Hugging Face's logo

hexgrad
/
Kokoro-82M

Text-to-Speech
English
Model card Files
xet
hexgrad commited on
Commit
8bc190a
ยท
1 Parent(s): 85cb1ba

Upload 2 files

Browse files
Files changed (2) hide show
  1. README.md +5 -1
  2. VOICES.md +55 -31
README.md CHANGED
@@ -24,7 +24,7 @@ pipeline_tag: text-to-speech
24
 
25
  | Model | Published | Training Data | Compute (A100 80GB) | Langs & Voices | SHA256 |
26
  | ----- | --------- | ------------- | ------------------- | -------------- | ------ |
27
- | **v1.0** | **2025 Jan 27** | **Few hundred hrs** | **$1000 for 1000 hrs** | [**6 & 47**](https://huggingface.co/hexgrad/Kokoro-82M/blob/main/VOICES.md) | `496dba11` |
28
  | [v0.19](https://huggingface.co/hexgrad/kLegacy/tree/main/v0.19) | 2024 Dec 25 | <100 hrs | $400 for 500 hrs | 1 & 10 | `3b0c392f` |
29
 
30
  ### Usage
@@ -37,9 +37,11 @@ You can run this cell on [Google Colab](https://colab.research.google.com/).
37
  !pip install -q kokoro>=0.3.1 soundfile
38
  # 2๏ธโƒฃ Install espeak, used for English OOD fallback and some non-English languages
39
  !apt-get -qq -y install espeak-ng > /dev/null 2>&1
 
40
  # ๐Ÿ‡ซ๐Ÿ‡ท 'f' => French fr-fr
41
  # ๐Ÿ‡ฎ๐Ÿ‡ณ 'h' => Hindi hi
42
  # ๐Ÿ‡ฎ๐Ÿ‡น 'i' => Italian it
 
43
 
44
  # 3๏ธโƒฃ Initalize a pipeline
45
  from kokoro import KPipeline
@@ -60,9 +62,11 @@ These were to have an enormous impact, not only because they were associated wit
60
  '''
61
  # text = 'ใ€Œใ‚‚ใ—ใŠใ‚ŒใŒใŸใ ๅถ็„ถใ€ใใ—ใฆใ“ใ†ใ—ใ‚ˆใ†ใจใ„ใ†ใคใ‚‚ใ‚Šใงใชใใ“ใ“ใซ็ซ‹ใฃใฆใ„ใ‚‹ใฎใชใ‚‰ใ€ใกใ‚‡ใฃใจใฐใ‹ใ‚Š็ตถๆœ›ใ™ใ‚‹ใจใ“ใ‚ใ ใชใ€ใจใ€ใใ‚“ใชใ“ใจใŒๅฝผใฎ้ ญใซๆ€ใ„ๆตฎใ‹ใ‚“ใ ใ€‚'
62
  # text = 'ไธญๅœ‹ไบบๆฐ‘ไธไฟก้‚ชไนŸไธๆ€•้‚ช๏ผŒไธๆƒนไบ‹ไนŸไธๆ€•ไบ‹๏ผŒไปปไฝ•ๅค–ๅœ‹ไธ่ฆๆŒ‡ๆœ›ๆˆ‘ๅ€‘ๆœƒๆ‹ฟ่‡ชๅทฑ็š„ๆ ธๅฟƒๅˆฉ็›Šๅšไบคๆ˜“๏ผŒไธ่ฆๆŒ‡ๆœ›ๆˆ‘ๅ€‘ๆœƒๅžไธ‹ๆๅฎณๆˆ‘ๅœ‹ไธปๆฌŠใ€ๅฎ‰ๅ…จใ€็™ผๅฑ•ๅˆฉ็›Š็š„่‹ฆๆžœ๏ผ'
 
63
  # text = 'Le dromadaire resplendissant dรฉambulait tranquillement dans les mรฉandres en mastiquant de petites feuilles vernissรฉes.'
64
  # text = 'เคŸเฅเคฐเคพเค‚เคธเคชเฅ‹เคฐเฅเคŸเคฐเฅ‹เค‚ เค•เฅ€ เคนเคกเคผเคคเคพเคฒ เคฒเค—เคพเคคเคพเคฐ เคชเคพเค‚เคšเคตเฅ‡เค‚ เคฆเคฟเคจ เคœเคพเคฐเฅ€, เคฆเคฟเคธเค‚เคฌเคฐ เคธเฅ‡ เค‡เคฒเฅ‡เค•เฅเคŸเฅเคฐเฅ‰เคจเคฟเค• เคŸเฅ‹เคฒ เค•เคฒเฅ‡เค•เฅเคถเคจเคฒ เคธเคฟเคธเฅเคŸเคฎ'
65
  # text = "Allora cominciava l'insonnia, o un dormiveglia peggiore dell'insonnia, che talvolta assumeva i caratteri dell'incubo."
 
