Published May 26, 2021 | Version 2.1
Dataset Open

Clotho dataset

  • 1. Audio Research Group, Faculty of Information Technology and Communication Sciences, Tampere University


Clotho is an audio captioning dataset, now reached version 2. Clotho consists of 6974 audio samples, and each audio sample has five captions (a total of 34 870 captions). Audio samples are of 15 to 30 s duration and captions are eight to 20 words long. 

Clotho is thoroughly described in our paper:

K. Drossos, S. Lipping and T. Virtanen, "Clotho: an Audio Captioning Dataset," IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 2020, pp. 736-740, doi: 10.1109/ICASSP40776.2020.9052990.

available online at: and at: 

If you use Clotho, please cite our paper.


To use the dataset, you can use our code at: 


These are the files for the development, validation, and evaluation splits of Clotho dataset. 


== Changes in version 2.1 ==

In version 2.1 of Clotho, we fixed some files that were corrupted from the compression and transferring processes (around 150 files) and we also replaced some characters that were illegal for most filesystems, e.g. ":" (around 10 files). 

Please use this version for your experiments. 

== Changes in version 2 ==

In version 2 of Clotho, there are audio files added in the development split and a new validation split is added. There are no changes in the evaluation split. 


  • Now there are 3840 audio files in the development split. In Clotho version 1, there were 2893 audio files. Now, 947 new audio files are added. 
  • There are 1046 new audio files in the validation split. 

All new captions are treated as in version 1 of Clotho, i.e. having word consistency, no named entities, no speech transcription, and no hapax legomena between splits (i.e. words appearing only in one of the splits). 


== Usage ==

To use the dataset you have to:

  1. Download the audio files: clotho_audio_development.7z,clotho_audio_validation.7z, and clotho_audio_evalution.7z
  2. Download the files with the captions: clotho_captions_development.csv, clotho_captions_validation.csv, and clotho_captions_evaluation.csv
  3. Download the files with the associated metadata: clotho_metadata_development.csv, clotho_metadata_validation.csv, and clotho_metadata_evaluation.csv
  4. Extract the audio files
  5. Then you can use each audio file with its corresponding captions


== License ==

The audio files in the archives:

  • clotho_audio_development.7z,
  • clotho_audio_validation.7z, and
  • clotho_audio_evalution.7z

and the associated meta-data in the CSV files:

  • clotho_metadata_development.csv
  • clotho_metadata_validation.csv
  • clotho_metadata_evaluation.csv

are under the corresponding licences (mostly CreativeCommons with attribution) of Freesound [1] platform, mentioned explicitly in the CSV files for each of the audio files. That is, each audio file in the 7z archives is listed in the CSV files with the meta-data. The meta-data for each file are: 

  • File name
  • Keywords
  • URL for the original audio file
  • Start and ending samples for the excerpt that is used in the Clotho dataset
  • Uploader/user in the Freesound platform (manufacturer)
  • Link to the licence of the file

The captions in the files:

  • clotho_captions_development.csv
  • clotho_captions_validation.csv
  • clotho_captions_evaluation.csv

are under the Tampere University licence, described in the LICENCE file (mainly a non-commercial with attribution licence). 


== References ==
[1] Frederic Font, Gerard Roma, and Xavier Serra. 2013. Freesound technical demo. In Proceedings of the 21st ACM international conference on Multimedia (MM '13). ACM, New York, NY, USA, 411-412. DOI:



Files (7.1 GB)

Name Size Download all
4.5 GB Download
1.2 GB Download
1.3 GB Download
1.3 MB Preview Download
362.0 kB Preview Download
367.6 kB Preview Download
830.8 kB Preview Download
225.3 kB Preview Download
224.8 kB Preview Download
1.9 kB Download

Additional details

Related works

Is supplement to
Conference paper: (URL)
Is supplemented by
Software: (URL)


EVERYSOUND – Computational Analysis of Everyday Soundscapes 637422
European Commission


  • Frederic Font, Gerard Roma, and Xavier Serra. 2013. Freesound technical demo. In Proceedings of the 21st ACM international conference on Multimedia (MM '13). ACM, New York, NY, USA, 411-412. DOI: