MusicEval: 2,748 Generated Music Clips with Expert Ratings for TTM Evaluation

MusicEval is a dataset of 2,748 generated music clips with a total duration of 16.62 hours, created by BAAI. It is the first generative music assessment dataset designed to address text-to-music evaluation challenges. The dataset was last updated on August 18, -2025.

Use Cases

• Benchmarking text-to-music generation models based on expert ratings.
• Training automatic music evaluation metrics based on the expert-rated clips.
• Studying the relationship between text prompts and generated music based on the dataset's design.

Strengths

• Contains 2,748 generated music clips.
• Provides expert ratings for text-to-music evaluation.
• Total audio duration is 16.62 hours.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

BAAI

Collection Method

Likely contains music clips generated by AI models and rated by experts, as described.

Time Range

null

Freshness

Last updated 2025-08-18 17:52:54; freshness should be verified.

Geography

null