66
 
67
  # 4๏ธโƒฃ Generate, display, and save audio files in a loop.
68
  generator = pipeline(
 
24
 
25
  | Model | Published | Training Data | Compute (A100 80GB) | Langs & Voices | SHA256 |
26
  | ----- | --------- | ------------- | ------------------- | -------------- | ------ |
27
+ | **v1.0** | **2025 Jan 27** | **Few hundred hrs** | **$1000 for 1000 hrs** | [**8 & 53**](https://huggingface.co/hexgrad/Kokoro-82M/blob/main/VOICES.md) | `496dba11` |
28
  | [v0.19](https://huggingface.co/hexgrad/kLegacy/tree/main/v0.19) | 2024 Dec 25 | <100 hrs | $400 for 500 hrs | 1 & 10 | `3b0c392f` |
29
 
30
  ### Usage
 
37
  !pip install -q kokoro>=0.3.1 soundfile
38
  # 2๏ธโƒฃ Install espeak, used for English OOD fallback and some non-English languages
39
  !apt-get -qq -y install espeak-ng > /dev/null 2>&1
40
+ # ๐Ÿ‡ช๐Ÿ‡ธ 'e' => Spanish es
41
  # ๐Ÿ‡ซ๐Ÿ‡ท 'f' => French fr-fr
42
  # ๐Ÿ‡ฎ๐Ÿ‡ณ 'h' => Hindi hi
43
  # ๐Ÿ‡ฎ๐Ÿ‡น 'i' => Italian it
44
+ # ๐Ÿ‡ง๐Ÿ‡ท 'p' => Brazilian Portuguese pt-br
45
 
46
  # 3๏ธโƒฃ Initalize a pipeline
47
  from kokoro import KPipeline
 
62
  '''
63
  # text = 'ใ€Œใ‚‚ใ—ใŠใ‚ŒใŒใŸใ ๅถ็„ถใ€ใใ—ใฆใ“ใ†ใ—ใ‚ˆใ†ใจใ„ใ†ใคใ‚‚ใ‚Šใงใชใใ“ใ“ใซ็ซ‹ใฃใฆใ„ใ‚‹ใฎใชใ‚‰ใ€ใกใ‚‡ใฃใจใฐใ‹ใ‚Š็ตถๆœ›ใ™ใ‚‹ใจใ“ใ‚ใ ใชใ€ใจใ€ใใ‚“ใชใ“ใจใŒๅฝผใฎ้ ญใซๆ€ใ„ๆตฎใ‹ใ‚“ใ ใ€‚'
64
  # text = 'ไธญๅœ‹ไบบๆฐ‘ไธไฟก้‚ชไนŸไธๆ€•้‚ช๏ผŒไธๆƒนไบ‹ไนŸไธๆ€•ไบ‹๏ผŒไปปไฝ•ๅค–ๅœ‹ไธ่ฆๆŒ‡ๆœ›ๆˆ‘ๅ€‘ๆœƒๆ‹ฟ่‡ชๅทฑ็š„ๆ ธๅฟƒๅˆฉ็›Šๅšไบคๆ˜“๏ผŒไธ่ฆๆŒ‡ๆœ›ๆˆ‘ๅ€‘ๆœƒๅžไธ‹ๆๅฎณๆˆ‘ๅœ‹ไธปๆฌŠใ€ๅฎ‰ๅ…จใ€็™ผๅฑ•ๅˆฉ็›Š็š„่‹ฆๆžœ๏ผ'
65
+ # text = 'Los partidos polรญticos tradicionales compiten con los populismos y los movimientos asamblearios.'
66
  # text = 'Le dromadaire resplendissant dรฉambulait tranquillement dans les mรฉandres en mastiquant de petites feuilles vernissรฉes.'
67
  # text = 'เคŸเฅเคฐเคพเค‚เคธเคชเฅ‹เคฐเฅเคŸเคฐเฅ‹เค‚ เค•เฅ€ เคนเคกเคผเคคเคพเคฒ เคฒเค—เคพเคคเคพเคฐ เคชเคพเค‚เคšเคตเฅ‡เค‚ เคฆเคฟเคจ เคœเคพเคฐเฅ€, เคฆเคฟเคธเค‚เคฌเคฐ เคธเฅ‡ เค‡เคฒเฅ‡เค•เฅเคŸเฅเคฐเฅ‰เคจเคฟเค• เคŸเฅ‹เคฒ เค•เคฒเฅ‡เค•เฅเคถเคจเคฒ เคธเคฟเคธเฅเคŸเคฎ'
68
  # text = "Allora cominciava l'insonnia, o un dormiveglia peggiore dell'insonnia, che talvolta assumeva i caratteri dell'incubo."
69
+ # text = 'Elabora relatรณrios de acompanhamento cronolรณgico para as diferentes unidades do Departamento que propรตem contratos.'
70
 
