Speech-to-speech over one socket
You bring the line.
We bring the voice.
Dvaarik runs the hard part of a phone agent — hearing, thinking, speaking, and knowing when to stop talking. Keep your own telephony, your own numbers, your own prompt. Stream the audio to us and it comes back speaking.
₹2/min · 11 Indian languages · no per-seat, no minimum
Everything between the ears
A voice agent is four hard problems stacked. You get all four behind one socket, tuned on real Indian phone calls.
- Hearing
- Speech recognition tuned for Indian accents and code-mixed speech, on 8 kHz phone audio.
- Thinking
- Your prompt, run on a fast model. Swap models later without touching your integration.
- Speaking
- Natural voices in eleven languages, streamed back sentence by sentence so replies start fast.
- Turn-taking
- Barge-in detection that survives background noise, silence handling, and graceful hangups.
From nothing to a live call
- 01
Create the call
POST your prompt, language and voice. You get back a WebSocket URL, good for five minutes.
- 02
Bridge your audio
Point your telephony provider's media stream at that URL, or pipe audio from anywhere else.
- 03
Play what comes back
Agent audio arrives as PCM16 frames. Transcripts and turn events arrive as JSON alongside.
- 04
Pay for the minutes
The call is metered against your prepaid balance when it ends. Nothing else is billed.
Create a call
# your server, once per phone call
curl -X POST https://api.developers.dvaarik.com/v1/calls \
-H "X-Api-Key: dvk_live_..." \
-d '{"prompt":"You are Riya at Sunrise Dental...","language":"en-IN"}'
# → { ws_url: "wss://..." } then stream PCM16 both ways
Your carrier. Your numbers. Your rules.
We do not sell you phone numbers and we never touch your carrier relationship. If your provider can hand you call audio over a WebSocket — Twilio, Plivo, Exotel and most others can — you are one bridge away from a working agent. A SIP trunk plus a media gateway works the same way.
That keeps DLT registration, number allocation and TRAI compliance where they already sit: with you.
Priced by the minute
Prepaid balance, drawn down as calls complete. No subscription, no seats, no charge for the minutes you do not use.
Standard
₹2/min
Indic-tuned recognition and voices. The right default for Indian phone traffic.
Premium
₹3.50/min
HD neural voices with fuller prosody, for brands that live on the phone.
Rounded up to the whole minute · calls capped at 10 minutes · concurrency rises with usage
Build the agent. Skip the stack.
Sign up, put ₹500 in the wallet, and place a call in the next ten minutes.