Gnani AI launches Prisma v3.0 speech-to-text for Indian calls
Gnani AI has rolled out Prisma v3.0, a speech-to-text model built for Indian phone calls and heavily code-mixed chats.
It supports nine Indian languages plus Indian English, offers four ways to transcribe, and lets users tweak five different settings.
In recent tests, it beat out ElevenLabs Scribe v2 with a lower word error rate, so your transcripts are more accurate than ever.
Gnani Auto-LID tops 98% on Vistaar
Gnani AI's speech models, which power 50 million voice interactions every day, also work with all the usual tech like REST APIs and WebSocket streaming.
On Vistaar audio clips longer than three seconds, automatic language identification (Auto-LID) named the spoken language correctly more than 98% of the time, a big deal for anyone who mixes languages while talking.
Ganesh Gopalan, Co-founder and CEO, Gnani AI says this makes it perfect for real-world conversations in India, while Bharath Shankar, Chief Product and Engineering Officer, Gnani AI points out that it keeps native scripts intact even when people switch between languages in one sentence.