README.md

metadata

license: mit
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 22050
    - name: output
      dtype: float64
    - name: channel
      dtype: float64
    - name: dataset
      dtype: string
  splits:
    - name: train
      num_bytes: 2656554344.8
      num_examples: 15050
  download_size: 2656134575
  dataset_size: 2656554344.8
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
language:
  - en
tags:
  - climate
size_categories:
  - 10K<n<100K
pretty_name: Nature Sound Generation

A dataset from downloaded youtube videos of nature sounds. Each audio clip is two seconds long.