Skip to content

Dataset · laion

1,175 hours of synthetic vocals with lyrics from LAION

laion/singing-voice-1k

Collection of 146,000 English audio clips of generated singing with lyrics and metadata.

What's inside

English audio dataset from LAION containing synthetic vocals generated by the LeVo 2 model in vocal mode. The release includes 146,640 clips with a total duration of 1,174.92 hours, packed into 128 TAR archives in WebDataset format. Each FLAC audio file comes with English song lyrics and generation metadata.

What you can do with it

  • Training and fine-tuning singing synthesis models (text-to-audio / singing voice synthesis)
  • Training speech recognition and lyric recognition (ASR) systems on vocal tracks
  • Acoustic characteristics research and synthetic vocal detection
Tasks
Text → audio, Automatic speech recognition
Languages
English
Size
100K–1M rows
Downloads this month
801
Likes
2

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("laion/singing-voice-1k")