Nigerian artificial intelligence startup Intron AI has launched Sahara v2.5, an upgraded voice AI model designed to better understand and generate multilingual conversations across African languages.
The new release introduces expanded code-switching capabilities, allowing the model to follow conversations in which speakers move between English and African languages — a common feature of everyday communication across the continent.
Intron AI says the model can also generate speech in African languages and switch between languages within the same conversation.
The upgrade builds on the company’s previous Sahara v2 model, which had more limited multilingual capabilities.
Solving a Major African Language Challenge
Across Africa, many people communicate using a mixture of English, French or Portuguese and indigenous languages.
A customer speaking to a call centre agent, for example, may begin in English before switching to Igbo, Yoruba, Hausa or another local language. Similar communication patterns can be found in hospitals, courtrooms, government services and other sectors.
According to Intron AI, these switching points have traditionally created challenges for speech recognition systems.
Sahara v2.5 is designed to address this problem by supporting code-switching across approximately 20 African languages.
The company says the model can follow conversations more naturally instead of treating language switches as errors or gaps in speech recognition.
The release covers languages used across more than 10 African countries, including Hausa, Swahili, Zulu and Luganda.
AI That Can Speak Back
Beyond understanding speech, Sahara v2.5 also introduces improved voice generation.
The technology can generate responses in African languages, including Igbo and Hausa, while also supporting mixed-language conversations.
This could open up new possibilities for businesses and institutions building voice-based services for African users.
A call centre, for example, could deploy an AI-powered voice assistant capable of communicating with customers in their preferred language rather than requiring the entire interaction to take place in English.
The technology could also have applications in government services, public information campaigns, healthcare, agriculture and media.
Intron AI CEO Tobi Olatunji said the company’s goal is to build technology that understands how people naturally communicate.
“If you give us voice, we want to give you text. If you give us text, we want to give you voice.”
Healthcare Among Early Use Cases
One of the early deployments of the technology is at Meridian Hospital in Enugu, where doctors and patients often communicate in Igbo.
The hospital had previously used Intron’s English-language technology to help doctors dictate medical notes after consultations.
With the expanded multilingual capabilities, the technology could now help capture and process conversations more directly, allowing medical professionals to review AI-generated notes instead of manually documenting every consultation from scratch.
Intron AI believes the technology could help improve efficiency in healthcare, particularly for medical professionals handling large numbers of patients.
The startup is also exploring applications across financial services, government, call centres, legal services and agriculture.
An API-First Strategy
Rather than building separate products for every industry, Intron AI is increasingly adopting an API-first approach.
This allows other companies to integrate the startup’s speech recognition and voice generation technology into their own platforms and services.
A financial technology company, for instance, could integrate Intron’s voice AI into its application without Intron having to develop an entirely separate banking product.
The company is also exploring opportunities in voice banking, where users could potentially interact with financial services using spoken commands instead of navigating apps and menus.
Intron’s direct sales efforts are focused on sectors including government and financial services, where the company sees strong commercial opportunities.
Building African Voice Technology
Developing effective voice AI for African languages requires significant amounts of language data and computing infrastructure.
Intron AI says it uses a combination of human contributors and proprietary synthetic data generation technology to train its models.
The company has also filed a US patent application related to its data-generation technology.
However, scaling the technology presents additional challenges, including the high cost of computing infrastructure and limited access to advanced graphics processing units required to run large AI systems.
Data privacy is another major consideration, particularly for sensitive sectors such as healthcare.
Intron says it is already working on deployments where AI models can operate offline, while cloud-based customers can also request zero data retention.
Making Digital Services More Accessible
The latest release reflects a growing push to build AI systems that adapt to African users rather than forcing users to adapt to technology designed primarily for other markets.
For millions of Africans, everyday communication is multilingual and often involves switching between English and indigenous languages.
By enabling AI to better understand those conversations, Intron AI is positioning its technology for a future in which voice could become an increasingly important interface for accessing digital services.
The company believes this could make technology more accessible to people who may be excluded from services designed primarily around smartphones, screens and text.
What’s Next?
Intron AI plans to continue expanding the capabilities and applications of its voice technology as it grows across sectors including healthcare, government and financial services.
The company expects to release additional performance benchmarks for Sahara v2.5 and is targeting profitability as it expands its commercial partnerships.