Indian voices for your voice agent.

Text to speech in 12 Indian languages, spoken sentence by sentence as your LLM writes. One model, one price: ₹0.30 per 1,000 characters.

₹1,000 of free credit, no card needed.

Hear it: one sample in each language

Bunty · Hindi
नमस्ते! आपका ऑर्डर आज शाम तक डिलीवर हो जाएगा, और डिलीवरी पार्टनर पहुँचने से पहले आपको कॉल करेगा। अगर आप उस वक़्त घर पर नहीं हैं, तो बस बता दीजिए, हम डिलीवरी कल के लिए रीशेड्यूल कर देंगे।
Languages
12
Voices
14
Audio
24 kHz
Per 1,000 characters
₹0.30

What it is

A voice for agents that speak India’s languages.

Vaakini is a text-to-speech API with one model, vaakini-tts-1. Send text, get audio: 24 kHz PCM, MP3 or WAV.

Twelve languages, one model
Assamese, Bengali, English, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil and Telugu. Name the language, or leave it out and the model picks it from the script.
Spoken as the text arrives
Stream your LLM’s reply as it is written. Vaakini splits it into sentences and speaks each as soon as it is complete, so the first audio follows the reply’s first sentence.
Made for agents
A LiveKit plugin and a Pipecat service that cancel a reply when your caller interrupts, and a Python SDK, an HTTP API and a WebSocket for everything else.
Numbers read as words
Amounts, dates, times, phone numbers, URLs and emails are spoken as words in the request’s language, unless you turn that off.

Voices

14 voices, each made for its language.

Pass a voice’s name as voice. Each is made from a recording in its own language, and can speak the other eleven too.

  • BuntyHindi

    Young male voice, friendly, casual

  • MonikaHindi

    Young female voice, warm, helpful

  • UmanandaAssamese

    Adult male voice, steady, friendly

  • RimjhimBengali

    Young female voice, clear, expressive

  • MuskanGujarati

    Adult female voice, warm, sweet

  • SrivatsaKannada

    Adult male voice, casual, natural

  • MahendranMalayalam

    Young male voice, relatable, natural

  • SunainaMalayalam

    Adult female voice, warm, clear

  • AmoghMarathi

    Young male voice, casual, engaging

  • KalyaniMarathi

    Young female voice, warm, gentle

  • SwastikOdia

    Adult male voice, casual, warm

  • RaqibPunjabi

    Young male voice, warm, conversational

  • VennilaTamil

    Young female voice, warm, easygoing

  • BhaavanaTelugu

    Young female voice, natural, warm

Or your own voice, from a 10-second recording.

Clone a voice you hold the rights to. Only your org can speak it, at the same price as every other voice. Custom voices are switched on per org: ask us, and we enable yours.

Write to hello@vaakini.com
  1. 1

    Record

    3 to 10 seconds of the speaker, with its transcript.

  2. 2

    Create

    On the dashboard’s Voices page, or with one API call. Ready at once.

  3. 3

    Speak

    Its id as voice, in any of the 12 languages.

Quickstart

Drop it into your agent.

Install the LiveKit plugin, the Pipecat service or the Python SDK, set VAAKINI_API_KEY, and name a voice.

  • HTTP API

    POST /v1/audio/speech returns the audio as it is spoken: MP3, PCM or WAV.

  • WebSocket

    /v1/stream takes text as it is written and sends each sentence’s audio back, from any language.

# pip install livekit-plugins-vaakinifrom livekit.agents import AgentSessionfrom livekit.plugins import vaakini# The API key comes from VAAKINI_API_KEY.tts = vaakini.TTS(    voice="Monika",    language="hi",)session = AgentSession(    tts=tts,    # ... stt, llm, vad)

Pricing

One price, in rupees.

Every new org gets ₹1,000 of credit, about 3.3 million characters, with no card. After that, buy prepaid credit from ₹1,000 plus 18% GST. It never expires.

Pricing in full
Speech, in every voice and language
₹0.30/ 1,000 characters
Creating and keeping custom voices
₹0

Give your agent
a voice in its language.

Sign up, copy your API key, and hear every voice in the playground. Your first ₹1,000 of speech is free, with no card.