CallMissed logoCallMissed
Back to all models
Speech to Textindian-languagesfree-tier
Sarvam AI — saaras:v4 logo

Saaras V4

by Sarvam AI · Released 2026

Sarvam AI's speech-to-text model with five output modes — transcribe, translate, verbatim, transliterate, and code-mix — across 24 languages. One model call returns the transcript in whichever written form your pipeline needs. Available on the CallMissed free tier.

Sarvam AI — saaras:v4 logo
Speech to Text

Saaras V4

Powered by Sarvam AI · Proprietary ASR model

Context Window

N/A

Parameters

Undisclosed

Max Output

N/A

Category

Speech to Text

Overview

Saaras V4 is Sarvam AI's speech-to-text model, covering 24 languages. Its defining feature is output mode: rather than returning a single fixed transcript shape, it can return the same audio in five different written forms.

Transcribe returns the speech as spoken, in its native script. Translate returns an English rendering of the speech. Verbatim keeps disfluencies, repetitions, and filler words intact — useful for compliance recordings, legal transcripts, and quality audits where what was actually said matters more than what was meant. Transliterate writes the speech in Roman script rather than the native one, which keeps Indic-language transcripts readable in systems that do not render Devanagari or Tamil script. Code-mix preserves mixed-language speech as it was spoken, so Hinglish and similar code-switched utterances survive the transcription step instead of being forced into one language.

That mode switch removes a whole post-processing layer: instead of transcribing and then running a separate translation or transliteration pass, you request the form you need directly. Saaras V4 is priced at $0.53 per audio hour and is included in the CallMissed free tier.

Pricing

MetricPrice
Price /hour₹53.0000

1 credit = ₹1 = $0.01 USD. Prices shown from provider; CallMissed passes through with ~35% markup.

Key Highlights

  • Five output modes — transcribe, translate, verbatim, transliterate, code-mix
  • 24 languages
  • Code-mix mode preserves Hinglish and other code-switched speech
  • Free on CallMissed — available on the free tier

Benchmarks

BenchmarkScore
Languages24
Output modes5
Price$0.53

Technical Details

  • Five output modes: transcribe, translate, verbatim, transliterate, code-mix
  • Transcribe — speech as spoken, in the native script
  • Translate — English rendering of the source speech
  • Verbatim — keeps disfluencies and fillers for compliance and audit use
  • Transliterate — native-language speech written in Roman script
  • Code-mix — preserves mixed-language speech as spoken
  • 24 supported languages
  • Free tier eligible on CallMissed

Strengths

  • Five output modes from one model call — no separate translation or transliteration pass
  • Verbatim mode preserves exactly what was said for compliance work
  • Code-mix mode handles Hinglish and similar code-switched speech
  • 24-language coverage with free-tier access

Limitations

  • Focused on Indian languages — not a general-purpose global STT
  • Accuracy may vary across less common supported languages
  • Output mode must be chosen per request — one call returns one form

Use Cases

Call center transcriptionVoice agent backendsCompliance and audit transcriptsMultilingual and code-mixed dictation

API Example

curl https://api.callmissed.com/v1/audio/transcriptions \
  -H "Authorization: Bearer cm_YOUR_KEY" \
  -F file=@audio.wav \
  -F model=saaras:v4 \
  -F language=hi

Endpoint: POST /v1/audio/transcriptions · Model ID: saaras:v4

Try Saaras V4 now

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

Start free Book a DemoRead docs