Bengaluru-based Gnani AI has unveiled its latest speech-to-text model, Prisma v2.5, which is already making waves in the healthcare sector. This model is not just another tech launch; it is a critical tool designed to bridge the communication gap in a country where linguistic diversity and ambient noise are the norm. With a remarkable performance that ranks first in 8 out of 9 Indian languages, Prisma v2.5 is trained on 14 million hours of proprietary Indic speech, making it uniquely suited for real-world applications.
The implications for healthcare are profound. Ananth Nagaraj, Co-founder and CTO of Gnani AI, highlighted the stakes involved: a single transcription error in financial discussions can misrepresent amounts by millions. In healthcare, where accurate communication can be a matter of life and death, the stakes are even higher. The model's ability to accurately transcribe short utterances and domain-specific vocabulary means that critical patient information can be captured without distortion, significantly reducing risks associated with miscommunication.
Moreover, Prisma v2.5 addresses a significant pain point for Indian enterprises. Traditional speech models often falter when faced with the realities of Indian accents and telephony audio quality. As Ganesh Gopalan, CEO of Gnani AI, pointed out, this model is the first to align its training data with how Indians actually speak, making it a game-changer for sectors like BFSI and healthcare.



