Word-Level Time-Aligned Lyrics for Multilingual Songs

79 songs across multiple genres and languages feature lyrics aligned to audio on a word-by-word basis with start and end times. Created by jamendolyrics, this dataset serves primarily as a benchmark for automatic lyrics alignment tasks. It was last updated in March 2025.

Use Cases

• Benchmarking automatic lyrics alignment models using word-level start and end timestamps.
• Training models to synchronize multilingual lyrics text with corresponding audio segments.
• Analyzing genre-specific patterns in lyrics timing and structure across different languages.
• Developing music players or karaoke systems that display synchronized, word-by-word lyrics.

Strengths

• 79 songs provide a multi-genre, multilingual collection for alignment tasks.
• Word-level time alignment with start and end timestamps offers precise ground truth.

Limitations

• Small sample size of 79 songs limits statistical power for model training.
• Primary focus on alignment benchmarking may limit utility for other tasks like transcription.

Provenance

Source

jamendolyrics

Collection Method

Lyrics time-aligned to music on a word-by-word level.

Freshness

Last updated March 2025.

Geography

Region tags suggest US focus, but multilingual content indicates broader coverage.