Intron Launches Sahara v2, Advancing African Speech Recognition
The platform draws on an extensive dataset of 40,000 voices, 14 million audio samples, and over 50,000 hours of recorded speech, representing more than 500 accents and 24 African languages, including Zulu, Swahili, Hausa, African French, Pidgin, Yoruba, Ga, Kinyarwanda, Xhosa, and Shona.
After months of development, Intron has launched Sahara v2, a next-generation speech and voice intelligence platform built specifically for African languages, accents, and real-world use cases. The platform draws on an extensive dataset of 40,000 voices, 14 million audio samples, and over 50,000 hours of recorded speech, representing more than 500 accents and 24 African languages, including Zulu, Swahili, Hausa, African French, Pidgin, Yoruba, Ga, Kinyarwanda, Xhosa, and Shona.
Sahara v2 delivers state-of-the-art speech recognition, consistently outperforming global models such as Gemini-3 and Meta’s OmniASR across healthcare, finance, legal, and call center benchmarks. Its accuracy and robustness make it suitable for complex real-world scenarios, including code-switching, background noise, numeric-heavy speech, and culturally specific names and terms.
The platform introduces new capabilities for African users, including the world’s first bilingual Swahili–English ASR for rapid code-switching, Hausa text-to-speech and voice bots, voice commands for apps and websites, voice autofill for long online forms, and offline deployment, enabling high-performance AI even in low-connectivity environments.
With paying customers across health, finance, and legal sectors, Sahara v2 is already transforming how organizations engage with African users. The launch highlights the continent’s linguistic diversity, providing voice AI that truly understands African speech while opening doors for new applications and digital experiences.
