license: cc-by-4.0
pretty_name: ChoraleBricks
task_categories:
- audio-to-audio
- audio-classification
tags:
- audio
- music
- source-separation
- multitrack
- wind-music
- brass
- music-information-retrieval
- music-transcription
- audio-alignment
configs:
- config_name: archives
data_files:
- split: train
path: metadata.csv
ChoraleBricks
This repository mirrors ChoraleBricks: A Modular Multitrack Dataset for Wind Music Research.
ChoraleBricks contains multitrack recordings of ten chorales arranged in soprano, alto, tenor, and bass parts. It provides isolated recordings of individual wind instruments, including flute, oboe, clarinet, trumpet, saxophone, baritone, trombone, and tuba, which can be combined into ensemble mixes with different instrumentations.
The upstream record describes 2 hours and 10 minutes of isolated recordings and 52 hours and 18 minutes of possible full ensemble permutations, plus sheet music, aligned symbolic representations, conducting videos, and reference annotations for tasks such as source separation, multi-pitch estimation, note transcription, and audio alignment.
This mirror preserves the upstream archive files and provides metadata.csv for the dataset viewer.
Source
- Zenodo: https://zenodo.org/records/15081741
- DOI:
10.5281/zenodo.15081741 - Project page: https://audiolabs-erlangen.de/resources/MIR/2025-ChoraleBricks
- Code: https://github.com/stefan-balke/choralebricks
- License: Creative Commons Attribution 4.0 International
Citation
@article{balke2025choralebricks,
title = {ChoraleBricks: A Modular Multitrack Dataset for Wind Music Research},
author = {Balke, Stefan and Berndt, Axel and Mueller, Meinard},
journal = {Transactions of the International Society for Music Information Retrieval},
year = {2025}
}