Nigeria’s AI startup Intron expands it’s speech platform to 57 languages

Intron, a Nigerian artificial intelligence startup has expanded its speech recognition platform, Sahara, to support 57 languages, adding 24 new ones as it scales operations across healthcare, legal, financial services, and telecommunications.

The upgraded Sahara v2, includes 23 African languages and supports more than 500 African accents. Newly added languages include Hausa, Swahili, Yoruba, Igbo, isiZulu, Kinyarwanda, Twi, isiXhosa, Amharic, and Wolof, among others. The company said language selection was driven by commercial demand.

Intron also introduces the world’s first bilingual Swahili-English automatic speech recognition model designed to handle code-switching, alongside new text-to-speech features and offline deployment options for enterprise users.

The development comes amid growing interest in voice technology across Africa, where most of the continent’s estimated 2,000 languages are primarily spoken rather than written. Industry projections show the global speech recognition market could reach $81.59 billion by 2032.

According to Intron, Sahara was trained with locally sourced voice data to better capture African accents, dialects, and speech patterns. The company’s internal benchmarks using African voice datasets showed the platform outperformed global systems such as Gemini, GPT-4, Whisper, ElevenLabs, and Azure by up to 64% in recognising African names, organisations, and locations.

It also reported 35% better performance with numbers, 20% stronger accuracy in noisy or multi-speaker environments, and 25% higher cross-domain accuracy across key sectors.

“We curated datasets of African voices and made them publicly testable so global models can be evaluated on African speech,” said Tobi Olatunji, founder and chief executive officer of Intron.

Founded in 2020 by Olatunji and Olakunle Asekun, the company initially built clinical documentation tools before expanding into broader voice infrastructure. Sahara now powers speech-to-text, text-to-speech, and voice authentication systems for enterprise and government clients.

Intron said it currently records consistent usage in at least six African countries, including Nigeria, Kenya, South Africa, Ghana, Rwanda, and Uganda.

Its clients include the Ogun State Judiciary and Audere, which uses the platform to transcribe WhatsApp voice messages across multiple accents.

Sahara v2 was built using more than 14 million audio clips, representing over 50,000 hours of speech from more than 40,000 speakers across 30 African countries. The company said much of its early medical speech data had to be created internally due to a lack of existing datasets.

The bilingual Swahili-English model was developed in partnership with Penda Health to address real-time language switching common in clinical settings. Intron said additional bilingual models for languages such as Yoruba, Hausa, Zulu, and Kinyarwanda are in development.

The company also launched a Hausa text-to-speech model aimed at powering multilingual voice applications for call centres, healthcare services, and financial platforms.

While Sahara is primarily cloud-based, Intron has introduced offline deployment through a partnership with Nvidia, enabling the system to run on edge devices in low-connectivity environments.

Intron said it complies with local data protection regulations and allows clients to choose between local and cloud data storage.

The company plans to raise $3 million later in 2026 to expand language coverage and further develop bilingual and sector-specific models.