Skip to the content.

jaCappella corpus / jaCappellaコーパス

The jaCappella corpus is a corpus of Japanese a cappella vocal ensembles. It consists of 50 copyright-cleared vocal ensemble songs and their audio recordings of individual voice parts. (Note that the jaCappella corpus team reserves all copyrights and neighboring rights of the vocal ensemble songs.) These songs were arranged from out-of-copyright Japanese children’s songs. All songs have six voice parts (lead vocal, soprano, alto, tenor, bass, and vocal percussion). They are divided into 10 subsets (jazz, punk rock, bossa nova, popular, reggae, enka, neutral, ballad, edm, and soulfunk), each of which features typical characteristics of a music genre.

jaCappellaコーパスは,日本語のアカペラ重唱曲のコーパスです. 本コーパスは,50曲の著作権処理済みアカペラ重唱曲と各声部の単独歌唱音源からなります.(jaCappellaコーパスチームは,作成した重唱曲に関する全ての著作権と著作隣接権を保有しています.) これらの重唱曲は,著作権保護期間の終了した童謡・唱歌から編曲して作成されています. 声部は,lead vocal,soprano,alto,tenor,bass,vocal percussionの6つです. 本コーパスは,10個のサブセット(jazz, punk rock, bossa nova, popular, reggae, enka, neutral, ballad, edm, and soulfunk)からなり,それぞれ対応するジャンルの典型的な特徴を備えた曲群からなります.

Audio files of singing voices / 歌唱音源ファイル

The singing voices were recorded in a recording studio. The sampling frequency is 48 kHz and the audio file is provided in a monaural WAVE format. All singers are native Japanese speakers.

歌唱音源は曲・声部毎にレコーディングスタジオで収録されたものであり,サンプリング周波数は48 kHz,モノラル形式のWAVEファイルとして保存されています. 歌唱者は全て日本語母語話者です.

Musical score / 譜面

The musical scores with Japanese lyrics are provided in the PDF and MusicXML formats. This dataset contains MusicXML files whose lyrics were modifed for singing voice synthesis (e.g. “は”->”わ”, “へ”->”え”). For non-Japanese researchers, MusicXML files with Romaji transcriptions of lyrics are provided.

Remark: Romaji transcription does not exactly match Japanese pronunciation. For the exact Japanese pronounciation, please refer to the singing voices.

譜面は,PDFおよびMusicXML形式で配布されます. また,歌声合成用に歌詞を発音表記に変換したMusicXMLファイル(e.g., 「は」->「わ」,「へ」->「え」),非日本語話者用に歌詞をヘボン式ローマ字で表記したMusicXMLファイルも同梱しています.


Metadata / メタデータ

Metadata of the songs are given as a csv file. The items in the csv file are as follows.

原曲の情報や歌唱者のID・性別などは,csvファイルにメタデータとしてまとめられています. csvファイルの項目は以下の通りです.

Terms of Use / 利用規約

English version


For commercial Use / 商用利用に関して

We welcome commercial use of this corpus. The commercial use is chargeable and please feel free to contact the following members for your commercial use.


Download / ダウンロード

You can download here.


Update logs

Example applications

Contributors (The jaCappella corpus team)

Citation / 引用

If you use the jaCappella corpus, please cite the following paper.


    author={Nakamura, Tomohiko and Takamichi, Shinnosuke and Tanji, Naoko and Fukayama, Satoru and Saruwatari, Hiroshi},
    title={jaCappella corpus: A Japanese a cappella vocal ensemble corpus},
    booktitle = ICASSP,
    month     = jun,
    year      = 2023,
    url_arXiv = {},
    url_Demo  = {demo/jaCappella_sep},
    url_Code  = {},
    doi       = {10.1109/ICASSP49357.2023.10095569},
    lang      = {en}

Links / リンク

Acknowledgement / 謝辞

The jaCappella corpus team reserves all copyrights and neighboring rights of the vocal ensemble songs. The music arrangement and sound production were performed by Alpha Enterprises, Inc.

jaCappellaコーパスチームは,作成した重唱曲に関する全ての著作権と著作隣接権を保有しています. 譜面,音源制作は,株式会社アルファエンタープライズが行いました.