README.md

metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: mid
      dtype: string
    - name: midi_file
      struct:
        - name: bytes
          list: uint8
        - name: filename
          dtype: string
        - name: size
          dtype: int64
    - name: audio
      dtype: audio
    - name: audio_duration
      dtype: float64
    - name: abc
      dtype: string
    - name: token_length_abc
      dtype: int64
  splits:
    - name: train
      num_bytes: 340021450637.4
      num_examples: 86480
  download_size: 380389725217
  dataset_size: 340021450637.4
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
tags:
  - music
  - art

midi, synthesized audio, ABC code triples

(this dataset contains those with audio duration in 5 min - 2 hours, less than 5 min data are in 300s and there are several subsets with smaller duration 60s 30s 10s)

(token_length_abc field represents the token count of the abc text w.r.t. Qwen3's tokenizer)

• midi files are from bread-midi-dataset
• synthesized audio: use Don Allen's Timbres of Heaven as soundfont and FluidSynth as synthesizer
• abc notation: mid2abc by EasyABC (midi2abc.py)

Citation

@misc{jiang2025advancingfoundationmodelmusic,
      title={Advancing the Foundation Model for Music Understanding}, 
      author={Yi Jiang and Wei Wang and Xianwen Guo and Huiyun Liu and Hanrui Wang and Youri Xu and Haoqi Gu and Zhongqian Xie and Chuanjiang Luo},
      year={2025},
      eprint={2508.01178},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2508.01178}, 
}

@misc {matthew_mitton_2025,
    author       = { {Matthew Mitton} },
    title        = { bread-midi-dataset (Revision 95c2155) },
    year         = 2025,
    url          = {\url{https://huggingface.co/datasets/breadlicker45/bread-midi-dataset }},
    doi          = { 10.57967/hf/4748 },
    publisher    = { Hugging Face }
}