Hugging Face's logo

Spaces:
matsuap
/
Style-Bert-VITS2-Editor-Demo
Paused

App Files Community
matsuap commited on
Commit
de1987c
•
1 Parent(s): f201e5f

Upload 11 files

Browse files
Files changed (11) hide show
  1. model_assets/Matsumura_v1/Matsumura_v1_e100_s1400.safetensors +3 -0
  2. model_assets/Matsumura_v1/Matsumura_v1_e72_s1000.safetensors +3 -0
  3. model_assets/Matsumura_v1/config.json +118 -0
  4. model_assets/Matsumura_v1/style_vectors.npy +3 -0
  5. model_assets/Matsumura_v1_jvnv-M2-jp/Matsumura_v1_jvnv-M2-jp.safetensors +3 -0
  6. model_assets/Matsumura_v1_jvnv-M2-jp/config.json +118 -0
  7. model_assets/Matsumura_v1_jvnv-M2-jp/recipe.json +10 -0
  8. model_assets/Matsumura_v1_jvnv-M2-jp/style_vectors.npy +3 -0
  9. model_assets/NotAnimeJPManySpeaker/NotAnimeJPManySpeaker_e120_s22200.safetensors +3 -0
  10. model_assets/NotAnimeJPManySpeaker/config.json +145 -0
  11. model_assets/NotAnimeJPManySpeaker/style_vectors.npy +3 -0
model_assets/Matsumura_v1/Matsumura_v1_e100_s1400.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d349a2c58ba1a43d1fa11b82ba8bf12698511135166e170fdb5d39500ba2f74
3
+ size 251150980
model_assets/Matsumura_v1/Matsumura_v1_e72_s1000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80d58b2b604f90b4c5bd0e75ebf62ddacf35aed0db9b04db30c8daf683d9712c
3
+ size 251150980
model_assets/Matsumura_v1/config.json ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "Matsumura_v1",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 100,
8
+ "learning_rate": 0.0001,
9
+ "betas": [
10
+ 0.8,
11
+ 0.99
12
+ ],
13
+ "eps": 1e-09,
14
+ "batch_size": 2,
15
+ "bf16_run": false,
16
+ "fp16_run": false,
17
+ "lr_decay": 0.99996,
18
+ "segment_size": 16384,
19
+ "init_lr_ratio": 1,
20
+ "warmup_epochs": 0,
21
+ "c_mel": 45,
22
+ "c_kl": 1.0,
23
+ "c_commit": 100,
24
+ "skip_optimizer": false,
25
+ "freeze_ZH_bert": false,
26
+ "freeze_JP_bert": false,
27
+ "freeze_EN_bert": false,
28
+ "freeze_emo": false,
29
+ "freeze_style": false,
30
+ "freeze_decoder": false
31
+ },
32
+ "data": {
33
+ "use_jp_extra": true,
34
+ "training_files": "Data\\Matsumura_v1\\train.list",
35
+ "validation_files": "Data\\Matsumura_v1\\val.list",
36
+ "max_wav_value": 32768.0,
37
+ "sampling_rate": 44100,
38
+ "filter_length": 2048,
39
+ "hop_length": 512,
40
+ "win_length": 2048,
41
+ "n_mel_channels": 128,
42
+ "mel_fmin": 0.0,
43
+ "mel_fmax": null,
44
+ "add_blank": true,
45
+ "n_speakers": 1,
46
+ "cleaned_text": true,
47
+ "spk2id": {
48
+ "Matsumura_v1": 0
49
+ },
50
+ "num_styles": 1,
51
+ "style2id": {
52
+ "Neutral": 0
53
+ }
54
+ },
55
+ "model": {
56
+ "use_spk_conditioned_encoder": true,
57
