Sarvam AI Text-to-Speech

Overview

SarvamTTSService provides text-to-speech synthesis specialized for Indian languages and voices. The service uses the bulbul:v3 model by default, which offers temperature control for output randomness and pace adjustment. The previous bulbul:v2 model is deprecated as Sarvam’s API no longer serves it.

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:

uv add "pipecat-ai"

Prerequisites

Sarvam AI Account Setup

Before using Sarvam AI TTS services, you need:

Sarvam AI Account: Sign up at Sarvam AI Console
API Key: Generate an API key from your account dashboard
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

str

required

Sarvam AI API subscription key.

str

default:"None"

deprecated

Deprecated since 0.0.105, removed in 2.0.0. Use settings=SarvamTTSService.Settings(model=...) instead.

str

default:"None"

deprecated

Speaker voice ID. If None, uses shubh (the default for v3). Deprecated in v0.0.105. Use settings=SarvamTTSService.Settings(voice=...) instead.

str

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

WebSocket URL for the TTS backend.

TextAggregationMode

default:"None"

Controls how incoming text is aggregated before synthesis. SENTENCE (used when None) buffers text until sentence boundaries, producing more natural speech. TOKEN streams tokens directly for lower latency. Import from pipecat.services.tts_service.

bool

default:"None"

deprecated

Deprecated since 0.0.104, removed in 2.0.0. Use text_aggregation_mode instead.

int

default:"None"

Audio sample rate in Hz (8000, 16000, 22050, 24000, and for v3: 32000, 44100, 48000). If None, uses model-specific default (24000 for v3 and v4-flash, 22050 for deprecated v2).

InputParams

default:"None"

deprecated

Deprecated since 0.0.105, removed in 2.0.0. Use settings=SarvamTTSService.Settings(...) instead.

SarvamTTSService.Settings

default:"None"

Runtime-configurable settings. See SarvamTTSService Settings below.

SarvamHttpTTSService

str

required

Sarvam AI API subscription key.

aiohttp.ClientSession

required

An aiohttp session for HTTP requests.

str

default:"None"

deprecated

Deprecated since 0.0.105, removed in 2.0.0. Use settings=SarvamHttpTTSService.Settings(model=...) instead.

str

default:"None"

deprecated

Speaker voice ID. If None, uses shubh (the default for v3). Deprecated in v0.0.105. Use settings=SarvamHttpTTSService.Settings(voice=...) instead.

str

default:"https://api.sarvam.ai"

Sarvam AI API base URL.

int

default:"None"

Audio sample rate in Hz (8000, 16000, 22050, 24000, and for v3: 32000, 44100, 48000). If None, uses model-specific default (24000 for v3 and v4-flash, 22050 for deprecated v2).

InputParams

default:"None"

deprecated

Deprecated since 0.0.105, removed in 2.0.0. Use settings=SarvamHttpTTSService.Settings(...) instead.

SarvamHttpTTSService.Settings

default:"None"

Runtime-configurable settings. See SarvamHttpTTSService Settings below.

SarvamTTSService Settings

Runtime-configurable settings passed via the settings constructor argument using SarvamTTSService.Settings(...). These can be updated mid-conversation with TTSUpdateSettingsFrame. See Service Settings for details.

Parameter	Type	Default	Description
`model`	`str`	`"bulbul:v3"`	Model: `bulbul:v3`, `bulbul:v4-flash`, `bulbul:v3-beta`, or the deprecated `bulbul:v2`. Other values raise `ValueError` at construction and are ignored with a warning in a runtime update. (Inherited.)
`voice`	`str`	`"shubh"`	Voice identifier. (Inherited.)
`language`	`Language \| str`	`"en-IN"`	Language for synthesis. (Inherited.)
`enable_preprocessing`	`bool`	`False`	Enable text preprocessing. Set to `True` at construction for every model except `bulbul:v2`.
`pace`	`float`	`1.0`	Pace of speech.
`pitch`	`float`	`None`	Pitch of speech.
`loudness`	`float`	`None`	Loudness of speech.
`temperature`	`float`	`None`	Temperature for speech synthesis.
`min_buffer_size`	`int`	`50`	Minimum buffer size for WebSocket.
`max_chunk_length`	`int`	`150`	Maximum chunk length for WebSocket.

SarvamHttpTTSService Settings

Runtime-configurable settings passed via the settings constructor argument using SarvamHttpTTSService.Settings(...). These can be updated mid-conversation with TTSUpdateSettingsFrame. See Service Settings for details.

