Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?


Failed to load latest commit information.
Latest commit message
Commit time
August 12, 2020 10:28
December 1, 2021 13:11
October 12, 2021 11:35
July 20, 2022 19:13
June 21, 2016 13:48
August 12, 2020 10:28
October 17, 2021 14:57

Free Spoken Digit Dataset (FSDD)


A simple audio/speech dataset consisting of recordings of spoken digits in wav files at 8kHz. The recordings are trimmed so that they have near minimal silence at the beginnings and ends.

FSDD is an open dataset, which means it will grow over time as data is contributed. In order to enable reproducibility and accurate citation the dataset is versioned using Zenodo DOI as well as git tags.

Current status

  • 6 speakers
  • 3,000 recordings (50 of each digit per speaker)
  • English pronunciations


Files are named in the following format: {digitLabel}_{speakerName}_{index}.wav Example: 7_jackson_32.wav

How to use with Hub

A simple way of using this dataset is with Activeloop's python package Hub!

First, run pip install hub (or pip3 install hub).

import hub
ds = hub.load("hub://activeloop/spoken_mnist")

# check out the first spectrogram, it's label, and who spoke it!
import matplotlib.pyplot as plt
plt.title(f"{ds.speakers[0].data()} spoke {ds.labels[0].numpy()}")

# train a model in pytorch
for sample in ds.pytorch():
    # ... model code here ...

# train a model in tensorflow
for sample in ds.tensorflow():
    # ... model code here ...

available tensors can be shown by printing dataset:

# prints: Dataset(path='hub://activeloop/spoken_mnist', tensors=['spectrograms', 'labels', 'audio', 'speakers'])

For more information, check out the hub documentation.


Please contribute your homemade recordings. All recordings should be mono 8kHz wav files and be trimmed to have minimal silence. Don't forget to update with the speaker meta-data.

To add your data, follow the recording instructions in acquire_data/ and then run to make your files.

Metadata contains meta-data regarding the speakers gender and accents.

Included utilities Trims silences at beginning and end of an audio file. Splits an audio file into multiple audio files by periods of silence. A simple class that provides an easy to use API to access the data. Used for creating spectrograms of the audio data. Spectrograms are often a useful pre-processing step.


The test set officially consists of the first 10% of the recordings. Recordings numbered 0-4 (inclusive) are in the test and 5-49 are in the training set.

Made with FSDD

Did you use FSDD in a paper, project or app? Add it here!

External tools


Creative Commons Attribution-ShareAlike 4.0 International