71
  # 4๏ธโƒฃ Generate, display, and save audio files in a loop.
72
  generator = pipeline(
VOICES.md CHANGED
@@ -2,11 +2,13 @@
2
 
3
  - ๐Ÿ‡บ๐Ÿ‡ธ [American English](#american-english): 10F 9M
4
  - ๐Ÿ‡ฌ๐Ÿ‡ง [British English](#british-english): 4F 4M
 
 
 
5
  - ๐Ÿ‡ซ๐Ÿ‡ท [French](#french): 1F
6
  - ๐Ÿ‡ฎ๐Ÿ‡ณ [Hindi](#hindi): 2F 2M
7
  - ๐Ÿ‡ฎ๐Ÿ‡น [Italian](#italian): 1F 1M
8
- - ๐Ÿ‡ฏ๐Ÿ‡ต [Japanese](#japanese): 4F 1M
9
- - ๐Ÿ‡จ๐Ÿ‡ณ [Mandarin Chinese](#mandarin-chinese): 4F 4M
10
 
11
  For each voice, the given grades are intended to be estimates of the **quality and quantity** of its associated training data, both of which impact overall inference quality.
12
 
@@ -27,7 +29,7 @@ Most voices perform best on a "goldilocks range" of 100-200 tokens out of ~500 p
27
  - 10 hours <= **HH hours** < 100 hours
28
  - 1 hour <= H hours < 10 hours
29
  - 10 minutes <= MM minutes < 100 minutes
30
- - 1 minute <= _M minutes_ < 10 minutes ๐Ÿค
31
 
32
  ### American English
33
 
@@ -45,7 +47,7 @@ Most voices perform best on a "goldilocks range" of 100-200 tokens out of ~500 p
45
  | af_nova | ๐Ÿšบ | B | MM minutes | C | `e0233676` |
46
  | af_river | ๐Ÿšบ | C | MM minutes | D | `e149459b` |
47
  | af_sarah | ๐Ÿšบ | B | H hours | C+ | `49bd364e` |
48
- | af_sky | ๐Ÿšบ๐Ÿค | B | _M minutes_ | C- | `c799548a` |
49
  | am_adam | ๐Ÿšน | D | H hours | F+ | `ced7e284` |
50
  | am_echo | ๐Ÿšน | C | MM minutes | D | `8bcfdc85` |
51
  | am_eric | ๐Ÿšน | C | MM minutes | D | `ada66f0e` |
@@ -54,7 +56,7 @@ Most voices perform best on a "goldilocks range" of 100-200 tokens out of ~500 p
54
  | am_michael | ๐Ÿšน | B | H hours | C+ | `9a443b79` |
55
  | am_onyx | ๐Ÿšน | C | MM minutes | D | `e8452be1` |
56
  | am_puck | ๐Ÿšน | B | H hours | C+ | `dd1d8973` |
57
- | am_santa | ๐Ÿšน๐Ÿค | C | _M minutes_ | D- | `7f2f7582` |
58
 
59
  ### British English
60
 
@@ -72,6 +74,46 @@ Most voices perform best on a "goldilocks range" of 100-200 tokens out of ~500 p
72
  | bm_george | ๐Ÿšน | B | MM minutes | C | `f1bc8122` |
73
  | bm_lewis | ๐Ÿšน | C | H hours | D+ | `b5204750` |
74
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
75
  ### French
76
 
77
  - `lang_code='f'` in [`misaki[en]`](https://github.com/hexgrad/misaki)
@@ -106,31 +148,13 @@ Most voices perform best on a "goldilocks range" of 100-200 tokens out of ~500 p
106
  | if_sara | ๐Ÿšบ | B | MM minutes | C | `6c0b253b` |
107
  | im_nicola | ๐Ÿšน | B | MM minutes | C | `234ed066` |
108
 
109
- ### Japanese
110
 
111
- - `lang_code='j'` in [`misaki[ja]`](https://github.com/hexgrad/misaki)
112
- - Total Japanese training data: H hours
113
 