+ "use_noise_scaled_mas": true,
58
+ "use_mel_posterior_encoder": false,
59
+ "use_duration_discriminator": false,
60
+ "use_wavlm_discriminator": true,
61
+ "inter_channels": 192,
62
+ "hidden_channels": 192,
63
+ "filter_channels": 768,
64
+ "n_heads": 2,
65
+ "n_layers": 6,
66
+ "kernel_size": 3,
67
+ "p_dropout": 0.1,
68
+ "resblock": "1",
69
+ "resblock_kernel_sizes": [
70
+ 3,
71
+ 7,
72
+ 11
73
+ ],
74
+ "resblock_dilation_sizes": [
75
+ [
76
+ 1,
77
+ 3,
78
+ 5
79
+ ],
80
+ [
81
+ 1,
82
+ 3,
83
+ 5
84
+ ],
85
+ [
86
+ 1,
87
+ 3,
88
+ 5
89
+ ]
90
+ ],
91
+ "upsample_rates": [
92
+ 8,
93
+ 8,
94
+ 2,
95
+ 2,
96
+ 2
97
+ ],
98
+ "upsample_initial_channel": 512,
99
+ "upsample_kernel_sizes": [
100
+ 16,
101
+ 16,
102
+ 8,
103
+ 2,
104
+ 2
105
+ ],
106
+ "n_layers_q": 3,
107
+ "use_spectral_norm": false,
108
+ "gin_channels": 512,
109
+ "slm": {
110
+ "model": "./slm/wavlm-base-plus",
111
+ "sr": 16000,
112
+ "hidden": 768,
113
+ "nlayers": 13,
114
+ "initial_channel": 64
115
+ }
116
+ },
117
+ "version": "2.6.1-JP-Extra"
118
+ }
model_assets/Matsumura_v1/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07a2f3f00bfa62f96d861a75299eb81a6a3b57762bcd619c2c2f87caa61e786b
3
+ size 1152
model_assets/Matsumura_v1_jvnv-M2-jp/Matsumura_v1_jvnv-M2-jp.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c2c3fb01c7583fc71928c5012a2f1677ecd0b9fb0be459f7edb96bb1209df79
3
+ size 251150980
model_assets/Matsumura_v1_jvnv-M2-jp/config.json ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "Matsumura_v1_jvnv-M2-jp",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 100,
8
+ "learning_rate": 0.0001,
9
+ "betas": [
10
+ 0.8,
11
+ 0.99
12
+ ],
13
+ "eps": 1e-09,
14
+ "batch_size": 2,
15
+ "bf16_run": false,
16
+ "fp16_run": false,
17
+ "lr_decay": 0.99996,
18
+ "segment_size": 16384,
19
+ "init_lr_ratio": 1,
20
+ "warmup_epochs": 0,
21
+ "c_mel": 45,
22
+ "c_kl": 1.0,
23
+ "c_commit": 100,
24
+ "skip_optimizer": false,
25
+ "freeze_ZH_bert": false,
26
+ "freeze_JP_bert": false,
27
+ "freeze_EN_bert": false,
28
+ "freeze_emo": false,
29
+ "freeze_style": false,
30
+ "freeze_decoder": false
31
+ },
32
+ "data": {
33
+ "use_jp_extra": true,
34
+ "training_files": "Data\\Matsumura_v1\\train.list",
35
+ "validation_files": "Data\\Matsumura_v1\\val.list",
36
+ "max_wav_value": 32768.0,
37
+ "sampling_rate": 44100,
38
+ "filter_length": 2048,
39
+ "hop_length": 512,
40
+ "win_length": 2048,
41
+ "n_mel_channels": 128,
42
+ "mel_fmin": 0.0,
43
+ "mel_fmax": null,
44
+ "add_blank": true,
45
+ "n_speakers": 1,
46
+ "cleaned_text": true,
47
+ "spk2id": {
48
+ "Matsumura_v1": 0
49
+ },
50
+ "num_styles": 1,
51
+ "style2id": {
52
+ "Neutral": 0
