Google Releases WAXAL for African Language Speech

Google has launched WAXAL, an open-source voice database created to support speech technologies in African languages. The project took three years to complete and involved universities and research groups across the continent. It responds to a long-standing lack of usable speech data, which has limited the development of voice tools in sub-Saharan Africa.

The dataset is available free of charge on Hugging Face. It contains more than 11,000 hours of recorded speech drawn from almost two million audio files and covers 21 African languages, including Hausa, Yoruba, Luganda, Acholi, Swahili, Igbo, and Fulani.

African Institutions Led Data Collection

African partners managed most of the data collection. Makerere University and University of Ghana coordinated work on 13 languages, while Rwanda’s Digital Umuganda added five more. Regional studios handled recording and quality control, and the African Institute for Mathematical Sciences contributed to the preparation of multilingual datasets planned for future releases.

Under this arrangement, participating institutions keep ownership of their language data while releasing it under an open licence.

Practical Uses for Voice-Based Services

WAXAL includes around 1,250 hours of transcribed speech for speech recognition and more than 20 hours of studio recordings for speech generation. Developers can use the dataset to build tools such as voice assistants, dictation software, and automated phone services. These applications are especially relevant in health, education, and agriculture, where voice systems can improve access for users with limited reading ability.

“This dataset gives students, researchers, and entrepreneurs the materials they need to build technology in their own languages and for their own communities,” said Aisha Walcott-Bryant, head of Google Research Africa.

Part of a Wider Push for African Language Tools

The release of WAXAL follows similar efforts across the continent. In 2025, Nigeria introduced N-ATLAS, an open-source system that transcribes Yoruba, Hausa, Igbo, and Nigerian English. Several African startups are also working on speech and translation products designed for local use rather than global markets.

Sub-Saharan Africa is home to more than 2,000 languages, yet only a small number currently have enough data to support speech and language software. This gap continues to exclude many users from voice-based services that are now common elsewhere. As Joyce Nakatumba-Nabende of Makerere University has noted, language tools only work in Africa when they reflect local languages and everyday realities.

Read more about google HERE