anonymous-vivoice34/ViVoice34

ViVoice-34: Vietnamese Speech Dataset

Dataset Description

ViVoice-34 is a Vietnamese speech dataset featuring recordings from speakers across provinces of Vietnam. This repository contains a preview subset with playable WAV samples and metadata.

The preview data is stored as Parquet with the audio column encoded as Hugging Face Audio, so Dataset Viewer and Data Studio can render an audio player instead of plain file paths.

Data Fields

Field	Type	Description
audio	Audio	Playable waveform data
file_name	string	Original sample path inside the preview set
transcript	string	Vietnamese transcript
speaker_id	string	Anonymized speaker identifier
province	string	Province of origin
gender	string	Speaker gender
age_group	string	Speaker age group
duration_s	float	Duration in seconds
total_word	int	Transcript word count
province_code	string	Source/topic code

speaker_name is intentionally excluded from the preview metadata.

Usage

from datasets import load_dataset

ds = load_dataset("anonymous-vivoice34/ViVoice34")
sample = ds["train"][0]

print(sample["audio"])
print(sample["transcript"])

Citation

@dataset{vivoice34_2026,
  title={ViVoice-34: Vietnamese Speech Dataset from 34 Provinces},
  year={2026},
  url={https://huggingface.co/datasets/anonymous-vivoice34/ViVoice34}
}

License

This dataset is released under the CC-BY-4.0 license.