Upload README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,59 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- af
|
| 4 |
+
- ar
|
| 5 |
+
- ar
|
| 6 |
+
- bg
|
| 7 |
+
- ca
|
| 8 |
+
- cs
|
| 9 |
+
- cy
|
| 10 |
+
- da
|
| 11 |
+
- de
|
| 12 |
+
- el
|
| 13 |
+
- en
|
| 14 |
+
- es
|
| 15 |
+
- eu
|
| 16 |
+
- fi
|
| 17 |
+
- fr
|
| 18 |
+
- he
|
| 19 |
+
- hi
|
| 20 |
+
- hr
|
| 21 |
+
- hu
|
| 22 |
+
- id
|
| 23 |
+
- it
|
| 24 |
+
- ja
|
| 25 |
+
- ko
|
| 26 |
+
- ms
|
| 27 |
+
- nb
|
| 28 |
+
- ne
|
| 29 |
+
- nl
|
| 30 |
+
- pl
|
| 31 |
+
- pt
|
| 32 |
+
- ro
|
| 33 |
+
- ru
|
| 34 |
+
- sk
|
| 35 |
+
- sl
|
| 36 |
+
- sq
|
| 37 |
+
- sr
|
| 38 |
+
- sv
|
| 39 |
+
- sw
|
| 40 |
+
- ta
|
| 41 |
+
- te
|
| 42 |
+
- th
|
| 43 |
+
- tr
|
| 44 |
+
- vi
|
| 45 |
+
- zh
|
| 46 |
+
|
| 47 |
+
license: mit
|
| 48 |
+
---
|
| 49 |
+
|
| 50 |
+
# Phonetic Piper Studio Recordings Prompts
|
| 51 |
+
|
| 52 |
+
This dataset is a processed version of an utterance dataset made available for various languages as prompts for the [Piper recording studio](hhttps://github.com/rhasspy/piper-recording-studio/tree/master/prompts). Along with the original prompts, we include:
|
| 53 |
+
|
| 54 |
+
- columns `ipa_espeak` and `ipa_epitran` containing phonemized versions of the original sentences according to [`espeak-ng`](https://github.com/espeak-ng/espeak-ng) and [`Epitran`](https://github.com/dmort27/epitran) phonemizers, respectively
|
| 55 |
+
- columns `lang`, `espeak_lang_code`, `epitran_lang_code` containing the language codes as reported by `piper recording studio`, `espeak-ng` and `Epitran`
|
| 56 |
+
- word boundary marker set as the character `|`
|
| 57 |
+
- a 90%/10% train/test split
|
| 58 |
+
|
| 59 |
+
The distribution license is inherited from the original dataset and so free to copy and use under the MIT license.
|