Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -1,41 +1,14 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
2 |
language:
|
3 |
- zh
|
4 |
- en
|
5 |
- ja
|
6 |
- ko
|
7 |
-
task_categories:
|
8 |
-
- audio-classification
|
9 |
-
- automatic-speech-recognition
|
10 |
-
- text-to-speech
|
11 |
pretty_name: Genshin Voice
|
12 |
-
configs:
|
13 |
-
- config_name: default
|
14 |
-
data_files:
|
15 |
-
- split: train
|
16 |
-
path: data/train-*
|
17 |
-
dataset_info:
|
18 |
-
features:
|
19 |
-
- name: audio
|
20 |
-
dtype: audio
|
21 |
-
- name: transcription
|
22 |
-
dtype: string
|
23 |
-
- name: language
|
24 |
-
dtype: string
|
25 |
-
- name: speaker
|
26 |
-
dtype: string
|
27 |
-
- name: speaker_type
|
28 |
-
dtype: string
|
29 |
-
- name: type
|
30 |
-
dtype: string
|
31 |
-
- name: inGameFilename
|
32 |
-
dtype: string
|
33 |
-
splits:
|
34 |
-
- name: train
|
35 |
-
num_bytes: 247309786538.984
|
36 |
-
num_examples: 413429
|
37 |
-
download_size: 201995753083
|
38 |
-
dataset_size: 247309786538.984
|
39 |
---
|
40 |
|
41 |
# Genshin Voice
|
@@ -45,13 +18,13 @@ Genshin Voice is a dataset of voice lines from the popular game [Genshin Impact]
|
|
45 |
Hugging Face 🤗 [Genshin-Voice](https://huggingface.co/datasets/simon3000/genshin-voice)
|
46 |
|
47 |
<!-- STATS -->
|
48 |
-
Last update at `2024-04-
|
49 |
|
50 |
`413429` wavs
|
51 |
|
52 |
-
`
|
53 |
|
54 |
-
`
|
55 |
|
56 |
`720` without inGameFilename (0%)
|
57 |
<!-- STATS_END -->
|
@@ -68,20 +41,63 @@ The voice lines are spoken by the characters in the game and cover a wide range
|
|
68 |
|
69 |
## Uses
|
70 |
|
71 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
|
73 |
```json
|
74 |
{
|
75 |
-
"
|
76 |
-
"
|
|
|
77 |
"language": "English(US)",
|
78 |
-
"
|
79 |
-
"
|
80 |
-
"
|
81 |
-
"
|
82 |
-
"
|
83 |
-
"
|
84 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
}
|
86 |
```
|
87 |
|
@@ -107,6 +123,8 @@ The dataset contains official annotations from the game, including language, spe
|
|
107 |
|
108 |
Annotations are incomplete. Some voice lines are missing speaker names and transcriptions.
|
109 |
|
|
|
|
|
110 |
### Recommendations
|
111 |
|
112 |
Users should be made aware of the risks, biases and limitations of the dataset.
|
|
|
1 |
---
|
2 |
+
task_categories:
|
3 |
+
- audio-classification
|
4 |
+
- automatic-speech-recognition
|
5 |
+
- text-to-speech
|
6 |
language:
|
7 |
- zh
|
8 |
- en
|
9 |
- ja
|
10 |
- ko
|
|
|
|
|
|
|
|
|
11 |
pretty_name: Genshin Voice
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
---
|
13 |
|
14 |
# Genshin Voice
|
|
|
18 |
Hugging Face 🤗 [Genshin-Voice](https://huggingface.co/datasets/simon3000/genshin-voice)
|
19 |
|
20 |
<!-- STATS -->
|
21 |
+
Last update at `2024-04-30`
|
22 |
|
23 |
`413429` wavs
|
24 |
|
25 |
+
`18016` without speaker (4%)
|
26 |
|
27 |
+
`22956` without transcription (6%)
|
28 |
|
29 |
`720` without inGameFilename (0%)
|
30 |
<!-- STATS_END -->
|
|
|
41 |
|
42 |
## Uses
|
43 |
|
44 |
+
To install Hugging Face's datasets library, follow the instructions from [this link](https://huggingface.co/docs/datasets/installation#audio).
