MIDI Music Collection for Symbolic AI Training

A collection of over 6.74 million unique and deduplicated MIDI files curated for music information retrieval and AI training. The dataset was created by 'projectlosangeles' and was last updated in December 2025. It includes normalized MIDI data and comprehensive metadata for symbolic music analysis.

Use Cases

• Train music generation models on the 6.74M normalized MIDI files to produce new symbolic compositions.
• Develop music information retrieval systems using the dataset's comprehensive metadata for tasks like genre classification or style transfer.
• Benchmark symbolic music AI algorithms for tasks like melody extraction or chord recognition against this large-scale collection.
• Perform music discovery and analysis by leveraging the de-duplicated and normalized structure of the MIDI files.

Strengths

• Contains over 6.74 million unique MIDI files.
• Files are de-duplicated and normalized for consistency.
• Includes comprehensive metadata for each file.

Limitations

• Specific row count, column details, and file size are unknown.
• License information is unspecified, which may restrict usage.
• Potential geographic or stylistic bias is unstated due to unknown source composition.

Provenance

Source

huggingface user 'projectlosangeles'

Collection Method

Curated collection, specific gathering method unknown.

Freshness

Last updated on 2025-12-28.