Parameter	Type	Default	Description
`model`	`str`	`"bulbul:v3"`	Model: `bulbul:v3`, `bulbul:v4-flash`, `bulbul:v3-beta`, or the deprecated `bulbul:v2`. Other values raise `ValueError` at construction and are ignored with a warning in a runtime update. (Inherited.)
`voice`	`str`	`"shubh"`	Voice identifier. (Inherited.)
`language`	`Language \| str`	`"en-IN"`	Language for synthesis. (Inherited.)
`enable_preprocessing`	`bool`	`False`	Enable text preprocessing. Set to `True` at construction for every model except `bulbul:v2`.
`pace`	`float`	`1.0`	Pace of speech.
`pitch`	`float`	`None`	Pitch of speech.
`loudness`	`float`	`None`	Loudness of speech.
`temperature`	`float`	`None`	Temperature for speech synthesis.

Usage

Basic Setup (WebSocket)

import os

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

tts = SarvamTTSService(
    api_key=os.getenv("SARVAM_API_KEY"),
    settings=SarvamTTSService.Settings(
        voice="shubh",
        model="bulbul:v3",
        language=Language.HI,
    ),
)

With Temperature Control

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

tts = SarvamTTSService(
    api_key=os.getenv("SARVAM_API_KEY"),
    settings=SarvamTTSService.Settings(
        voice="shubh",
        model="bulbul:v3",
        language=Language.HI,
        pace=1.2,
        temperature=0.8,
    ),
)

HTTP Service

import aiohttp
from pipecat.services.sarvam.tts 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,
        settings=SarvamHttpTTSService.Settings(
            voice="shubh",
            model="bulbul:v3",
            language=Language.HI,
            pace=1.2,
            temperature=0.8,
        ),
    )

Notes

Model default: bulbul:v3 is the default model. The previous bulbul:v2 is deprecated since 1.9.0: Sarvam’s API rejects it.
Model differences: bulbul:v2 supported pitch and loudness control; bulbul:v3 supports temperature control but does not support pitch or loudness. bulbul:v4-flash is a low-latency model that supports pitch, loudness, and pace (0.5 to 2.0) but ignores temperature. Unsupported parameters are not sent; setting one at construction logs a warning.
bulbul:v4-flash speakers: bulbul:v4-flash has its own speaker catalogue, listed in Sarvam’s documentation; the bulbul:v3 speaker names are not accepted. The default is shubh_en_narration_gentle.
Changing models at runtime: Updating model through TTSUpdateSettingsFrame switches to the new model’s capabilities and resets the voice to the new model’s default unless you also set voice. SarvamTTSService reconnects its WebSocket, since the model is part of the connection URL.
Default voice: v3 defaults to shubh; deprecated v2 defaulted to anushka.
Default sample rate: v3 and v4-flash default to 24000 Hz; deprecated v2 defaulted to 22050 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:v3 and bulbul:v4-flash support pace from 0.5 to 2.0; deprecated v2 supported 0.3 to 3.0. Values outside the range are clamped at construction; runtime updates are sent as given.

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")

Get Started

Pipecat Server

Client SDKs

Pipecat Flows

Pipecat Cloud

CLI

Pipecat Context Hub

Sarvam AI Text-to-Speech

Overview

Sarvam TTS API Reference

Example Implementation

Sarvam Documentation

Sarvam Console

Installation

Prerequisites

Sarvam AI Account Setup

Required Environment Variables

Configuration

SarvamTTSService

SarvamHttpTTSService

SarvamTTSService Settings

SarvamHttpTTSService Settings

Usage

Basic Setup (WebSocket)

With Temperature Control

HTTP Service

Notes

Event Handlers

​Overview

Sarvam TTS API Reference

Example Implementation

Sarvam Documentation

Sarvam Console

​Installation

​Prerequisites

​Sarvam AI Account Setup

​Required Environment Variables

​Configuration

​SarvamTTSService

​SarvamHttpTTSService

​SarvamTTSService Settings

​SarvamHttpTTSService Settings

​Usage

​Basic Setup (WebSocket)

​With Temperature Control

​HTTP Service

​Notes

​Event Handlers

Overview

Installation

Prerequisites

Sarvam AI Account Setup

Required Environment Variables

Configuration

SarvamTTSService

SarvamHttpTTSService

SarvamTTSService Settings

SarvamHttpTTSService Settings

Usage

Basic Setup (WebSocket)

With Temperature Control

HTTP Service

Notes

Event Handlers