---
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
---