114
- | Name | Traits | Target Quality | Training Duration | Overall Grade | SHA256 | CC BY |
115
- | ---- | ------ | -------------- | ----------------- | ------------- | ------ | ----- |
116
- | jf_alpha | ๐Ÿšบ | B | H hours | C+ | `1bf4c9dc` | |
117
- | jf_gongitsune | ๐Ÿšบ | B | MM minutes | C | `1b171917` | [gongitsune](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__gongitsune.txt) |
118
- | jf_nezumi | ๐Ÿšบ๐Ÿค | B | _M minutes_ | C- | `d83f007a` | [nezuminoyomeiri](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__nezuminoyomeiri.txt) |
119
- | jf_tebukuro | ๐Ÿšบ | B | MM minutes | C | `0d691790` | [tebukurowokaini](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__tebukurowokaini.txt) |
120
- | jm_kumo | ๐Ÿšน๐Ÿค | B | _M minutes_ | C- | `98340afd` | [kumonoito](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__kumonoito.txt) |
121
-
122
- ### Mandarin Chinese
123
-
124
- - `lang_code='z'` in [`misaki[zh]`](https://github.com/hexgrad/misaki)
125
- - Total Mandarin Chinese training data: H hours
126
-
127
- | Name | Traits | Target Quality | Training Duration | Overall Grade | SHA256 |
128
- | ---- | ------ | -------------- | ----------------- | ------------- | ------ |
129
- | zf_xiaobei | ๐Ÿšบ | C | MM minutes | D | `9b76be63` |
130
- | zf_xiaoni | ๐Ÿšบ | C | MM minutes | D | `95b49f16` |
131
- | zf_xiaoxiao | ๐Ÿšบ | C | MM minutes | D | `cfaf6f2d` |
132
- | zf_xiaoyi | ๐Ÿšบ | C | MM minutes | D | `b5235dba` |
133
- | zm_yunjian | ๐Ÿšน | C | MM minutes | D | `76cbf8ba` |
134
- | zm_yunxi | ๐Ÿšน | C | MM minutes | D | `dbe6e1ce` |
135
- | zm_yunxia | ๐Ÿšน | C | MM minutes | D | `bb2b03b0` |
136
- | zm_yunyang | ๐Ÿšน | C | MM minutes | D | `5238ac22` |
 
2
 
3
  - ๐Ÿ‡บ๐Ÿ‡ธ [American English](#american-english): 10F 9M
4
  - ๐Ÿ‡ฌ๐Ÿ‡ง [British English](#british-english): 4F 4M
5
+ - ๐Ÿ‡ฏ๐Ÿ‡ต [Japanese](#japanese): 4F 1M
6
+ - ๐Ÿ‡จ๐Ÿ‡ณ [Mandarin Chinese](#mandarin-chinese): 4F 4M
7
+ - ๐Ÿ‡ช๐Ÿ‡ธ [Spanish](#spanish): 1F 2M
8
  - ๐Ÿ‡ซ๐Ÿ‡ท [French](#french): 1F
9
  - ๐Ÿ‡ฎ๐Ÿ‡ณ [Hindi](#hindi): 2F 2M
10
  - ๐Ÿ‡ฎ๐Ÿ‡น [Italian](#italian): 1F 1M
11
+ - ๐Ÿ‡ง๐Ÿ‡ท [Brazilian Portuguese](#brazilian-portuguese): 1F 2M
 
12
 
13
  For each voice, the given grades are intended to be estimates of the **quality and quantity** of its associated training data, both of which impact overall inference quality.
14
 
 
29
  - 10 hours <= **HH hours** < 100 hours
30
  - 1 hour <= H hours < 10 hours
31
  - 10 minutes <= MM minutes < 100 minutes
32
+ - 1 minute <= _M minutes_ ๐Ÿค < 10 minutes
33
 
34
  ### American English
35
 
 
47
  | af_nova | ๐Ÿšบ | B | MM minutes | C | `e0233676` |
48
  | af_river | ๐Ÿšบ | C | MM minutes | D | `e149459b` |
49
  | af_sarah | ๐Ÿšบ | B | H hours | C+ | `49bd364e` |
50
+ | af_sky | ๐Ÿšบ | B | _M minutes_ ๐Ÿค | C- | `c799548a` |
51
  | am_adam | ๐Ÿšน | D | H hours | F+ | `ced7e284` |
52
  | am_echo | ๐Ÿšน | C | MM minutes | D | `8bcfdc85` |
53
  | am_eric | ๐Ÿšน | C | MM minutes | D | `ada66f0e` |
 
