Speech Data for 21 African Languages Is Now Open Access. What Does That Mean for Us? | Metaglossia: The Translation World | Scoop.it

"Google has released WAXAL, an open speech dataset covering 21 Sub-Saharan African languages including Acholi, Hausa, Igbo, Luganda, Shona, Swahili, and Yoruba. The dataset contains over 11,000 hours of speech data from nearly 2 million recordings, developed over three years in partnership with Makerere University in Uganda, the University of Ghana, Digital Umuganda in Rwanda, and other African organizations. The name WAXAL comes from the Wolof word for “speak,” and the collection is now available here.


The dataset includes approximately 1,250 hours of transcribed speech for automatic speech recognition and over 20 hours of studio recordings for text-to-speech synthesis. Participants were asked to describe pictures in their native languages to capture natural speech patterns, while professional voice actors provided high-quality audio recordings. Google partnered with Media Trust and Loud n Clear for voice recording work, with the framework allowing partner organizations to retain ownership of the data they collected while making it available to researchers globally.


For African writers, artists, and technologists, the implications are worth examining carefully. Speech recognition and text-to-speech technologies determine whose voices get translated into text, whose stories can be transcribed, and which languages digital assistants will understand. The availability of this data could enable African developers to build tools on their own terms rather than waiting for Silicon Valley to decide which markets merit investment.


The potential applications for writers and literary communities are tangible. Imagine transcription tools that actually work for interviews conducted in Igbo or Yoruba, making the labor of documentation less extractive. Audiobook production in Swahili or Hausa without requiring English as an intermediary language. Voice-based storytelling platforms where oral traditions can be archived and shared without being filtered through colonial languages. Translation tools built by and for African language speakers rather than optimized for European language pairs. The dataset provides the foundation for developers to build them now.


That said, questions remain about how this data will be used and who ultimately benefits. Large tech companies have a pattern of extracting resources from the continent, whether raw materials or, increasingly, cultural and linguistic data, while the tools built from these resources often serve markets elsewhere first. The dataset’s open license means anyone can use it, but whether African developers and researchers will have the computational resources and infrastructure to compete with well-funded corporations is another matter. WAXAL offers infrastructure that didn’t exist before, but infrastructure alone doesn’t redistribute power."
https://brittlepaper.com/2026/02/speech-data-for-21-african-languages-is-now-open-access-what-does-that-mean-for-us/
#Metaglossia
#metaglossia_mundus
#métaglossie