Changes

              
    

          
          
        
        

            
f
1
{
f
1
{
            
2
  "access_rights": "",
2
  "access_rights": "",
            
3
  "author": "Ziyue Jiang",
3
  "author": "Ziyue Jiang",
            
4
  "author_email": "",
4
  "author_email": "",
            
5
  "citation": [],
5
  "citation": [],
            
6
  "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700",
6
  "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700",
            
7
  "defined_in": "https://doi.org/10.48550/arXiv.2306.03509",
7
  "defined_in": "https://doi.org/10.48550/arXiv.2306.03509",
            
8
  "doi": "10.57702/0yzgeweh",
8
  "doi": "10.57702/0yzgeweh",
            
n
9
  "doi_date_published": null,
n
9
  "doi_date_published": "2024-12-16",
            
10
  "doi_publisher": "TIB",
10
  "doi_publisher": "TIB",
            
n
11
  "doi_status": false,
n
11
  "doi_status": true,
            
12
  "domain": "https://service.tib.eu/ldmservice",
12
  "domain": "https://service.tib.eu/ldmservice",
            
13
  "extra_authors": [
13
  "extra_authors": [
            
14
    {
14
    {
            
15
      "extra_author": "Yi Ren",
15
      "extra_author": "Yi Ren",
            
16
      "orcid": ""
16
      "orcid": ""
            
17
    },
17
    },
            
18
    {
18
    {
            
19
      "extra_author": "Zhenhui Ye",
19
      "extra_author": "Zhenhui Ye",
            
20
      "orcid": ""
20
      "orcid": ""
            
21
    },
21
    },
            
22
    {
22
    {
            
23
      "extra_author": "Jinglin Liu",
23
      "extra_author": "Jinglin Liu",
            
24
      "orcid": ""
24
      "orcid": ""
            
25
    },
25
    },
            
26
    {
26
    {
            
27
      "extra_author": "Chen Zhang",
27
      "extra_author": "Chen Zhang",
            
28
      "orcid": ""
28
      "orcid": ""
            
29
    },
29
    },
            
30
    {
30
    {
            
31
      "extra_author": "Qian Yang",
31
      "extra_author": "Qian Yang",
            
32
      "orcid": ""
32
      "orcid": ""
            
33
    },
33
    },
            
34
    {
34
    {
            
35
      "extra_author": "Shengpeng Ji",
35
      "extra_author": "Shengpeng Ji",
            
36
      "orcid": ""
36
      "orcid": ""
            
37
    },
37
    },
            
38
    {
38
    {
            
39
      "extra_author": "Rongjie Huang",
39
      "extra_author": "Rongjie Huang",
            
40
      "orcid": ""
40
      "orcid": ""
            
41
    },
41
    },
            
42
    {
42
    {
            
43
      "extra_author": "Chunfeng Wang",
43
      "extra_author": "Chunfeng Wang",
            
44
      "orcid": ""
44
      "orcid": ""
            
45
    },
45
    },
            
46
    {
46
    {
            
47
      "extra_author": "Xiang Yin",
47
      "extra_author": "Xiang Yin",
            
48
      "orcid": ""
48
      "orcid": ""
            
49
    },
49
    },
            
50
    {
50
    {
            
51
      "extra_author": "Zejun Ma",
51
      "extra_author": "Zejun Ma",
            
52
      "orcid": ""
52
      "orcid": ""
            
53
    },
53
    },
            
54
    {
54
    {
            
55
      "extra_author": "Zhou Zhao",
55
      "extra_author": "Zhou Zhao",
            
56
      "orcid": ""
56
      "orcid": ""
            
57
    }
57
    }
            
58
  ],
58
  ],
            
59
  "groups": [
59
  "groups": [
            
60
    {
60
    {
            
61
      "description": "",
61
      "description": "",
            
62
      "display_name": "Speech Synthesis",
62
      "display_name": "Speech Synthesis",
            
63
      "id": "3b8f8411-cbd1-410f-92f8-5ecbc4095d46",
63
      "id": "3b8f8411-cbd1-410f-92f8-5ecbc4095d46",
            
