Google has introduced the WAXAL artificial intelligence speech dataset to support 21 African languages, a major effort to correct long‑standing language gaps in AI systems. The dataset is aimed at enabling voice technology development for more than 100 million speakers across Sub‑Saharan Africa.
The WAXAL dataset was developed in partnership with a coalition of African universities and research organisations. These partners played a central role in gathering, curating and validating the speech data that underpins the project.
The launch responds to a stark imbalance in global speech technology resources. Most voice‑enabled systems have been trained on data for a few dominant languages. In contrast, the over 2,000 languages spoken across Africa have largely been excluded from mainstream AI applications because of limited high‑quality speech data.
According to Google, the initiative started more than three years ago when researchers documented the deficiency of African language data in existing global speech datasets. That scarcity has inhibited the development of localised tools like voice assistants, speech recognition and translation for speakers outside Western language spheres.
Google describes WAXAL as a foundational resource for students, researchers and developers. The dataset is intended to fuel the creation of AI solutions that operate in local languages and reflect the realities of African users.
“The ultimate impact of WAXAL is the empowerment of people in Africa,” said Aisha Walcott‑Bryantt, Head of Google Research Africa. “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.”
The dataset consists of substantial speech resources. It offers approximately 1,250 hours of transcribed natural speech intended for training automatic speech recognition (ASR) models. In addition, there are over 20 hours of professional studio recordings designed to aid the development of high‑fidelity synthetic voices.
Languages covered include widely spoken tongues such as Hausa, Yoruba, Igbo and Swahili, as well as less widely represented languages like Acholi, Kikuyu, Lingala, Luganda and Shona, among others. This range is meant to ensure broad applicability across different regions and linguistic communities.
A defining aspect of the WAXAL project is the leadership of African institutions in the data collection process. Universities such as Makerere University in Uganda and the University of Ghana led the collection efforts for multiple languages, while organisations such as Digital Umuganda in Rwanda also played a key role. These institutions retain ownership of the data they helped compile.
This local ownership model contrasts with many global AI projects where data is gathered and controlled by external entities. By ceding data rights to local partners, the project aims to establish a framework for equitable, community‑led AI development.
Academics involved in the project stressed the value of high‑quality speech data for research and innovation. Joyce Nakatumba‑Nabende, a senior lecturer at Makerere University, noted that WAXAL enhances research capacity by providing the data needed to build contextually relevant speech technologies.
Prof. Isaac Wiafe of the University of Ghana highlighted the project’s ability to mobilise volunteers and spark innovation in sectors such as education, agriculture and health. The open resource reduces reliance on costly and time‑intensive in‑house dataset creation, which has been a barrier for many African innovators.
The potential impact of WAXAL extends beyond research labs and tech startups. When systems can understand and produce speech in local languages, they can significantly improve access to digital public services, interactive learning platforms and automated healthcare information. These applications can increase digital inclusion and participation in the AI economy.
By lowering the barriers to AI development in African languages, the project may also stimulate the growth of technology ecosystems across the continent. These ecosystems could support new business models, create jobs, and foster innovation that is grounded in local linguistic and cultural contexts.
