TiMauzi/imslp-midi-cc0-1.0

IMSLP MIDI Dataset (CC0-1.0)

This dataset contains MIDI files and metadata crawled from IMSLP (International Music Score Library Project) on July 21-22, 2024.

Data Fields

• midi_source: URL to the original MIDI file on IMSLP (incl. original uploader).
• metadata_source: URL to the original metadata on IMSLP.
• file_name, title, composer, year, era, style, key, license: Metadata fields.
• midi: Raw MIDI bytes.
• midi_mido: JSON-serialized mido object.

How to Retrieve Origin Data and Authors

• For each entry, use midi_source to find the original MIDI file and metadata_source for the metadata/author information.

Statistics & Visualizations

• Total entries: 1113
• Unique composers: 406
• Eras: Romantic (474), Baroque (241), Modern (231), Classical (75), Renaissance (51), Early 20th century (28), Other (13)
• Styles: Romantic (473), Baroque (257), Modern (148), Classical (95), Early 20th century (62), Renaissance (57), Traditional (8), Ancient (5), Jazz (4), Non-western classical (2), Medieval (2)

How to Use

from datasets import load_dataset
dataset = load_dataset('TiMauzi/imslp-midi-cc0-1.0')

The data in the column midi are raw bytes. You may want to use packages like mido to analyze them. For example, to view the content of the first MIDI file, you can use a script as follows:

import mido
import io

train_data = dataset['train']  # assuming you already loaded the dataset as described

# Take the first example
midi_bytes = train_data[0]['midi']  # this is raw bytes

# Load with mido
midi_file = mido.MidiFile(file=io.BytesIO(midi_bytes))

# Inspect tracks and messages
for i, track in enumerate(midi_file.tracks):
    print(f"Track {i}: {track.name}")
    for msg in track[:10]:  # print first 10 messages
        print(msg)

License Merging

While most files in IMSLP are stored as public domain or open source, different and sometimes incompatible licensing agreements exacerbate the conflation of individual pieces into a single data collection. While there are several customized licenses for some of the MIDI files, some of the crawled and observed data were purely from the public domain globally, whereas the vast majority of all MIDI files make use of Creative Commons (CC) agreements. The latter may be combined under specific circumstances: For example, CC-BY-licensed pieces may be remixed with CC-BY-SA- and CC-BY-NC-licensed pieces into a CC-BY-SA or CC-BY-NC remix, respectively. Furthermore, CC-BY-NC-licensed pieces or remixes may again be remixed using the CC-BY-NC-SA license. Finally, ShareAlike (SA) licenses CC-BY-SA and CC-BY-NC-SA of version 3.0 are upgradable to those of version 4.0. Since the compatibility of all other, mostly custom licenses, remains ambiguous, they have been dropped.

This specific dataset version only contains data in the public domain and cc0-1.0. For other versions, please check out TiMauzi/imslp-midi-by-nc-sa or TiMauzi/imslp-midi-by-sa.

In case of any copyright issues or concerns, feel free to contact me, so I can take this dataset offline.

Citation

If you want to reference this work, you may use the following BibTeX citation of the proof-of-concept repository available on Zenodo:

@misc{tim_sockel_timauzi/midi-models:_2026,
    title = {{TiMauzi}/midi-models: {Initial} {Release}},
    copyright = {MIT License},
    shorttitle = {{TiMauzi}/midi-models},
    url = {https://zenodo.org/doi/10.5281/zenodo.18913009},
    abstract = {This release contains the first version of the proof-of-concept notebook to train a simple regression model on MIDI data. Furthermore, it contains a README.md describing the process of crawling and preprocessing the datasets.},
    urldate = {2026-03-08},
    publisher = {Zenodo},
    author = {{Tim Sockel}},
    month = mar,
    year = {2026},
    doi = {10.5281/ZENODO.18913009},
}