Swivuriso: ZA-African Next Voices

Vukosi Marivate (co-PI) Kayode Olaleye (co-PI) Sitwala Mundia Unarine Netshifhefhe Nia Zion van Wyk Mahmooda Milanzie Tsholofelo Mogale Chijioke Okorie Thapelo Sindane Andinda Bakainaga Graham Morrissey Dale Dunbar Franscois Smit Tsosheletso Chidi Rooweither Mabuya Andiswa Bukula Respect Mlambo Tebogo Macucwa
UP Emblem
DSFSI Logo
Project Links
Hugging Face Hugging Face za-next-voices-2025 @DSFSI_Research
Dataset Size: TBC

Swivuriso: ZA-African Next Voices

⚠️ IMPORTANT: Work in Progress

This dataset is not final. Releases and updates will continue until end of September 2025.

  • Users must regularly check this repository for new versions, corrections, and updates.
  • For attribution, benchmarking, and publications, always reference the latest available information (as of September 2025 or later).
  • Do not cite or benchmark against old/incomplete versions. Use only the most recent release and documentation.

We appreciate your patience while we expand and improve this dataset. Stay updated by watching this repo.

About the project

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

Partners, Funders & Supporters

Way With Words (Partner), Data Science Law Lab @ UP (Partner), Gates Foundation (Funder), Meta (Funder)

Data Sources

Vukuzenzele Newspaper [Website][Data Repo], Wikipedia, African Wordnet, GrainSA, Agricultural Research Council, SADILAR, Masakhane