Sarvam AI

Sarvam AI is a text-to-speech plugin that synthesizes the LLM’s reply into the agent’s voice across 11 Indic languages using the bulbul model family. It occupies the pipeline’s tts slot.

Setup

Set your Sarvam AI API key in the worker environment. Generate a key from the Sarvam AI dashboard:

export SARVAM_API_KEY=<key>

Usage

Import the plugin and pass it to the pipeline’s tts slot.

Python

from zrt.plugins import SarvamAITTS

tts = SarvamAITTS(
    model="bulbul:v3",
    speaker="shubh",
    language="en-IN",
)

# Pipeline(tts=tts, ...)

Parameters

Constructor parameters for the Python SDK (SarvamAITTS).

Parameter	Type	Default	Description
`api_key`	`str \| None`	`None`	Sarvam AI API key. Falls back to the `SARVAM_API_KEY` environment variable when unset.
`model`	`SarvamAITTSModel \| str`	`"bulbul:v3"`	TTS model: `"bulbul:v3"` (latest, 43+ voices, temperature control) or `"bulbul:v2"` (legacy, pitch/loudness controls).
`language`	`str`	`"en-IN"`	BCP-47 language code. Supported: `"bn-IN"`, `"en-IN"`, `"gu-IN"`, `"hi-IN"`, `"kn-IN"`, `"ml-IN"`, `"mr-IN"`, `"od-IN"`, `"pa-IN"`, `"ta-IN"`, `"te-IN"`.
`speaker`	`str`	`"shubh"`	Speaker/voice name, e.g. `"aditya"`, `"ritu"`, `"priya"`, `"neha"`, `"rahul"`.
`streaming`	`bool`	`True`	Stream audio back in chunks. Set to `False` for a single-response download.
`sample_rate`	`int`	`24000`	Output sample rate in Hz. Supported: `8000`, `16000`, `22050`, `24000`; `bulbul:v3` also supports `32000`, `44100`, `48000` via the REST API.
`output_audio_codec`	`str`	`"linear16"`	Audio encoding: `"linear16"`, `"mp3"`, `"mulaw"`, `"alaw"`, `"opus"`, `"flac"`, `"aac"`, `"wav"`.
`pitch`	`float \| None`	`0.0`	Pitch adjustment (`bulbul:v2` only). Range `-0.75` to `0.75`. No effect on `bulbul:v3`.
`pace`	`float \| None`	`1.0`	Speaking rate multiplier. `bulbul:v3` range: `0.5`–`2.0`; `bulbul:v2` range: `0.3`–`3.0`.
`loudness`	`float \| None`	`1.0`	Volume multiplier (`bulbul:v2` only). Range `0.3`–`3.0`. No effect on `bulbul:v3`.
`temperature`	`float \| None`	`0.6`	Expressiveness/variability control (`bulbul:v3` only). Range `0.01`–`2.0`.
`bitrate`	`str`	`"128k"`	MP3 bitrate when `output_audio_codec="mp3"`.
`min_buffer_size`	`int`	`50`	Minimum characters to accumulate before sending a synthesis request in streaming mode.
`max_chunk_length`	`int`	`150`	Maximum characters per synthesis chunk.
`preprocessing`	`bool`	`False`	Enable server-side text pre-processing (number normalization, acronym expansion) before synthesis.

The synthesized audio is streamed back to the caller as the final stage of the pipeline.

Plugins

Inference Gateway

Speech-to-text

LLM

Text-to-speech

Realtime models

Avatars

Turn detection

Voice activity detection

Noise cancellation

Setup

Usage

Parameters

​Setup

​Usage

​Parameters

Setup

Usage

Parameters