metadata
dataset_info:
features:
- name: audiocap_id
dtype: string
- name: youtube_id
dtype: string
- name: start_time
dtype: string
- name: audio
dtype:
audio:
sampling_rate: 48000
- name: caption
dtype: string
splits:
- name: train
- name: validation
- name: test
license: mit
task_categories:
- text-to-speech
language:
- en
multilinguality:
- monolingual
tags:
- youtube
- captions
pretty_name: AudioCaps
size_categories:
- 10K<n<100K
source_datasets:
- original
paperswithcode_id: audiocaps