Google, alongside leading African universities and community organizations, has launched WAXAL, a groundbreaking open speech dataset aimed at making artificial intelligence finally understand and speak African languages.

The project is a big step toward closing one of AI’s most overlooked gaps: language access. While voice assistants and speech-powered tech are everywhere globally, most of Africa’s 2,000+ languages have been left out due to a lack of quality speech data. That exclusion affects hundreds of millions of people. Built over three years with funding from Google, the dataset includes 1,250 hours of transcribed natural speech and over 20 hours of studio-quality recordings, covering 21 Sub-Saharan African languages such as Yoruba, Hausa, Luganda, Acholi, Igbo, Swahili, and more. In total, it lays the foundation for AI tools that could reach over 100 million speakers across the continent.

“This is about empowerment,” said Aisha Walcott-Bryant, Head of Google Research Africa. “WAXAL gives African students, researchers, and builders the tools they need to create technology on their own terms, in their own languages — from education platforms to voice-powered services that can unlock real economic opportunities.”

Instead of extracting data, Google partnered directly with African institutions that led the work themselves. Universities and community groups such as Makerere University (Uganda), University of Ghana, and Digital Umuganda (Rwanda) handled data collection, with technical support from Google. Importantly, the institutions retain full ownership of the data, setting a new standard for ethical and equitable AI development.

“For AI to truly work in Africa, it has to understand our languages and realities,” said Joyce Nakatumba-Nabende, Senior Lecturer at Makerere University. “WAXAL has already strengthened local research and inspired student-led innovation here in Uganda.”

At the University of Ghana, the project became a movement of its own.