5_coarse_prompt.npy",
1187
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_5_fine_prompt.npy",
1188
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_5_semantic_prompt.npy"
1189
+ },
1190
+ "v2/tr_speaker_6": {
1191
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_6_coarse_prompt.npy",
1192
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_6_fine_prompt.npy",
1193
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_6_semantic_prompt.npy"
1194
+ },
1195
+ "v2/tr_speaker_7": {
1196
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_7_coarse_prompt.npy",
1197
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_7_fine_prompt.npy",
1198
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_7_semantic_prompt.npy"
1199
+ },
1200
+ "v2/tr_speaker_8": {
1201
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_8_coarse_prompt.npy",
1202
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_8_fine_prompt.npy",
1203
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_8_semantic_prompt.npy"
1204
+ },
1205
+ "v2/tr_speaker_9": {
1206
+ "coarse_prompt": "speaker_embeddings/v2/tr_speaker_9_coarse_prompt.npy",
1207
+ "fine_prompt": "speaker_embeddings/v2/tr_speaker_9_fine_prompt.npy",
1208
+ "semantic_prompt": "speaker_embeddings/v2/tr_speaker_9_semantic_prompt.npy"
1209
+ },
1210
+ "v2/zh_speaker_0": {
1211
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_0_coarse_prompt.npy",
1212
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_0_fine_prompt.npy",
1213
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_0_semantic_prompt.npy"
1214
+ },
1215
+ "v2/zh_speaker_1": {
1216
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_1_coarse_prompt.npy",
1217
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_1_fine_prompt.npy",
1218
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_1_semantic_prompt.npy"
1219
+ },
1220
+ "v2/zh_speaker_2": {
1221
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_2_coarse_prompt.npy",
1222
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_2_fine_prompt.npy",
1223
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_2_semantic_prompt.npy"
1224
+ },
1225
+ "v2/zh_speaker_3": {
1226
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_3_coarse_prompt.npy",
1227
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_3_fine_prompt.npy",
1228
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_3_semantic_prompt.npy"
1229
+ },
1230
+ "v2/zh_speaker_4": {
1231
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_4_coarse_prompt.npy",
1232
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_4_fine_prompt.npy",
1233
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_4_semantic_prompt.npy"
1234
+ },
1235
+ "v2/zh_speaker_5": {
1236
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_5_coarse_prompt.npy",
1237
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_5_fine_prompt.npy",
1238
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_5_semantic_prompt.npy"
1239
+ },
1240
+ "v2/zh_speaker_6": {
1241
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_6_coarse_prompt.npy",
1242
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_6_fine_prompt.npy",
1243
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_6_semantic_prompt.npy"
1244
+ },
1245
+ "v2/zh_speaker_7": {
1246
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_7_coarse_prompt.npy",
1247
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_7_fine_prompt.npy",
1248
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_7_semantic_prompt.npy"
1249
+ },
1250
+ "v2/zh_speaker_8": {
1251
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_8_coarse_prompt.npy",
