Sarvam AI

Text-to-speech service implementation using Sarvam AI’s TTS API

Overview

SarvamTTSService provides text-to-speech synthesis specialized for Indian languages and voices. The service offers extensive voice customization options including pitch, pace, and loudness control, with support for multiple Indian languages and preprocessing for mixed-language content. The bulbul:v3-beta model adds temperature control and 25 new speaker voices.

Sarvam TTS API Reference

Pipecat’s API methods for Sarvam AI TTS integration

Example Implementation

Complete example with Indian language support

Sarvam Documentation

Official Sarvam AI text-to-speech API documentation

Sarvam Console

Access Indian language voices and API keys

Installation

To use Sarvam AI services, no additional dependencies are required beyond the base installation:

pip install "pipecat-ai"

Prerequisites

Sarvam AI Account Setup

Before using Sarvam AI TTS services, you need:

1. Sarvam AI Account: Sign up at Sarvam AI Console
2. API Key: Generate an API key from your account dashboard
3. Language Selection: Choose from available Indian language voices

Required Environment Variables

• SARVAM_API_KEY: Your Sarvam AI API key for authentication

Configuration

Sarvam offers two service implementations: SarvamTTSService (WebSocket) for real-time streaming and SarvamHttpTTSService (HTTP) for simpler batch synthesis.

SarvamTTSService

api_key · str · required

Sarvam AI API subscription key.

model · str · default: "bulbul:v2"

TTS model to use. Options: bulbul:v2, bulbul:v3-beta, bulbul:v3.

voice_id · str · default: "None"

Speaker voice ID. If None, uses the model-appropriate default (anushka for v2, shubh for v3).

url · str · default: "wss://api.sarvam.ai/text-to-speech/ws"

WebSocket URL for the TTS backend.

aggregate_sentences · bool · default: "True"

Buffer text until sentence boundaries before sending.

sample_rate · int · default: "None"

Audio sample rate in Hz (8000, 16000, 22050, 24000). If None, uses model-specific default (22050 for v2, 24000 for v3).

params · InputParams · default: "None"

Runtime-configurable voice and generation settings. See InputParams (WebSocket) below.

SarvamHttpTTSService

api_key · str · required

Sarvam AI API subscription key.

aiohttp_session · aiohttp.ClientSession · required

An aiohttp session for HTTP requests.

model · str · default: "bulbul:v2"

TTS model to use. Options: bulbul:v2, bulbul:v3-beta, bulbul:v3.

voice_id · str · default: "None"

Speaker voice ID. If None, uses the model-appropriate default.

base_url · str · default: "https://api.sarvam.ai"

Sarvam AI API base URL.

sample_rate · int · default: "None"

Audio sample rate in Hz (8000, 16000, 22050, 24000). If None, uses model-specific default.

params · InputParams · default: "None"

Runtime-configurable voice and generation settings. See InputParams (HTTP) below.

InputParams (WebSocket)

Parameter	Type	Default	Description
language	Language	Language.EN	Target language for synthesis. Supports Bengali, English, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu.
pitch	float	0.0	Voice pitch adjustment (-0.75 to 0.75). Only for bulbul:v2.
pace	float	1.0	Speech pace multiplier. v2: 0.3-3.0, v3: 0.5-2.0.
loudness	float	1.0	Volume multiplier (0.3 to 3.0). Only for bulbul:v2.
enable_preprocessing	bool	False	Enable text preprocessing. Always enabled for v3 models.
min_buffer_size	int	50	Minimum characters to buffer before TTS processing.
max_chunk_length	int	150	Maximum characters processed in a single chunk.
output_audio_codec	str	"linear16"	Audio codec: linear16, mulaw, alaw, opus, flac, aac, wav, mp3.
output_audio_bitrate	str	"128k"	Audio bitrate: 32k, 64k, 96k, 128k, 192k.
temperature	float	0.6	Output randomness for v3 models (0.01-1.0). Ignored for v2.

InputParams (HTTP)

Parameter	Type	Default	Description
language	Language	Language.EN	Target language for synthesis.
pitch	float	0.0	Voice pitch adjustment (-0.75 to 0.75). Only for bulbul:v2.
pace	float	1.0	Speech pace multiplier. v2: 0.3-3.0, v3: 0.5-2.0.
loudness	float	1.0	Volume multiplier (0.3 to 3.0). Only for bulbul:v2.
enable_preprocessing	bool	False	Enable text preprocessing. Always enabled for v3 models.
temperature	float	0.6	Output randomness for v3 models (0.01-1.0). Ignored for v2.

Usage

Basic Setup (WebSocket)

from pipecat.services.sarvam import SarvamTTSService
from pipecat.transcriptions.language import Language

tts = SarvamTTSService(
    api_key=os.getenv("SARVAM_API_KEY"),
    voice_id="anushka",
    params=SarvamTTSService.InputParams(
        language=Language.HI,
    ),
)

With v3 Model and Temperature Control

from pipecat.services.sarvam import SarvamTTSService
from pipecat.transcriptions.language import Language

tts = SarvamTTSService(
    api_key=os.getenv("SARVAM_API_KEY"),
    voice_id="aditya",
    model="bulbul:v3-beta",
    params=SarvamTTSService.InputParams(
        language=Language.HI,
        pace=1.2,
        temperature=0.8,
    ),
)

HTTP Service

import aiohttp
from pipecat.services.sarvam import SarvamHttpTTSService
from pipecat.transcriptions.language import Language

async with aiohttp.ClientSession() as session:
    tts = SarvamHttpTTSService(
        api_key=os.getenv("SARVAM_API_KEY"),
        aiohttp_session=session,
        voice_id="anushka",
        params=SarvamHttpTTSService.InputParams(
            language=Language.HI,
            pitch=0.1,
            pace=1.2,
            loudness=1.5,
        ),
    )

Notes

• Model differences: bulbul:v2 supports pitch and loudness control; bulbul:v3-beta and bulbul:v3 add temperature control but do not support pitch or loudness. Setting unsupported parameters for a model will log a warning.
• Default speakers vary by model: v2 defaults to anushka; v3 models default to shubh.
• Default sample rates vary by model: v2 defaults to 22050 Hz; v3 models default to 24000 Hz.
• Indian language focus: Sarvam AI specializes in Indian languages, supporting Bengali, English (India), Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu.
• Pace ranges differ: bulbul:v2 supports pace from 0.3 to 3.0, while v3 models support 0.5 to 2.0. Values outside the range are clamped automatically.

Event Handlers

Sarvam WebSocket TTS supports the standard service connection events:

Event	Description
on_connected	Connected to Sarvam WebSocket
on_disconnected	Disconnected from Sarvam WebSocket
on_connection_error	WebSocket connection error occurred

@tts.event_handler("on_connected")
async def on_connected(service):
    print("Connected to Sarvam")