64
      "image_display_url": "",
64
      "image_display_url": "",
            
65
      "name": "speech-synthesis",
65
      "name": "speech-synthesis",
            
66
      "title": "Speech Synthesis"
66
      "title": "Speech Synthesis"
            
67
    },
67
    },
            
68
    {
68
    {
            
69
      "description": "",
69
      "description": "",
            
70
      "display_name": "Text-to-Speech",
70
      "display_name": "Text-to-Speech",
            
71
      "id": "486fb9b3-8453-4c88-bd0d-b9d59c07299c",
71
      "id": "486fb9b3-8453-4c88-bd0d-b9d59c07299c",
            
72
      "image_display_url": "",
72
      "image_display_url": "",
            
73
      "name": "text-to-speech",
73
      "name": "text-to-speech",
            
74
      "title": "Text-to-Speech"
74
      "title": "Text-to-Speech"
            
75
    }
75
    }
            
76
  ],
76
  ],
            
77
  "id": "76610fc5-4116-47ee-861e-02dbd990afda",
77
  "id": "76610fc5-4116-47ee-861e-02dbd990afda",
            
78
  "isopen": false,
78
  "isopen": false,
            
79
  "landing_page": "https://mega-tts.github.io/demo-page",
79
  "landing_page": "https://mega-tts.github.io/demo-page",
            
80
  "license_title": null,
80
  "license_title": null,
            
81
  "link_orkg": "",
81
  "link_orkg": "",
            
82
  "metadata_created": "2024-12-16T18:34:07.249430",
82
  "metadata_created": "2024-12-16T18:34:07.249430",
            
n
83
  "metadata_modified": "2024-12-16T18:34:07.249436",
n
83
  "metadata_modified": "2024-12-16T18:34:07.608347",
            
84
  "name": 
84
  "name": 
            
85
tts--zero-shot-text-to-speech-at-scale-with-intrinsic-inductive-bias",
85
tts--zero-shot-text-to-speech-at-scale-with-intrinsic-inductive-bias",
            
86
  "notes": "Scaling text-to-speech to a large and wild dataset has 
86
  "notes": "Scaling text-to-speech to a large and wild dataset has 
            
87
been proven to be highly effective in achieving timbre and speech 
87
been proven to be highly effective in achieving timbre and speech 
            
88
style generalization, particularly in zero-shot TTS. However, previous 
88
style generalization, particularly in zero-shot TTS. However, previous 
            
89
works usually encode speech into latent using audio codec and use 
89
works usually encode speech into latent using audio codec and use 
            
90
autoregressive language models or diffusion models to generate it, 
90
autoregressive language models or diffusion models to generate it, 
            
91
which ignores the intrinsic nature of speech and may lead to inferior 
91
which ignores the intrinsic nature of speech and may lead to inferior 
            
92
or uncontrollable results.",
92
or uncontrollable results.",
            
n
93
  "num_resources": 0,
n
93
  "num_resources": 1,
            
94
  "num_tags": 3,
94
  "num_tags": 3,
            
95
  "organization": {
95
  "organization": {
            
96
    "approval_status": "approved",
96
    "approval_status": "approved",
            
97
    "created": "2024-11-25T12:11:38.292601",
97
    "created": "2024-11-25T12:11:38.292601",
            
98
    "description": "",
98
    "description": "",
            
99
    "id": "079d46db-32df-4b48-91f3-0a8bc8f69559",
99
    "id": "079d46db-32df-4b48-91f3-0a8bc8f69559",
            
100
    "image_url": "",
100
    "image_url": "",
            
101
    "is_organization": true,
101
    "is_organization": true,
            
102
    "name": "no-organization",
102
    "name": "no-organization",
            
103
    "state": "active",
103
    "state": "active",
            
104
    "title": "No Organization",
104
    "title": "No Organization",
            
105
    "type": "organization"
105
    "type": "organization"
            
106
  },
106
  },
            
107
  "owner_org": "079d46db-32df-4b48-91f3-0a8bc8f69559",
107
  "owner_org": "079d46db-32df-4b48-91f3-0a8bc8f69559",
            
