Nano-Speech Dataset

A compact, general-purpose speech dataset designed for lightweight ASR and speech processing tasks.

Dataset Description

• Domain: General speech
• Format: Compressed ZIP archive
• Size: ~25 GB
• Access: Public (no authentication required)

Contents

The dataset is provided as a single archive:

• nano_speech.zip — Contains speech audio files, ready for use in training and evaluation pipelines

Usage

from huggingface_hub import snapshot_download
snapshot_download("edwixx/nano-speech")

Or load directly with 🤗 Datasets (if a config is defined):

from datasets import load_dataset
dataset = load_dataset("edwixx/nano-speech")

Intended Uses

• Automatic Speech Recognition (ASR) model training
• Speech embedding and representation learning
• Evaluating compact speech models on diverse audio
• Lightweight / on-device speech processing research

License

This dataset is released under the MIT License.