Redação Portal ERP

Aug 26, 2026

T|Fonte:?18px+

????Modo Noturno

3 min read

Intron AI, a Nigerian startup that builds speech recognition and voice generation models for African languages, has released Sahara v2.5, a version of its model that follows conversations in which speakers move between English and an African language and that can generate speech in African languages.

The previous version handled switching between Swahili and English alone. Once the model reached courtrooms, call centres and hospitals, that proved too narrow, because speakers move into the language they are most comfortable in mid-sentence and the system lost the thread. Intron says customer feedback from those deployments shaped the update.

Sahara v2.5 supports code-switching across about 20 African languages, treating a change of language as part of the conversation rather than as an error. The company says its engineering effort concentrated on holding accuracy at those switching points, where speech recognition models tend to degrade, and that performance benchmarks will be published separately.

Voice generation is the second addition. Synthetic speech has reached a usable standard in English while remaining difficult in many African languages. Intron says its model now produces speech in languages including Igbo and Hausa, and that the generated voice can change language partway through an exchange. A call centre could run a voice bot that meets a customer in the language that customer prefers, and the same capability applies to public information campaigns and government services.

One of the first deployments is at Meridian Hospital in Enugu, where doctors and patients speak mostly in Igbo. The hospital had been using Intron's English model for dictation, which meant a doctor finished a consultation and then recorded notes. Capturing the consultation itself removes that step and leaves the doctor editing a draft instead of writing one.

Intron began as a healthcare company and found its model in use elsewhere, among them call centres and government agencies, because it recognises African names and accents. It now sells through an API that lets other companies embed its speech and voice capabilities, while its direct sales team works donor-funded government partnerships and financial services. The company expects to reach profitability next year and declined to give revenue figures. Its models are trained on a mix of paid human recordings and a proprietary synthetic data system, for which Intron has filed a US patent application.

Data residency is handled case by case. Intron runs one project with a hospital in Port Harcourt where the model operates offline, and it offers a zero-retention option for customers who send audio to the cloud, under which the company deletes the material from its side once results are returned.

Share:

Redação Portal ERP

Editorial Team

Portal ERP's editorial team brings the latest news and analysis on technology and business management.