53
+ }
54
+ },
55
+ "model": {
56
+ "use_spk_conditioned_encoder": true,
57
+ "use_noise_scaled_mas": true,
58
+ "use_mel_posterior_encoder": false,
59
+ "use_duration_discriminator": false,
60
+ "use_wavlm_discriminator": true,
61
+ "inter_channels": 192,
62
+ "hidden_channels": 192,
63
+ "filter_channels": 768,
64
+ "n_heads": 2,
65
+ "n_layers": 6,
66
+ "kernel_size": 3,
67
+ "p_dropout": 0.1,
68
+ "resblock": "1",
69
+ "resblock_kernel_sizes": [
70
+ 3,
71
+ 7,
72
+ 11
73
+ ],
74
+ "resblock_dilation_sizes": [
75
+ [
76
+ 1,
77
+ 3,
78
+ 5
79
+ ],
80
+ [
81
+ 1,
82
+ 3,
83
+ 5
84
+ ],
85
+ [
86
+ 1,
87
+ 3,
88
+ 5
89
+ ]
90
+ ],
91
+ "upsample_rates": [
92
+ 8,
93
+ 8,
94
+ 2,
95
+ 2,
96
+ 2
97
+ ],
98
+ "upsample_initial_channel": 512,
99
+ "upsample_kernel_sizes": [
100
+ 16,
101
+ 16,
102
+ 8,
103
+ 2,
104
+ 2
105
+ ],
106
+ "n_layers_q": 3,
107
+ "use_spectral_norm": false,
108
+ "gin_channels": 512,
109
+ "slm": {
110
+ "model": "./slm/wavlm-base-plus",
111
+ "sr": 16000,
112
+ "hidden": 768,
113
+ "nlayers": 13,
114
+ "initial_channel": 64
115
+ }
116
+ },
117
+ "version": "2.6.1-JP-Extra"
118
+ }
model_assets/Matsumura_v1_jvnv-M2-jp/recipe.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "method": "usual",
3
+ "model_a": "model_assets\\Matsumura_v1\\Matsumura_v1_e100_s1400.safetensors",
4
+ "model_b": "model_assets\\jvnv-M2-jp\\jvnv-M2-jp_e159_s17000.safetensors",
5
+ "voice_weight": 0.3,
6
+ "voice_pitch_weight": 0.3,
7
+ "speech_style_weight": 0.2,
8
+ "tempo_weight": 0.5,
9
+ "use_slerp_instead_of_lerp": false
10
+ }
model_assets/Matsumura_v1_jvnv-M2-jp/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31315469261b18cf0348d4ae1b9437a37b3b0871cc1e366454b32011eaa99ecd
3
+ size 1152
model_assets/NotAnimeJPManySpeaker/NotAnimeJPManySpeaker_e120_s22200.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37cc6704519dcf5968d1b2a0f25dd5dd2658fbc4f070cd931ba2784508907246
3
+ size 251159172
model_assets/NotAnimeJPManySpeaker/config.json ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 4000,
5
+ "seed": 42,
6
+ "epochs": 120,
7
+ "learning_rate": 0.0001,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 17,
14
+ "bf16_run": false,
15
+ "fp16_run": false,
16
+ "lr_decay": 0.99996,
17
+ "segment_size": 16384,
18
+ "init_lr_ratio": 1,
19
+ "warmup_epochs": 0,
20
+ "c_mel": 45,
21
+ "c_kl": 1.0,
22
+ "c_commit": 100,
23
+ "skip_optimizer": true,
24
+ "freeze_ZH_bert": false,
25
+ "freeze_JP_bert": false,
26
+ "freeze_EN_bert": false,
27
+ "freeze_emo": false,
28
+ "freeze_style": false
29
+ },
30
+ "data": {
31
+ "use_jp_extra": true,
32
+ "training_files": "",
33
+ "validation_files": "",
