Karaoke Songs Long

A dataset of long-form karaoke instrumental tracks with auto-generated captions describing the musical characteristics of each song. This repository contains audio-to-text caption pairs for long-duration karaoke versions of popular songs.

What it is

This dataset provides instrumental karaoke tracks (long versions โ€” extended or full-length) paired with descriptive captions. Each caption details the genre, tempo (BPM), instrumentation, mood, and potential use cases of the audio.

Dataset Contents

The captions.csv file contains the following fields:

Field Description
original_filename Original audio filename (artist - song title - karaoke version)
cleaned_filename Cleaned/standardized filename
caption Auto-generated textual description of the audio track

Included Artists (sampling)

  • 50 Cent โ€” "In Da Club" (Karaoke)
  • 5 Seconds of Summer โ€” "Amnesia", "Beside You", "Don't Stop", "Good Girls", "Heartbreak Girl", "She's Kinda Hot", "Youngblood"
  • A Great Big World & Christina Aguilera โ€” "Say Something"
  • Adele โ€” "Chasing Pavements", "Don't You Remember", "Easy On Me"
  • (and many more)

Caption Features

Each caption includes information about:

  • Genre: e.g., pop rock, hip hop, indie rock, folk, electronic, soul/R&B
  • Tempo: BPM range (e.g., 72 BPM, 84 BPM, 120 BPM, 160 BPM)
  • Time Signature: typically 4/4
  • Instrumentation: guitars, bass, drums, synths, piano, etc.
  • Mood/Atmosphere: energetic, reflective, calming, introspective, uplifting
  • Use Cases: background music, workout, documentaries, soundtracks, commercials

Usage

This dataset is suitable for:

  • Audio captioning and music description models
  • Text-to-audio generation training
  • Music information retrieval (MIR) research
  • Karaoke applications and instrument separation
  • Audio retrieval from natural language queries

License

Please check the individual song copyrights and terms. This repository contains descriptive captions only; actual audio files (if present) are instrumental karaoke tracks and may be subject to additional licensing.

Citation

@misc{edwixx-karaoke_songs_long,
  author = {Anurag Kanade},
  title = {karaoke_songs_long},
  year = {2026},
  publisher = {Hugging Face},
  journal = {Hugging Face Hub},
  howpublished = {\url{https://huggingface.co/edwixx/karaoke_songs_long}}
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support