Welcome to Vakyam AI

Vakyam AI is a developer platform for natural, human-sounding speech in Indian languages. Our text-to-speech model, Raaga 1, turns text in English, Hindi, Tamil, Telugu, Kannada, Marathi, Gujarati, and Bengali into lifelike audio through a simple HTTP and WebSocket API.

These docs will take you from your first API call to streaming real-time audio in production.

New here? Start with these

Quickstart

Make your first request and save an audio file in a few minutes.

Authentication

Create an API key and learn how to send it.

Voices & languages

Browse the voices and languages Raaga 1 supports.

API Reference

Full endpoint, parameter, and schema details.

Ways to generate speech

Pick the approach that fits your product's latency and integration needs:

Full audio

One request, one complete audio file returned as base64 JSON.

HTTP streaming

Stream audio bytes as they are generated over a single HTTP response.

Realtime WebSocket

Sentence-by-sentence synthesis with binary audio chunks over a socket.

Prefer an SDK?

Skip the raw HTTP calls with an official client:

Python SDK

pip install vakyamai

JavaScript / TypeScript SDK

pnpm add @vakyam-ai/tts

Building a voice agent?

Drop Vakyam TTS into a realtime voice-agent framework:

LiveKit Agents

Use VakyamTTS as the voice in a LiveKit Agents pipeline.

Pipecat

Add VakyamTTSService from pipecat-vakyam to a pipeline.

Good to know

Eight Indian languages — en-IN, hi-IN, ta-IN, te-IN, kn-IN, mr-IN, gu-IN, bn-IN.
Multiple voices per language — the same person can speak across languages.
Flexible output — mp3, wav, raw pcm, or mulaw, at sample rates from 8 kHz to 48 kHz.
Fair credit counting — monthly character allotments by plan, plus optional extra credits at ₹0.75 per 1,000 characters, counted by Unicode character (aksara), not by byte.
Plan-based limits — Free, Developer, and Growth tiers set your monthly characters, requests per minute, and concurrency.

Ask AI about this page