Infographic showing Google's WAXAL speech project details covering 21 African languages with 11,000 hours of speech data

WAXAL: Bringing Voice to Africa’s Languages

What It Is
WAXAL (from the Wolof word for “speak”) is Google’s massive open-source speech dataset covering 21 Sub-Saharan African languages. Released February 2, 2026, it contains over 11,000 hours of speech from nearly 2 million recordings.

The Core Achievement
The dataset provides three critical components:

  • 1,250 hours of transcribed speech for automatic speech recognition (ASR)
  • 20+ hours of studio-quality recordings for text-to-speech (TTS)
  • Natural conversational content collected by native speakers

Languages included range from widely-spoken ones like Hausa, Swahili, and Yoruba to smaller languages like Luganda, Acholi, and Kikuyu.

Why This Matters

Addressing the Data Desert
While Africa has over 2,000 languages, most voice technologies (like Siri or Alexa) have completely ignored them due to lack of training data. WAXAL directly tackles this “data poverty” that has excluded over 100 million people from voice-enabled technology.

Empowering Local Innovation
Unlike typical “data extraction” projects, WAXAL uses a groundbreaking ownership model. Partner institutions like Makerere University (Uganda) and the University of Ghana retain ownership of the data they collected. This means local researchers and startups can build AI tools—for healthcare, education, or agriculture—without depending on foreign tech giants or waiting for permission.

Real-World Impact
This enables practical applications for populations where literacy rates may be low but mobile phone usage is high. Farmers could get agricultural advice in their native language, students could access educational content verbally, and healthcare information could reach remote communities more effectively.

Preserving Linguistic Heritage
By documenting authentic native-speaker conversations rather than scripted readings, WAXAL also serves as a preservation tool for linguistic diversity at a time when many languages face endangerment.

The Collaborative Approach
The three-year project was funded by Google but led by African institutions. Contributors recorded themselves speaking naturally in their native languages, creating a dataset that captures real-world usage patterns rather than artificial lab conditions.

The complete dataset is available on Hugging Face, making it freely accessible to researchers and developers worldwide while ensuring African institutions maintain control over their contributions.

This represents a significant step toward more equitable AI development—where technology serves all languages and communities, not just the economically dominant ones.