|
45 |
+
|
46 |
+
### Example: Load the dataset and filter for Chinese voices of Ganyu with transcriptions
|
47 |
+
|
48 |
+
```python
|
49 |
+
from datasets import load_dataset
|
50 |
+
import soundfile as sf
|
51 |
+
import os
|
52 |
+
|
53 |
+
# Load the dataset
|
54 |
+
dataset = load_dataset('simon3000/genshin-voice', split='train', streaming=True)
|
55 |
+
|
56 |
+
# Filter the dataset for Chinese voices of Ganyu with transcriptions
|
57 |
+
chinese_ganyu = dataset.filter(lambda voice: voice['language'] == 'Chinese' and voice['speaker'] == 'Ganyu' and voice['transcription'] != '')
|
58 |
+
|
59 |
+
# Create a folder to store the audio and transcription files
|
60 |
+
ganyu_folder = 'ganyu'
|
61 |
+
os.makedirs(ganyu_folder, exist_ok=True)
|
62 |
+
|
63 |
+
# Process each voice in the filtered dataset
|
64 |
+
for i, voice in enumerate(chinese_ganyu):
|
65 |
+
audio_path = os.path.join(ganyu_folder, f'{i}_audio.wav') # Path to save the audio file
|
66 |
+
transcription_path = os.path.join(ganyu_folder, f'{i}_transcription.txt') # Path to save the transcription file
|
67 |
+
|
68 |
+
# Save the audio file
|
69 |
+
sf.write(audio_path, voice['audio']['array'], voice['audio']['sampling_rate'])
|
70 |
+
|
71 |
+
# Save the transcription file
|
72 |
+
with open(transcription_path, 'w') as transcription_file:
|
73 |
+
transcription_file.write(voice['transcription'])
|
74 |
+
|
75 |
+
print(f'{i} done') # Print the progress
|
76 |
+
```
|
77 |
+
|
78 |
+
### You unpacked the game and just want to know what the wavs are about
|
79 |
+
|
80 |
+
result.json format: (subject to change)
|
81 |
|
82 |
```json
|
83 |
{
|
84 |
+
"9b5502fb1b83cb97.wav": {
|
85 |
+
"inGameFilename": "VO_friendship\\VO_raidenShogun\\vo_raidenEi_dialog_pendant.wem",
|
86 |
+
"filename": "9b5502fb1b83cb97.wav",
|
87 |
"language": "English(US)",
|
88 |
+
"transcription": "Really? So in all this time, no new Electro Visions have appeared in the outside world? Well, what I can say on this topic is subject to certain constraints, but... it is not by my will that Visions are granted or denied. The key is people's desire, and... well, there's another side to it too.",
|
89 |
+
"speaker": "Raiden Shogun",
|
90 |
+
"talkRoleType": "",
|
91 |
+
"talkRoleID": "",
|
92 |
+
"guid": "f8e72b65-6c0a-4df1-a2f0-2bb08dbeab75",
|
93 |
+
"voiceConfigs": [
|
94 |
+
{
|
95 |
+
"gameTrigger": "Fetter",
|
96 |
+
"gameTriggerArgs": 3001,
|
97 |
+
"avatarName": "Switch_raidenShogun"
|
98 |
+
}
|
99 |
+
]
|
100 |
+
}
|
101 |
}
|
102 |
```
|
103 |
|
|
|
123 |
|
124 |
Annotations are incomplete. Some voice lines are missing speaker names and transcriptions.
|
125 |
|
126 |
+
Speakers and transcriptions may contain markups and placeholders: `#<color=#37FFFF>パイモン:</color>{NICKNAME}、すごく怖い悪夢を見たことってあるか?\\n<color=#37FFFF>{NICKNAME}:...`
|
127 |
+
|
128 |
### Recommendations
|
129 |
|
130 |
Users should be made aware of the risks, biases and limitations of the dataset.
|