Hugging Face's logo

hexgrad
/
Kokoro-82M

Text-to-Speech
English
Model card Files
xet
hexgrad commited on
Commit
0cbe210
·
1 Parent(s): 6fc7acd

Upload VOICES.md

Browse files
Files changed (1) hide show
  1. VOICES.md +18 -10
VOICES.md CHANGED
@@ -10,10 +10,14 @@ Subjectively, voices will sound better or worse to different people.
10
 
11
  **Training Duration**
12
  - How much audio was seen during training? Smaller durations result in a lower overall grade.
 
 
 
 
13
 
14
- ### American 🇺🇸
15
 
16
- American G2P: [`misaki[en]`](https://github.com/hexgrad/misaki) with `en-us` espeak-ng fallback
17
 
18
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
19
  | ---- | ------ | -------------- | ----------------- | ------------- |
@@ -36,9 +40,9 @@ American G2P: [`misaki[en]`](https://github.com/hexgrad/misaki) with `en-us` esp
36
  | am_onyx | 🚹 | C | MM minutes | D |
37
  | am_puck | 🚹 | B | H hours | C+ |
38
 
39
- ### British 🇬🇧
40
 
41
- British G2P: [`misaki[en]`](https://github.com/hexgrad/misaki) with `en-gb` espeak-ng fallback
42
 
43
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
44
  | ---- | ------ | -------------- | ----------------- | ------------- |
@@ -53,17 +57,17 @@ British G2P: [`misaki[en]`](https://github.com/hexgrad/misaki) with `en-gb` espe
53
 
54
  ### French 🇫🇷
55
 
56
- French G2P: espeak-ng `fr-fr`
57
 
58
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
59
  | ---- | ------ | -------------- | ----------------- | ------------- |
60
  | [ff_siwis](https://datashare.ed.ac.uk/handle/10283/2353) | 🚺 | B | <11 hours | B- |
61
 
62
- This table lists all French training data seen by Kokoro.
63
 
64
  ### Hindi 🇮🇳
65
 
66
- Hindi G2P: espeak-ng `hi`
67
 
68
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
69
  | ---- | ------ | -------------- | ----------------- | ------------- |
@@ -72,19 +76,21 @@ Hindi G2P: espeak-ng `hi`
72
  | hm_omega | 🚹 | B | MM minutes | C |
73
  | hm_psi | 🚹 | B | MM minutes | C |
74
 
75
- This table lists all Hindi training data seen by Kokoro, which totals about 6 hours.
76
 
77
  ### Japanese 🇯🇵
78
 
79
- Japanese G2P: [`misaki[ja]`](https://github.com/hexgrad/misaki)
80
 
81
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
82
  | ---- | ------ | -------------- | ----------------- | ------------- |
83
  | jf_alpha | 🚺 | B | H hours | C+ |
84
 
 
 
85
  ### Mandarin Chinese 🇨🇳
86
 
87
- Mandarin Chinese G2P: [`misaki[zh]`](https://github.com/hexgrad/misaki)
88
 
89
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
90
  | ---- | ------ | -------------- | ----------------- | ------------- |
@@ -96,3 +102,5 @@ Mandarin Chinese G2P: [`misaki[zh]`](https://github.com/hexgrad/misaki)
96
  | zm_yunxi | 🚹 | C | MM minutes | D |
97
  | zm_yunxia | 🚹 | C | MM minutes | D |
98
  | zm_yunyang | 🚹 | C | MM minutes | D |
 
 
 
10
 
11
  **Training Duration**
12
  - How much audio was seen during training? Smaller durations result in a lower overall grade.
13
+ - 10 hours <= HH hours < 100 hours
14
+ - 1 hour <= H hours < 10 hours
15
+ - 10 minutes <= MM minutes < 100 minutes
16
+ - 1 minute <= M minutes < 10 minutes
17
 
18
+ ### American English 🇺🇸
19
 
20
+ G2P: [`misaki[en]`](https://github.com/hexgrad/misaki) `lang_code='a'` with `en-us` espeak-ng fallback
21
 
22
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
23
  | ---- | ------ | -------------- | ----------------- | ------------- |
 
40
  | am_onyx | 🚹 | C | MM minutes | D |
41
  | am_puck | 🚹 | B | H hours | C+ |
42
 
43
+ ### British English 🇬🇧
44
 
45
+ G2P: [`misaki[en]`](https://github.com/hexgrad/misaki) `lang_code='b'` with `en-gb` espeak-ng fallback
46
 
47
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
48
  | ---- | ------ | -------------- | ----------------- | ------------- |
 
57
 
58
  ### French 🇫🇷
59
 
60
+ G2P: espeak-ng `fr-fr`
61
 
62
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
63
  | ---- | ------ | -------------- | ----------------- | ------------- |
64
  | [ff_siwis](https://datashare.ed.ac.uk/handle/10283/2353) | 🚺 | B | <11 hours | B- |
65
 
66
+ Total French training data: <11 hours
67
 
68
  ### Hindi 🇮🇳
69
 
70
+ G2P: espeak-ng `hi`
71
 
72
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
73
  | ---- | ------ | -------------- | ----------------- | ------------- |
 
76
  | hm_omega | 🚹 | B | MM minutes | C |
77
  | hm_psi | 🚹 | B | MM minutes | C |
78
 
79
+ Total Hindi training data: H hours
80
 
81
  ### Japanese 🇯🇵
82
 
83
+ G2P: [`misaki[ja]`](https://github.com/hexgrad/misaki)
84
 
85
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
86
  | ---- | ------ | -------------- | ----------------- | ------------- |
87
  | jf_alpha | 🚺 | B | H hours | C+ |
88
 
89
+ Total Japanese training data: H hours
90
+
91
  ### Mandarin Chinese 🇨🇳
92
 
93
+ G2P: [`misaki[zh]`](https://github.com/hexgrad/misaki)
94
 
95
  | Name | Traits | Target Quality | Training Duration | Overall Grade |
96
  | ---- | ------ | -------------- | ----------------- | ------------- |
 
102
  | zm_yunxi | 🚹 | C | MM minutes | D |
103
  | zm_yunxia | 🚹 | C | MM minutes | D |
104
  | zm_yunyang | 🚹 | C | MM minutes | D |
105
+
106
+ Total Mandarin Chinese training data: H hours