MUSIC8K

This dataset is organized by music generation model name.
Each subfolder contains MP3 audio files generated by the corresponding model.

Folder structure

music-by-model-mp3/
  acestep1.5/            # MP3 files generated by acestep1.5
  heartmula/             # MP3 files generated by heartmula
  minimax_2.6/           # MP3 files generated by minimax_2.6
  mureka_v9/             # MP3 files generated by mureka_v9
  suno_v5.5/             # MP3 files generated by suno_v5.5
  sunov5_new/            # MP3 files generated by sunov5_new
  metadata/
    pop_1000_unique_artist.csv
    songs_musiccaps.jsonl
  scripts/
    tamper.py

Audio counts

• acestep1.5: 989 MP3 files
• heartmula: 989 MP3 files
• minimax_2.6: 304 MP3 files
• mureka_v9: 662 MP3 files
• suno_v5.5: 330 MP3 files
• sunov5_new: 340 MP3 files

Total: 3614 MP3 files

Additional files

metadata/pop_1000_unique_artist.csv

Reference metadata used for song generation, including lyrics and artist-related information.

metadata/songs_musiccaps.jsonl

Reference metadata used for song generation, including prompt and lyric-related information.

scripts/tamper.py

Implementation of three audio tampering operations used in this project.

Usage notes

• Audio files are grouped by model name for easier cross-model comparison.
• The two files in metadata/ provide prompt and lyric context for synthesized songs.
• scripts/tamper.py is included for reproducibility of the audio tampering pipeline.

Related datasets

Data for ACEStep 1.0 and Mureka O1 are available at:

• ACEStep 1.0: https://huggingface.co/datasets/Yi3852/ACEStep-Songs
• Mureka O1: https://huggingface.co/datasets/sleeping-ai/Mureka-384K

License and publishing notes

Before publishing to Hugging Face, please ensure you have the right to distribute all audio and metadata content.
If needed, add explicit license and attribution details here.