Hindi and Indian-language voice AI API for calls
For businesses whose callers speak Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali or another Indian language: the sarvam engine runs Sarvam's Saaras speech recognition, the Sarvam-105B model and the Bulbul voice as one pipeline, priced from the same book as every other engine and billed per second from one prepaid balance.
Last updated
How do you build a Hindi or Indian-language phone agent with Wixzel Voice?
- 01POST /v1/agents
Create the agent with the sarvam engine and the language set on the stt stage.
- 02POST /v1/sip-trunks
Connect an Indian carrier trunk or any trunk that terminates Indian numbers.
- 03POST /v1/calls
Place a call, or route inbound calls from the trunk to the agent.
- 04Compose
Mix within the family: sarvam models can be named stage by stage in the agent's voice object.
- 05GET /v1/usage/events
Every second, token and character itemised, in micro-dollars.
Which voice engine fits an Indian-language phone agent?
Use the sarvam engine on Wixzel Voice, at $0.0623 per connected minute. The stt provider selects the pipeline, and Sarvam's models are trained on Indian languages and accents rather than adapted from English. It is the engine to reach for whenever the caller's first language is not English.
What does an Indian-language phone agent cost on Wixzel Voice?
Wixzel Voice engine time for 3-minute calls on the sarvam engine, at typical speech rates, billed per second of actual usage. Telephony is your own trunk at your carrier’s rate and is not included, and shorter calls cost proportionally less.
| Calls per month | Connected minutes | Engine cost per month |
|---|---|---|
| 500 | 1,500 | $93.45 |
| 2,500 | 7,500 | $467 |
| 10,000 | 30,000 | $1,869 |
Prepaid: add credit from $5 and the calls spend it. Estimate your own mix.
Which guides show how to build a Hindi or Indian-language phone agent on Wixzel Voice?
Build phone agents that speak Hindi and Malayalam on Wixzel Voice's sarvam engine: language codes, Bulbul voices, code-mixed speech and the price per minute.
Four requests to a ringing phone, what a call record returns, and what a call costs.
Twilio, Telnyx, Plivo, Vonage, Bandwidth, Exotel, Vobiz or any SIP provider: what to set at the carrier.
Frequently asked questions
- Does Wixzel Voice support Hindi voice agents?
- Yes. The sarvam engine runs Sarvam's Saaras v3 speech-to-text, the Sarvam-105B language model and the Bulbul v3 voice, all built for Indian languages including Hindi. Set the language on the stt stage of the agent.
- Which Indian languages does the sarvam engine cover?
- The languages supported by Sarvam's Saaras, Sarvam-105B and Bulbul models, which include Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali, Gujarati, Punjabi and Odia. Check Sarvam's current model documentation for the full and up-to-date list; Wixzel Voice does not restrict it further.
- What does a minute in Hindi cost?
- $0.0623 per connected minute at typical speech rates on the sarvam engine, including the orchestration fee, plus your carrier's rate for the trunk.
- Can the agent switch between English and an Indian language mid-call?
- Within what the models handle, yes: Sarvam's models cope with code-mixed speech such as Hinglish. A hard switch to the classic engine mid-call is not possible; the engine is fixed when the call is admitted.
- Is Indian phone data kept in India?
- Sarvam processes speech and text in India during the call. The call log, transcript and any recording are stored on the platform's servers in Finland. See the Privacy Policy and DPA for the full list of processors and locations.
What else do developers build on Wixzel Voice?
Build an agent that answers the phone, checks availability and books an appointment, over your own SIP trunk, with the booking written back through the API.
Upload leads, start a campaign, and have an agent call each one over your own SIP trunk and ask your qualifying questions. Every call is a record with its status and transcript, linked to the lead.
Answer a support number with an agent that reads from a knowledge base you upload, handles the common questions, and leaves a transcript of every call.
Put an agent in a browser page, a Flutter app or a kiosk with no SIP trunk and no phone number: your server mints a one-minute session, the app streams audio over a WebSocket, and the session is billed like a call.
Connect Claude Code, claude.ai, Claude Desktop, Cursor or any MCP client to your account, and let an AI agent connect a carrier, build a voice agent, place calls, run campaigns and read transcripts, with every tool that spends money asking first.
Operator, prices, limits, refunds and data location, on one page.