Intron, a Nigerian artificial intelligence startup, has expanded its speech recognition platform, Sahara, to support 57 languages, adding 24 new ones in a move aimed at improving how voice technologies understand African languages, accents, and multilingual conversations.
According to the firm, the newly added languages—including Hausa, Swahili, isiZulu, Yoruba, Kinyarwanda, Twi, Igbo, isiXhosa, Amharic, Luganda, Oromo, Shona, and Wolof—were selected based on commercial demand from enterprise customers using the platform.
“We curated datasets of African voices, combining publicly available datasets with our in-house collection, and made them available so anyone can test global models on African speech,” Tobi Olatunji, founder and CEO of Intron, told TechCabal.
The expansion underscores a growing push to build voice technology that reflects how people actually speak across the continent, particularly in sectors such as healthcare, legal services, finance, and telecommunications where accurate transcription and voice-based systems are increasingly important.
The Problem: Why voice recognition struggles with African languages
Although speech recognition technology has advanced rapidly in recent years, most global systems were designed primarily around English and other widely digitised languages. As a result, they often perform poorly when applied to African speech patterns.
Africa is home to roughly 2,000 languages, many of which have limited written resources or the large structured datasets needed to train modern artificial intelligence systems. This lack of training data has created a major barrier for companies building voice-based tools for the continent.
The challenge is compounded by Africa’s linguistic diversity and the prevalence of code-switching—where speakers alternate between languages within the same conversation. In many countries, people routinely mix local languages with English or French, making it difficult for conventional speech recognition systems to accurately process conversations.
These limitations have slowed the adoption of voice-based digital services across industries such as healthcare, finance, and telecommunications, where accurate transcription and voice interaction tools are increasingly important.
The lack of digitised African languages also creates a communication barrier. For example, someone in northern Nigeria with little or no English literacy may struggle to interact with digital platforms or automated systems to access essential services such as healthcare.
The Solution: How Intron’s Sahara voice AI understands African speech
Intron’s Sahara platform is designed specifically to address these challenges by training speech recognition models on African voice data.
The upgraded Sahara v2 platform supports 57 languages, including 23 African languages, and can recognise more than 500 distinct African accents. According to the company, the system was trained using more than 14 million audio clips totaling over 50,000 hours from more than 40,000 speakers across 30 African countries.
One of the key features introduced in the update is what Intron describes as the world’s first bilingual Swahili-English automatic speech recognition model capable of handling code-switching, a common occurrence in real-world conversations across Africa.
“A doctor may ask questions in English, a patient replies in Swahili, then switches back to English, all within the same exchange,” Olatunji explained. “Many monolingual systems struggle in those scenarios.”
The company also introduced its first local-language text-to-speech model in Hausa, enabling voice assistants, call centre bots, and digital health tools to interact with users in local languages.
Additionally, Sahara can run fully offline through a partnership with Nvidia, using Nvidia Jetson Edge devices to support organisations operating in areas with limited internet connectivity.
But Intron is not the only company trying to address the challenge of limited African language data for voice AI.
Google in February 2026 launched WAXAL, a large-scale open-source speech dataset aimed at improving AI technologies for Sub-Saharan African languages.
WAXAL was trained on more than 11,000 hours of recorded speech compiled from over two million individual recordings. It covers 21 Sub-Saharan African languages, including Hausa, Yoruba, Luganda, and Acholi.
Why It Matters: Voice AI could expand digital access across Africa
Improving speech recognition for African languages could significantly expand how people interact with technology across the continent.
Voice-based systems have the potential to make digital services more accessible to millions of Africans who may not be comfortable using text-based interfaces or who speak languages that are poorly supported by mainstream software.
In healthcare, voice transcription tools can help doctors document consultations more quickly, reduce administrative burdens, and improve patient records. In call centres and financial services, multilingual voice bots could enable companies to communicate more effectively with customers in local languages.
More broadly, building AI infrastructure tailored to African speech could help local developers create new digital products, stimulate innovation, and generate jobs in areas such as data collection, AI engineering, and language technology.
“A lot of people actually don’t know how to read and write on the continent,” Diack Abdoulaye Diack, program manager at Google Research, told TechCabal. “Voice is basically the gateway to technology.”
As the global speech and voice recognition market continues to expand, platforms like Sahara could play a critical role in ensuring that African languages—and the millions of people who speak them—are fully represented in the next generation of artificial intelligence systems.