---
dataset_info:
  features:
  - name: audio
    dtype: audio
  splits:
  - name: train
    num_bytes: 47994025010.08
    num_examples: 108880
  download_size: 49944500339
  dataset_size: 47994025010.08
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This dataset is sampled from LibriVox audio (135+ hours). The purpose of this dataset is for training neural audio codecs or autoencoder to reconstruct high-fidelity audio data.