56
  | am_michael | ๐Ÿšน | B | H hours | C+ | `9a443b79` |
57
  | am_onyx | ๐Ÿšน | C | MM minutes | D | `e8452be1` |
58
  | am_puck | ๐Ÿšน | B | H hours | C+ | `dd1d8973` |
59
+ | am_santa | ๐Ÿšน | C | _M minutes_ ๐Ÿค | D- | `7f2f7582` |
60
 
61
  ### British English
62
 
 
74
  | bm_george | ๐Ÿšน | B | MM minutes | C | `f1bc8122` |
75
  | bm_lewis | ๐Ÿšน | C | H hours | D+ | `b5204750` |
76
 
77
+ ### Japanese
78
+
79
+ - `lang_code='j'` in [`misaki[ja]`](https://github.com/hexgrad/misaki)
80
+ - Total Japanese training data: H hours
81
+
82
+ | Name | Traits | Target Quality | Training Duration | Overall Grade | SHA256 | CC BY |
83
+ | ---- | ------ | -------------- | ----------------- | ------------- | ------ | ----- |
84
+ | jf_alpha | ๐Ÿšบ | B | H hours | C+ | `1bf4c9dc` | |
85
+ | jf_gongitsune | ๐Ÿšบ | B | MM minutes | C | `1b171917` | [gongitsune](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__gongitsune.txt) |
86
+ | jf_nezumi | ๐Ÿšบ | B | _M minutes_ ๐Ÿค | C- | `d83f007a` | [nezuminoyomeiri](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__nezuminoyomeiri.txt) |
87
+ | jf_tebukuro | ๐Ÿšบ | B | MM minutes | C | `0d691790` | [tebukurowokaini](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__tebukurowokaini.txt) |
88
+ | jm_kumo | ๐Ÿšน | B | _M minutes_ ๐Ÿค | C- | `98340afd` | [kumonoito](https://github.com/koniwa/koniwa/blob/master/source/tnc/tnc__kumonoito.txt) |
89
+
90
+ ### Mandarin Chinese
91
+
92
+ - `lang_code='z'` in [`misaki[zh]`](https://github.com/hexgrad/misaki)
93
+ - Total Mandarin Chinese training data: H hours
94
+
95
+ | Name | Traits | Target Quality | Training Duration | Overall Grade | SHA256 |
96
+ | ---- | ------ | -------------- | ----------------- | ------------- | ------ |
97
+ | zf_xiaobei | ๐Ÿšบ | C | MM minutes | D | `9b76be63` |
98
+ | zf_xiaoni | ๐Ÿšบ | C | MM minutes | D | `95b49f16` |
99
+ | zf_xiaoxiao | ๐Ÿšบ | C | MM minutes | D | `cfaf6f2d` |
100
+ | zf_xiaoyi | ๐Ÿšบ | C | MM minutes | D | `b5235dba` |
101
+ | zm_yunjian | ๐Ÿšน | C | MM minutes | D | `76cbf8ba` |
102
+ | zm_yunxi | ๐Ÿšน | C | MM minutes | D | `dbe6e1ce` |
103
+ | zm_yunxia | ๐Ÿšน | C | MM minutes | D | `bb2b03b0` |
104
+ | zm_yunyang | ๐Ÿšน | C | MM minutes | D | `5238ac22` |
105
+
106
+ ### Spanish
107
+
108
+ - `lang_code='e'` in [`misaki[en]`](https://github.com/hexgrad/misaki)
109
+ - espeak-ng `es`
110
+
111
+ | Name | Traits | SHA256 |
112
+ | ---- | ------ | ------ |
113
+ | ef_dora | ๐Ÿšบ | `d9d69b0f` |
114
+ | em_alex | ๐Ÿšน | `5eac53f7` |
115
+ | em_santa | ๐Ÿšน | `aa8620cb` |
116
+
117
  ### French
118
 
119
  - `lang_code='f'` in [`misaki[en]`](https://github.com/hexgrad/misaki)
 
148
  | if_sara | ๐Ÿšบ | B | MM minutes | C | `6c0b253b` |
149
  | im_nicola | ๐Ÿšน | B | MM minutes | C | `234ed066` |
150
 
151
+ ### Brazilian Portuguese
152
 
153
+ - `lang_code='p'` in [`misaki[en]`](https://github.com/hexgrad/misaki)
154
+ - espeak-ng `pt-br`
155
 
156
+ | Name | Traits | SHA256 |
157
+ | ---- | ------ | ------ |
158
+ | pf_dora | ๐Ÿšบ | `07e4ff98` |
159
+ | pm_alex | ๐Ÿšน | `cf0ba8c5` |
160
+ | pm_santa | ๐Ÿšน | `d4210316` |