• Google launches WAXAL open-source African language voice database

• Dataset offers 11,000 hours across 21 languages, free on Hugging Face

• Project aims to boost voice AI access, led by African institutions

Google has officially launched WAXAL, an open-source voice database designed to support the development of artificial intelligence (AI) technologies capable of understanding and reproducing African languages. The project was developed over three years in partnership with institutions across the continent. It aims to address a long-standing shortage of linguistic data, widely seen as a major obstacle to the growth of voice AI in sub-Saharan Africa.

Now available for free on the Hugging Face platform, WAXAL contains more than 11,000 hours of voice recordings drawn from nearly two million audio files. The database covers 21 African languages, including Hausa, Yoruba, Luganda, Acholi, Swahili, Igbo, and Fulani.

African partners led the data collection effort. Makerere University in Uganda and the University of Ghana coordinated work on 13 languages, while the Rwandan initiative Digital Umuganda contributed five additional languages. Regional studios also helped produce high-quality recordings. The African Institute for Mathematical Sciences (AIMS) took part in developing multilingual corpora for future versions.

Built as a foundational resource, WAXAL provides around 1,250 hours of transcribed speech for automatic speech recognition. It also includes more than 20 hours of studio recordings intended for speech synthesis. The goal is to enable the development of voice-based applications, such as voice assistants, dictation tools, or public services accessible to people with limited literacy, particularly in the fields of health, education, and agriculture.

“This dataset provides the critical foundation for students, researchers, and entrepreneurs to build technology on their own terms, in their own languages, finally reaching over 100 million people,” said Aisha Walcott-Bryant, head of Google Research Africa.

The launch of WAXAL comes as efforts to advance African linguistic technologies continue to expand. In 2025, Nigeria presented N-ATLAS, an open-source linguistic model capable of transcribing speech in Yoruba, Hausa, Igbo, and Nigerian English. In the private sector, African startups are also developing voice recognition and translation solutions aimed at local needs.

Sub-Saharan Africa has more than 2,000 languages, but only a small number currently have the resources needed for natural language processing. This limits access to voice technologies for millions of people, even as such tools become widespread in other regions of the world.

Under the partnership model adopted, the African institutions that contributed to the data collection retain ownership of the corpora while making them available under an open license. For Joyce Nakatumba-Nabende, a professor and researcher at Makerere University, “For AI to have a real impact in Africa, it must speak our languages and understand our contexts.”

Fiacre E. Kakpo