A final-year Mechanical Engineering student at the University of Lagos (UNILAG), Saheed Azeez, has developed a groundbreaking text-to-speech (TTS) model named YarnGPT. This AI-powered system is designed to read text aloud in a clear Nigerian-accented English, making it a significant step forward in voice technology tailored for African users. Unlike conventional TTS models that often struggle with local intonations, YarnGPT effectively captures the unique pronunciation and rhythm of Nigerian speech.
Azeez was inspired to build YarnGPT due to the lack of AI-powered speech models that support Nigerian and African accents. To train the model, he extracted audio and transcriptions from Nigerian movies, which provided rich linguistic data. Despite facing challenges with data quality and limited funding, he successfully developed the model using Python and machine learning libraries, investing approximately $150 of his personal savings.
The innovation has potential applications in various fields, including audiobooks, education, and assistive technology. YarnGPT could help improve accessibility for visually impaired individuals and make learning materials more engaging for Nigerian students. By offering a familiar accent, the model enhances comprehension and user experience, bridging the gap between technology and local culture.
Azeez intends to keep YarnGPT open-source, allowing developers and researchers to improve and expand on his work. However, he is already working on an upgraded version with enhanced accuracy and a wider range of Nigerian dialects. Balancing this project with his academic workload was challenging, but he remains committed to advancing AI solutions tailored for African users.
The Vice-Chancellor of UNILAG, Prof. Folasade Ogunsola, has praised Azeez’s innovation, commending his dedication to AI development. She expressed confidence that his work will inspire other students to explore cutting-edge technologies. As AI continues to shape global industries, YarnGPT represents a significant milestone in the localization of artificial intelligence for African communities.
