Fact · Speech recognition
105,829audio files
Version 0.02 of Google's Speech Commands dataset contains 105,829 audio files, each a one-second recording of a single spoken English word or background noise.
Checked against its source on October 17, 2026
In the source's own words
“each containing a single spoken English word or background noise. These words are from a small set of commands ... Version 0.02 of the data set (configuration "v0.02" ) was released on April 11th 2018 and contains 105,829 audio files.”
speech_commands dataset card, Hugging Face, dataset card for Google's Speech Commands dataset, 2018, Dataset description and Dataset Summary.
Open the original- What it measures
- The dataset was built for keyword spotting, the task of detecting a few command words. Version 0.01, released in 2017, had 64,727 audio files for thirty words. Version 0.02 added five more words.
- What it does not show
- The count is audio files, not speakers or hours. The clips are single words, not sentences, so the dataset teaches a model to recognize commands, not to transcribe dictation.