Skip to content

Dataset · Meranti

Audio recordings and text descriptions from Freesound

Meranti/CLAP_freesound

Part of the large LAION-Audio-630K corpus with English audio-text pairs.

What's inside

The repository contains two versions of the Freesound dataset, which is part of the well-known open LAION-Audio-630K corpus. Inside are audio files and their corresponding English text descriptions, with sample sizes ranging from 1 to 10 million records. Details on how the descriptions were created are provided in the card metadata.

What you can do with it

  • Train sound classification and acoustic event models
  • Train multimodal audio-text encoders (similar to CLAP)
  • Search and annotate sound effects using text queries
Tasks
Audio classification
Languages
English
Size
1M–10M rows
Downloads this month
16.2K
Likes
35

How to load it

from datasets import load_dataset

ds = load_dataset("Meranti/CLAP_freesound")