Skip to content
CallMissed
Book a demo

Explore CallMissed

Back to all models
Speech to Textindian-languages

Gnani Prisma v3

by Gnani · Released 2026

Gnani.ai's newest India-first telephony speech-to-text model. English plus 9 Indian languages, automatic language detection, mixed-language speech and keyterm boosting, for real-time voice agents and batch transcription.

Speech to Text

Gnani Prisma v3

Powered by Gnani · Gnani Prisma v3 telephony speech recognition

Context Window

N/A

Parameters

N/A

Max Output

N/A

Category

Speech to Text

Overview

Gnani Prisma v3 is Gnani.ai's newest India-first speech-to-text model, built for Indian telephony. It covers 10 languages: English and nine Indian languages (Hindi, Bengali, Tamil, Telugu, Kannada, Malayalam, Marathi, Gujarati and Punjabi), selected with locale codes such as hi-IN or en-IN.

Set the language to indic-auto and the model detects the language itself, and it handles speakers who mix languages in the same call. Keyterm boosting lets you bias recognition toward names and domain jargon. It runs in real time for voice agents and in batch for recorded audio.

On the CallMissed platform a voice agent set to multilingual follows the caller: Prisma v3 reports the language it heard, so the voice replies in the same language. At $0.2813 per audio hour it is priced for production call workloads. It replaces Gnani Prisma v2.5, whose ID keeps working and is answered by this model at the same price.

Pricing

MetricPrice
Price /hour₹27.0000

1 credit = ₹1 ≈ US$0.0104 (US$1 = ₹96). Transparent per-call pricing — you pay only for what you call, with no per-seat charge or minimums.

Key Highlights

  • English + 9 Indian languages
  • Automatic language detection with indic-auto
  • Mixed-language speech and keyterm boosting
  • Real-time and batch transcription

Benchmarks

BenchmarkScore
Languages10
Language Detectionindic-auto
Audio ProfileTelephony
DeploymentRealtime + Batch

Technical Details

  • Supports hi-IN, bn-IN, ta-IN, te-IN, kn-IN, ml-IN, mr-IN, gu-IN, pa-IN and en-IN
  • language accepts one code, a comma-separated list of up to 4 codes, or indic-auto; a base code such as hi is read as hi-IN
  • Mixed-language speech and keyterm boosting for names and jargon
  • Real-time for voice agents and batch for recorded audio
  • File transcription is limited to 30 seconds of audio per request
  • Processed in India, available in India-hosted voice agents

Strengths

  • Built for Indian telephony audio
  • Automatic language detection across 10 languages
  • Handles mixed-language speech
  • Keyterm boosting for names and domain terms

Limitations

  • Focused on English and Indian languages — not a general-purpose multilingual STT
  • File transcription accepts at most 30 seconds of audio per request
  • Paid plans only

Use Cases

Contact center transcriptionVoice agent backendsMultilingual callers that switch languageNames and jargon with keyterm boosting

API Example

curl https://api.callmissed.com/v1/audio/transcriptions \
  -H "Authorization: Bearer cm_YOUR_KEY" \
  -F [email protected] \
  -F model=gnani-prisma-v3.0 \
  -F language=indic-auto

Endpoint: POST /v1/audio/transcriptions · Model ID: gnani-prisma-v3.0

Try Gnani Prisma v3 now

Get 1000 free API credits on signup. No credit card required.

Start free Book a DemoRead docs