1252
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_8_fine_prompt.npy",
1253
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_8_semantic_prompt.npy"
1254
+ },
1255
+ "v2/zh_speaker_9": {
1256
+ "coarse_prompt": "speaker_embeddings/v2/zh_speaker_9_coarse_prompt.npy",
1257
+ "fine_prompt": "speaker_embeddings/v2/zh_speaker_9_fine_prompt.npy",
1258
+ "semantic_prompt": "speaker_embeddings/v2/zh_speaker_9_semantic_prompt.npy"
1259
+ },
1260
+ "zh_speaker_0": {
1261
+ "coarse_prompt": "speaker_embeddings/zh_speaker_0_coarse_prompt.npy",
1262
+ "fine_prompt": "speaker_embeddings/zh_speaker_0_fine_prompt.npy",
1263
+ "semantic_prompt": "speaker_embeddings/zh_speaker_0_semantic_prompt.npy"
1264
+ },
1265
+ "zh_speaker_1": {
1266
+ "coarse_prompt": "speaker_embeddings/zh_speaker_1_coarse_prompt.npy",
1267
+ "fine_prompt": "speaker_embeddings/zh_speaker_1_fine_prompt.npy",
1268
+ "semantic_prompt": "speaker_embeddings/zh_speaker_1_semantic_prompt.npy"
1269
+ },
1270
+ "zh_speaker_2": {
1271
+ "coarse_prompt": "speaker_embeddings/zh_speaker_2_coarse_prompt.npy",
1272
+ "fine_prompt": "speaker_embeddings/zh_speaker_2_fine_prompt.npy",
1273
+ "semantic_prompt": "speaker_embeddings/zh_speaker_2_semantic_prompt.npy"
1274
+ },
1275
+ "zh_speaker_3": {
1276
+ "coarse_prompt": "speaker_embeddings/zh_speaker_3_coarse_prompt.npy",
1277
+ "fine_prompt": "speaker_embeddings/zh_speaker_3_fine_prompt.npy",
1278
+ "semantic_prompt": "speaker_embeddings/zh_speaker_3_semantic_prompt.npy"
1279
+ },
1280
+ "zh_speaker_4": {
1281
+ "coarse_prompt": "speaker_embeddings/zh_speaker_4_coarse_prompt.npy",
1282
+ "fine_prompt": "speaker_embeddings/zh_speaker_4_fine_prompt.npy",
1283
+ "semantic_prompt": "speaker_embeddings/zh_speaker_4_semantic_prompt.npy"
1284
+ },
1285
+ "zh_speaker_5": {
1286
+ "coarse_prompt": "speaker_embeddings/zh_speaker_5_coarse_prompt.npy",
1287
+ "fine_prompt": "speaker_embeddings/zh_speaker_5_fine_prompt.npy",
1288
+ "semantic_prompt": "speaker_embeddings/zh_speaker_5_semantic_prompt.npy"
1289
+ },
1290
+ "zh_speaker_6": {
1291
+ "coarse_prompt": "speaker_embeddings/zh_speaker_6_coarse_prompt.npy",
1292
+ "fine_prompt": "speaker_embeddings/zh_speaker_6_fine_prompt.npy",
1293
+ "semantic_prompt": "speaker_embeddings/zh_speaker_6_semantic_prompt.npy"
1294
+ },
1295
+ "zh_speaker_7": {
1296
+ "coarse_prompt": "speaker_embeddings/zh_speaker_7_coarse_prompt.npy",
1297
+ "fine_prompt": "speaker_embeddings/zh_speaker_7_fine_prompt.npy",
1298
+ "semantic_prompt": "speaker_embeddings/zh_speaker_7_semantic_prompt.npy"
1299
+ },
1300
+ "zh_speaker_8": {
1301
+ "coarse_prompt": "speaker_embeddings/zh_speaker_8_coarse_prompt.npy",
1302
+ "fine_prompt": "speaker_embeddings/zh_speaker_8_fine_prompt.npy",
1303
+ "semantic_prompt": "speaker_embeddings/zh_speaker_8_semantic_prompt.npy"
1304
+ },
1305
+ "zh_speaker_9": {
1306
+ "coarse_prompt": "speaker_embeddings/zh_speaker_9_coarse_prompt.npy",
1307
+ "fine_prompt": "speaker_embeddings/zh_speaker_9_fine_prompt.npy",
1308
+ "semantic_prompt": "speaker_embeddings/zh_speaker_9_semantic_prompt.npy"
1309
+ }
1310
+ }
1311
+ }
speaker_embeddings/announcer_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c77d5e1f9ab39e489d27628ab26dfae7a8b4cae4ba72658a1a94927f7ff5132
3
+ size 3128
speaker_embeddings/announcer_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7aac73cd2a14c1939a718aeb48b064d149b39f773f1dcc9b93be6f79ebd33755
3
+ size 12128
speaker_embeddings/announcer_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5564642cbd84db6f4d1d8a522bd26329b4f4cd8cd81e1965b90baf1ee80c7ff
3
+ size 1126