108
  "private": false,
108
  "private": false,
            
109
  "relationships_as_object": [],
109
  "relationships_as_object": [],
            
110
  "relationships_as_subject": [],
110
  "relationships_as_subject": [],
            
t
111
  "resources": [],
t
111
  "resources": [
            
112
    {
            
113
      "cache_last_updated": null,
            
114
      "cache_url": null,
            
115
      "created": "2024-12-16T18:25:35",
            
116
      "data": [
            
117
        "dcterms:title",
            
118
        "dcterms:accessRights",
            
119
        "dcterms:creator",
            
120
        "dcterms:description",
            
121
        "dcterms:issued",
            
122
        "dcterms:language",
            
123
        "dcterms:identifier",
            
124
        "dcat:theme",
            
125
        "dcterms:type",
            
126
        "dcat:keyword",
            
127
        "dcat:landingPage",
            
128
        "dcterms:hasVersion",
            
129
        "dcterms:format",
            
130
        "mls:task",
            
131
        "datacite:isDescribedBy"
            
132
      ],
            
133
      "description": "The json representation of the dataset with its 
            
134
distributions based on DCAT.",
            
135
      "format": "JSON",
            
136
      "hash": "",
            
137
      "id": "386b0c58-3dcb-4ba8-bcb2-30624aa90191",
            
138
      "last_modified": "2024-12-16T18:34:07.600252",
            
139
      "metadata_modified": "2024-12-16T18:34:07.612963",
            
140
      "mimetype": "application/json",
            
141
      "mimetype_inner": null,
            
142
      "name": "Original Metadata",
            
143
      "package_id": "76610fc5-4116-47ee-861e-02dbd990afda",
            
144
      "position": 0,
            
145
      "resource_type": null,
            
146
      "size": 1242,
            
147
      "state": "active",
            
148
      "url": 
            
149
resource/386b0c58-3dcb-4ba8-bcb2-30624aa90191/download/metadata.json",
            
150
      "url_type": "upload"
            
151
    }
            
152
  ],
            
112
  "services_used_list": "",
153
  "services_used_list": "",
            
113
  "state": "active",
154
  "state": "active",
            
114
  "tags": [
155
  "tags": [
            
115
    {
156
    {
            
116
      "display_name": "intrinsic inductive bias",
157
      "display_name": "intrinsic inductive bias",
            
117
      "id": "6faa5efc-5519-4087-ba36-51af7fd0834f",
158
      "id": "6faa5efc-5519-4087-ba36-51af7fd0834f",
            
118
      "name": "intrinsic inductive bias",
159
      "name": "intrinsic inductive bias",
            
119
      "state": "active",
160
      "state": "active",
            
120
      "vocabulary_id": null
161
      "vocabulary_id": null
            
121
    },
162
    },
            
122
    {
163
    {
            
123
      "display_name": "speech style generalization",
164
      "display_name": "speech style generalization",
            
124
      "id": "ddd19543-2a01-4be1-953a-79f0d5d409b7",
165
      "id": "ddd19543-2a01-4be1-953a-79f0d5d409b7",
            
125
      "name": "speech style generalization",
166
      "name": "speech style generalization",
            
126
      "state": "active",
167
      "state": "active",
            
127
      "vocabulary_id": null
168
      "vocabulary_id": null
            
128
    },
169
    },
            
129
    {
170
    {
            
130
      "display_name": "zero-shot text-to-speech",
171
      "display_name": "zero-shot text-to-speech",
            
131
      "id": "2d2499e1-498e-4b56-8bd4-f9da097aad3b",
172
      "id": "2d2499e1-498e-4b56-8bd4-f9da097aad3b",
            
132
      "name": "zero-shot text-to-speech",
173
      "name": "zero-shot text-to-speech",
            
133
      "state": "active",
174
      "state": "active",
            
134
      "vocabulary_id": null
175
      "vocabulary_id": null
            
135
    }
176
    }
            
136
  ],
177
  ],
            
137
  "title": "Mega-TTS: Zero-Shot Text-to-Speech at Scale with Intrinsic 
178
  "title": "Mega-TTS: Zero-Shot Text-to-Speech at Scale with Intrinsic 
            
138
Inductive Bias",
179
Inductive Bias",
            
139
  "type": "dataset",
180
  "type": "dataset",
            
140
  "version": ""
181
  "version": ""
            
141
}
182
}