ayousanz's picture
Upload README.md with huggingface_hub
eacce3a verified
|
Raw
History Blame Contribute Delete
9.5 kB
---
license: cc-by-nc-4.0
language:
- ja
task_categories:
- audio-classification
tags:
- singing-voice
- japanese
- seed-vc
- voice-conversion
- svc
- rvc
- music
- webdataset
size_categories:
- 10K<n<100K
---
# Japanese Singing Voice Dataset / 日本語歌声データセット
[English](#english) | [日本語](#日本語)
---
## English
A large-scale Japanese singing voice dataset for training voice conversion models.
### Dataset Description
This dataset contains Japanese singing voice audio files collected for training singing voice conversion (SVC) models such as Seed-VC, RVC, So-VITS-SVC, and similar architectures.
### Dataset Statistics
| Metric | Value |
|--------|-------|
| Total Duration | ~1,000 hours |
| Number of Files | 15,311 |
| Number of Shards | 78 |
| Audio Format | MP3 (in WebDataset TAR) |
| Bitrate | ~170 kbps (VBR) |
| Total Size | ~73 GB |
| Language | Japanese |
### Data Format
This dataset uses **WebDataset** format for efficient streaming and training.
```
data/
├── train-0000.tar
├── train-0001.tar
├── ...
└── train-0077.tar
```
Each TAR archive contains paired files:
- `{video_id}.mp3` - Audio file
- `{video_id}.json` - Metadata
### Metadata Fields
| Field | Type | Description |
|-------|------|-------------|
| id | string | YouTube video ID |
| url | string | YouTube URL |
| title | string | Song title |
| channel_id | string | YouTube channel ID |
| channel_title | string | Channel name / Artist |
| duration_sec | int | Duration in seconds |
| published | string | Published date (ISO 8601) |
| view_count | int | View count |
| like_count | int | Like count |
| category_id | string | YouTube category ID ("10" = Music) |
| topic_categories | array | Wikipedia topic categories |
| language | string | "ja" |
| dataset_category | string | Classification (official/cover/diverse) |
| dataset_genre | string | Genre (pop/rock/electronic/other) |
| dataset_score | int | Quality score (0-100) |
### Usage
#### Streaming with WebDataset
```python
import webdataset as wds
from huggingface_hub import get_token
from torch.utils.data import DataLoader
hf_token = get_token()
url = "https://huggingface.co/datasets/tts-dataset/japanese-singing-voice/resolve/main/data/train-{0000..0077}.tar"
url = f"pipe:curl -s -L {url} -H 'Authorization:Bearer {hf_token}'"
dataset = wds.WebDataset(url).decode()
dataloader = DataLoader(dataset, batch_size=16, num_workers=4)
for sample in dataloader:
audio = sample["mp3"]
metadata = sample["json"]
# Process audio and metadata
```
#### With Hugging Face Datasets
```python
from datasets import load_dataset
dataset = load_dataset(
"tts-dataset/japanese-singing-voice",
split="train",
streaming=True
)
for sample in dataset:
audio = sample["mp3"]
metadata = sample["json"]
```
### Content Categories
- **Official** (~40%): Official music videos from record labels and artists
- **Cover** (~30%): Cover songs by various singers
- **Diverse** (~30%): Various genres and styles
### Intended Uses
- Training singing voice conversion models (Seed-VC, RVC, So-VITS-SVC)
- Training Japanese HuBERT models
- Research on singing voice synthesis
- Academic research on voice conversion
### License
This dataset is licensed under **CC-BY-NC-4.0** (Creative Commons Attribution-NonCommercial 4.0 International).
> **Important:** This dataset does not own the copyright to the audio files; the copyright remains with the original owners of the videos or audio. Users are permitted to use this dataset only for non-commercial purposes.
#### Terms
- **Attribution** — You must give appropriate credit and indicate if changes were made
- **NonCommercial** — You may not use the material for commercial purposes
- Commercial use requires explicit permission from original copyright holders
- Users are responsible for ensuring compliance with applicable laws in their jurisdiction
### Source
This dataset is derived from publicly available YouTube videos, filtered for Japanese singing content.
Original metadata source: [ayousanz/music-youtube-list](https://huggingface.co/datasets/ayousanz/music-youtube-list)
### Citation
If you use this dataset in your research, please cite:
```bibtex
@dataset{japanese_singing_voice,
title={Japanese Singing Voice Dataset},
author={tts-dataset},
year={2025},
publisher={Hugging Face},
url={https://huggingface.co/datasets/tts-dataset/japanese-singing-voice}
}
```
### Disclaimer
This dataset is provided "as is" without warranty of any kind. The maintainers are not responsible for any misuse of the data.