speaker_embeddings/de_speaker_0_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3eb78af32779b5fabbb92e7cdbdeaaa34f5f8ad1085e213b101165665c5ce4cb
3
+ size 9472
speaker_embeddings/de_speaker_0_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2421bcbeb7fadfcf1c0b33e38e316474d4a95139d4dd0b3d861e72c8fd77c36
3
+ size 18816
speaker_embeddings/de_speaker_0_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44e5539dbcd158a2471c59b46cde7f0564d108cd2fd935e3664b2ec733ad4393
3
+ size 3240
speaker_embeddings/de_speaker_1_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ac93157e39cd0fa7a5e9d0404d6adcb3daaec8e820b42b72b96de62e19ca89b
3
+ size 9472
speaker_embeddings/de_speaker_1_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f8eaca4b732dd3264213ac387f4c91f83a574563b0d72368b237df03fe92b4d
3
+ size 18816
speaker_embeddings/de_speaker_1_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98cde60e36a6265d6534e4c6422436f201a9bfcdc87f9817d6faf6ab4f22554b
3
+ size 3240
speaker_embeddings/de_speaker_2_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd5e9d07b0480dda2c3661de1da0671f34419c029c8e7410705139a38fb133d3
3
+ size 6944
speaker_embeddings/de_speaker_2_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f11b22a8a24428ca0114e4fcc770c0ca554376de95be7ac5ac560ae4f1988fb
3
+ size 13760
speaker_embeddings/de_speaker_2_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8efe5707a724632900ab5cb274ffdbac0d39fd6dda1be49833d63311df478ef3
3
+ size 2400
speaker_embeddings/de_speaker_3_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf0fb48dd18c79996fc800abbd3407dd967b21a36bda695584e29d302ead274a
3
+ size 8608
speaker_embeddings/de_speaker_3_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:229455388c628e9d168ad11940041b15ed3b40f724ab3a4747f6cb38b2f5230a
3
+ size 17088
speaker_embeddings/de_speaker_3_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88fee5c6b974dbfb1168b55f73990be270847a91a2e7895d298b188f3ad7da9f
3
+ size 2952
speaker_embeddings/de_speaker_4_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0268e63ef8aa7c6113af4a840c3581c08edea0eacec8aaa7d52da8012d8c16f5
3
+ size 5984
speaker_embeddings/de_speaker_4_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21c41cce8756afcd8b690ad8701e35f33b4c0be3c63bae5f6ebbc10e5cb3c150
3
+ size 11840
speaker_embeddings/de_speaker_4_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a03e84c7f1f4dd97f033d999fea8eb76c4dde6883e1dae67303b6a46bdb57803
3
+ size 2080
speaker_embeddings/de_speaker_5_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f4272d4c3928264a36a406b9c0d306db99316180dca8e96504c93244fc61d96
3
+ size 10416
speaker_embeddings/de_speaker_5_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dac52541ccb321598997fe0e27ebadf5540eb6e53a8b3a614d7cabfcbae4b2d0
3
+ size 20704
speaker_embeddings/de_speaker_5_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:412b3a1f45525390e84bd2109bb86af782da2245adcbdd8803c1b362a311075c
3
+ size 3552
speaker_embeddings/de_speaker_6_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc029718234023df88f9ae4ea830a90115bd2256b10fe8c3c725483392c0f6c2
3
+ size 9408
speaker_embeddings/de_speaker_6_fine_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3414b015ad3567a2590f5b044f3d2043270bef8b03cb1348b7265d21bbca2299
3
+ size 18688
speaker_embeddings/de_speaker_6_semantic_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2300ace5fda24f16a27b345a6f54c19e2e2363808cac0a58bac12518b4c8f77d
3
+ size 3216
speaker_embeddings/de_speaker_7_coarse_prompt.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd86cb193cb39a077a2cb2540af2953ce7f2f480ec3129c5191e730b87329493
3
+ size 17696