Published November 30, 2025 | Version 1.0
Dataset Restricted

The South African Next Voices Multilingual Speech Dataset [Compressed]

Authors/Creators

  • 1. ROR icon University of Pretoria
  • 2. ROR icon Technological University Dublin
  • 3. Penguide Advisory
  • 4. South African Centre for Digital Language Resources
  • 5. ROR icon Pennsylvania State University
  • 6. ROR icon North-West University

Description

Swivuriso: ZA-African Next Voices-Compressed

Swivuriso is a large-scale multilingual speech dataset targeting over 3000 hours of audio across 7 South African languages. The dataset is developed to support Automatic Speech Recognition (ASR) and inclusive speech technologies for low-resource African languages. It combines both scripted and unscripted speech, collected through ethical, community-centered processes.

Dataset Paper: ArXiv - Work in Progress

This is a compressed version of the original dataset https://huggingface.co/datasets/dsfsi-anv/za-african-next-voices/

⚠️ IMPORTANT: Visit the original dataset for full details

Language Coverage

Language Target Hours Released
isiZulu 500 ▇▇▇▇▇▇▇▇▇▇ 100%
isiXhosa 500 ▇▇▇▇▇▇▇▇▇▇ 100%
Sesotho 500 ▇▇▇▇▇▇▇▇▇▇ 100%
Setswana 500 ▇▇▇▇▇▇▇▇▇▇ 100%
Xitsonga 500 ▇▇▇▇▇▇▇▇▇▇ 100%
isiNdebele 250 ▇▇▇▇▇▇▇▇▇▇ 100%
Tshivenda 250 ▇▇▇▇▇▇▇▇▇▇ 100%

Use Restriction:

The persons whose voices are included in this dataset, and the creators and owners of this dataset* do not give consent in any manner or form to, and strictly prohibit any use of this dataset for any form of text-to-speech (TTS), voice cloning, voice synthesis, or any technology or activity intended to replicate, mimic or generate human voices or any technology or activity resulting in the replication, mimicry or generation of human voices.

This dataset includes scripted and unscripted speech across various domains such as agriculture, health, finance, sports, transport, culture, society, and general topics. It is primarily designed for use in automatic speech recognition (ASR) tasks.

Use of this dataset for any form of text-to-speech (TTS), voice cloning, voice synthesis, or any technology intended to replicate or generate human voices is strictly prohibited.

These restrictions are in place until further notice.

Citations

If you use Swivuriso in your work, please cite both of the below:

 

Dataset

 
@dataset{za-african-next-voices-2025,
  title     = {The South African Next Voices Multilingual Speech Dataset},
    author       = {Marivate, Vukosi and
                  Olaleye, Kayode and
                  Mundia, Sitwala and
                  Bakainga, Andinda and
                  Netshifhefhe, Unarine Leo and
                  Milanzie, Mahmooda and
                  Mogale, Hope and
                  SINDANE, THAPELO and
                  Abdulrasaq, Zainab and
                  Mokgosi, Kesego and
                  Okorie, Chijioke and
                  van Wyk, Nia Zion and
                  Morrissey, Graham and
                  Dunbar, Dale and
                  Smit, Francois and
                  Chidi, Tsosheletso and
                  Mabuya, Rooweither and
                  Bukula, Andiswa and
                  MLAMBO, RESPECT and
                  Macucwa, Solomon Tebogo and
                  Abdulmumin, Idris and
                  Rananga, Seani},
  url2      = {https://github.com/dsfsi/za-african-next-voices},
  url3      = {https://www.dsfsi.co.za/za-african-next-voices/},
  year      = {2025},
  type      = {dataset},
  publisher = {Zenodo},
  version   = {1.0},
  doi       = {10.5281/zenodo.17776289},
  url       = {https://doi.org/10.5281/zenodo.17776289},
}

 

Research Paper

 

Will be available soon.

 

Files

Restricted

The record is publicly accessible, but files are restricted. <a href="https://zenodo.org/account/settings/login?next=https://zenodo.org/records/17776290">Log in</a> to check if you have access.

Additional details

Software

Repository URL
https://github.com/dsfsi/za-african-next-voices/