---
## 日本語
歌声変換モデルの学習用大規模日本語歌声データセットです。
### データセット概要
このデータセットは、Seed-VC、RVC、So-VITS-SVCなどの歌声変換(SVC)モデルの学習用に収集された日本語歌声音声ファイルを含んでいます。
### データセット統計
| 項目 | 値 |
|------|-----|
| 総再生時間 | 約1,000時間 |
| ファイル数 | 15,311 |
| シャード数 | 78 |
| 音声形式 | MP3(WebDataset TAR内) |
| ビットレート | 約170 kbps (VBR) |
| 合計サイズ | 約73 GB |
| 言語 | 日本語 |
### データ形式
このデータセットは効率的なストリーミングと学習のために**WebDataset**形式を使用しています。
```
data/
├── train-0000.tar
├── train-0001.tar
├── ...
└── train-0077.tar
```
各TARアーカイブにはペアのファイルが含まれています:
- `{video_id}.mp3` - 音声ファイル
- `{video_id}.json` - メタデータ
### メタデータフィールド
| フィールド | 型 | 説明 |
|-----------|---|------|
| id | string | YouTube動画ID |
| url | string | YouTube URL |
| title | string | 曲タイトル |
| channel_id | string | YouTubeチャンネルID |
| channel_title | string | チャンネル名/アーティスト |
| duration_sec | int | 秒数 |
| published | string | 公開日時(ISO 8601) |
| view_count | int | 再生回数 |
| like_count | int | いいね数 |
| category_id | string | YouTubeカテゴリID("10"=音楽) |
| topic_categories | array | Wikipediaトピックカテゴリ |
| language | string | "ja" |
| dataset_category | string | 分類(official/cover/diverse) |
| dataset_genre | string | ジャンル(pop/rock/electronic/other) |
| dataset_score | int | 品質スコア(0-100) |
### 使用方法
#### WebDatasetでストリーミング
```python
import webdataset as wds
from huggingface_hub import get_token
from torch.utils.data import DataLoader
hf_token = get_token()
url = "https://huggingface.co/datasets/tts-dataset/japanese-singing-voice/resolve/main/data/train-{0000..0077}.tar"
url = f"pipe:curl -s -L {url} -H 'Authorization:Bearer {hf_token}'"
dataset = wds.WebDataset(url).decode()
dataloader = DataLoader(dataset, batch_size=16, num_workers=4)
for sample in dataloader:
audio = sample["mp3"]
metadata = sample["json"]
# 音声とメタデータを処理
```
#### Hugging Face Datasetsを使用
```python
from datasets import load_dataset
dataset = load_dataset(
"tts-dataset/japanese-singing-voice",
split="train",
streaming=True
)
for sample in dataset:
audio = sample["mp3"]
metadata = sample["json"]
```
### コンテンツカテゴリ
- **Official(公式)** (~40%): レコード会社やアーティストの公式ミュージックビデオ
- **Cover(カバー)** (~30%): 様々な歌手によるカバー曲
- **Diverse(多様)** (~30%): 様々なジャンルとスタイル
### 想定される用途
- 歌声変換モデルの学習(Seed-VC、RVC、So-VITS-SVC)
- 日本語HuBERTモデルの学習
- 歌声合成の研究
- 音声変換に関する学術研究
### ライセンス
このデータセットは**CC-BY-NC-4.0**(クリエイティブ・コモンズ 表示-非営利 4.0 国際)でライセンスされています。
> **重要:** このデータセットは音声ファイルの著作権を所有していません。著作権は動画または音声の元の所有者に帰属します。このデータセットは非商用目的でのみ使用できます。
#### 条件
- **表示** — 適切なクレジットを表示し、変更があった場合はその旨を示す必要があります
- **非営利** — 営利目的での使用はできません
- 商用利用には元の著作権者からの明示的な許可が必要です
- 利用者は各自の法域で適用される法律への準拠を確認する責任があります
### ソース
このデータセットは、日本語歌唱コンテンツをフィルタリングした公開YouTube動画から派生しています。
元のメタデータソース: [ayousanz/music-youtube-list](https://huggingface.co/datasets/ayousanz/music-youtube-list)
### 引用
研究でこのデータセットを使用する場合は、以下を引用してください:
```bibtex
@dataset{japanese_singing_voice,
title={Japanese Singing Voice Dataset},
author={tts-dataset},
year={2025},
publisher={Hugging Face},
url={https://huggingface.co/datasets/tts-dataset/japanese-singing-voice}
}
```
### 免責事項
このデータセットは「現状のまま」提供され、いかなる種類の保証もありません。メンテナーはデータの誤用について責任を負いません。