34
+ "max_wav_value": 32768.0,
35
+ "sampling_rate": 44100,
36
+ "filter_length": 2048,
37
+ "hop_length": 512,
38
+ "win_length": 2048,
39
+ "n_mel_channels": 128,
40
+ "mel_fmin": 0.0,
41
+ "mel_fmax": null,
42
+ "add_blank": true,
43
+ "n_speakers": 5,
44
+ "cleaned_text": true,
45
+ "spk2id": {
46
+ "NotAnime-JP-age20female-amazinGood": 0,
47
+ "NotAnime-JP-age20female-calmCloud": 1,
48
+ "NotAnime-JP-age20female-coolcute": 2,
49
+ "NotAnime-JP-age20female-fineCrystal": 3,
50
+ "NotAnime-JP-age20male-lightFire": 4
51
+ },
52
+ "num_styles": 25,
53
+ "style2id": {
54
+ "Neutral": 0,
55
+ "amazinGood(down)": 1,
56
+ "amazinGood(lol)": 2,
57
+ "amazinGood(onmygod)": 3,
58
+ "amazinGood(normal)": 4,
59
+ "calmCloud(lol)": 5,
60
+ "calmCloud(question)": 6,
61
+ "calmCloud(down)": 7,
62
+ "calmCloud(hate)": 8,
63
+ "calmCloud(ohmygod)": 9,
64
+ "calmCloud(normal)": 10,
65
+ "coolcute(onmygod)": 11,
66
+ "coolcute(normal)": 12,
67
+ "coolcute(fine)": 13,
68
+ "coolcute(sad)": 14,
69
+ "fineCrystal(fine)": 15,
70
+ "fineCrystal(ohmygod)": 16,
71
+ "fineCrystal(veryfine)": 17,
72
+ "fineCrystal(normal)": 18,
73
+ "fineCrystal(sad)": 19,
74
+ "lightFire(question)": 20,
75
+ "lightFire(hello)": 21,
76
+ "lightFire(normal)": 22,
77
+ "lightFire(strong)": 23,
78
+ "lightFire(lol)": 24
79
+ }
80
+ },
81
+ "model": {
82
+ "use_spk_conditioned_encoder": true,
83
+ "use_noise_scaled_mas": true,
84
+ "use_mel_posterior_encoder": false,
85
+ "use_duration_discriminator": false,
86
+ "use_wavlm_discriminator": true,
87
+ "inter_channels": 192,
88
+ "hidden_channels": 192,
89
+ "filter_channels": 768,
90
+ "n_heads": 2,
91
+ "n_layers": 6,
92
+ "kernel_size": 3,
93
+ "p_dropout": 0.1,
94
+ "resblock": "1",
95
+ "resblock_kernel_sizes": [
96
+ 3,
97
+ 7,
98
+ 11
99
+ ],
100
+ "resblock_dilation_sizes": [
101
+ [
102
+ 1,
103
+ 3,
104
+ 5
105
+ ],
106
+ [
107
+ 1,
108
+ 3,
109
+ 5
110
+ ],
111
+ [
112
+ 1,
113
+ 3,
114
+ 5
115
+ ]
116
+ ],
117
+ "upsample_rates": [
118
+ 8,
119
+ 8,
120
+ 2,
121
+ 2,
122
+ 2
123
+ ],
124
+ "upsample_initial_channel": 512,
125
+ "upsample_kernel_sizes": [
126
+ 16,
127
+ 16,
128
+ 8,
129
+ 2,
130
+ 2
131
+ ],
132
+ "n_layers_q": 3,
133
+ "use_spectral_norm": false,
134
+ "gin_channels": 512,
135
+ "slm": {
136
+ "model": "./slm/wavlm-base-plus",
137
+ "sr": 16000,
138
+ "hidden": 768,
139
+ "nlayers": 13,
140
+ "initial_channel": 64
141
+ }
142
+ },
143
+ "version": "2.1-JP-Extra",
144
+ "model_name": "NotAnimeJPManySpeaker"
145
+ }
model_assets/NotAnimeJPManySpeaker/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28823bf64b422f055e6fa0ca6c5e8ec2d8669e298e181812502581b79d577fc8
3
+ size 25728