Gnani.ai Launches Speech Model Supporting 12 Indian Languages

Gnani.ai has launched Prisma v2.5, a speech recognition model trained on 14 million hours of proprietary Indic speech data across 12 Indian languages. The model is designed to handle common Indian speech conditions including regional accents, dialect variations, background noise, compressed telephony audio, and code-switching between English and Indian languages.

According to the company, Prisma v2.5 outperformed global and Indian speech recognition models in independent benchmarks, ranking first in eight out of nine Indian languages tested. The model targets enterprise customers across contact centers, banking, financial services, insurance, healthcare, retail, and government services.

CEO Ganesh Gopalan emphasized that Indian speech environments require specialized training data reflecting real-world conversations with accents, noise, and code-switching. The launch follows Gnani.ai's $10 million funding round from Aavishkaar Capital earlier this year and positions the company in competition with Sarvam AI, ElevenLabs, and OpenAI in the Indic AI space.

The model will be available through APIs, enabling businesses to improve transcription quality in multilingual customer interactions at scale.