Skip to main content

Simplismart TTS plugin guide

How to use the Simplismart TTS plugin for LiveKit Agents.

On this page

Available inPython

Overview

This plugin allows you to use Simplismart  as a TTS provider for your voice agents. Simplismart hosts multiple TTS model families: Orpheus (the default) and Qwen 3. Orpheus models run on Simplismart's shared inference endpoint, while Qwen 3 TTS runs on a dedicated deployment you create.

Installation

Install the plugin from PyPI:

uv add "livekit-agents[simplismart]~=1.8"

Authentication

The Simplismart plugin requires a Simplismart API key .

Set SIMPLISMART_API_KEY in your .env file.

Orpheus models

Orpheus models run on Simplismart's shared inference endpoint.

Usage

Use Simplismart TTS within an AgentSession or as a standalone speech generator. For example, you can use this TTS in the Voice AI quickstart.

from livekit.plugins import simplismart
session = AgentSession(
tts=simplismart.TTS(
model="canopylabs/orpheus-3b-0.1-ft",
voice="tara",
)
# ... llm, stt, etc.
)

Parameters

This section describes some of the parameters for Orpheus models. For a complete list, see the plugin reference.

modelstr | TTSModelsDefault: canopylabs/orpheus-3b-0.1-ft

Model identifier for the backend TTS model. Examples include canopylabs/orpheus-3b-0.1-ft and maya-research/Veena.

voicestrDefault: tara

Voice identifier to use for generation.

temperaturefloatDefault: 0.7

Controls the randomness in the model output. Lower values make output more deterministic.

top_pfloatDefault: 0.9

Nucleus sampling probability threshold. Limits the sampling pool of predicted tokens.

repetition_penaltyfloatDefault: 1.5

Penalty applied to repeated text to reduce repetition.

max_tokensintDefault: 1000

Maximum number of output tokens allowed in the synthesized speech.

Qwen 3 TTS

Qwen 3 TTS runs on a dedicated Simplismart deployment.

Model deployment

Deploy the qwen-tts model from your Simplismart dashboard before using it with LiveKit Agents. Your deployment includes a model endpoint URL. Pass this URL as base_url, including the /v1/audio/speech path. Authenticate with the same SIMPLISMART_API_KEY.

Usage

Set model="qwen-tts" to use the Qwen request format and the default Chelsie voice. Pass language and leading_silence to configure Qwen-specific behavior.

from livekit.plugins import simplismart
session = AgentSession(
tts=simplismart.TTS(
model="qwen-tts",
voice="Chelsie",
base_url="https://<your-deployment-endpoint>/v1/audio/speech",
language="English",
leading_silence=True,
)
# ... llm, stt, etc.
)

Parameters

This section describes some of the parameters for Qwen 3 TTS. For a complete list, see the plugin reference.

base_url
Required
str

Endpoint URL for your deployed model. Takes the shape https://<your-deployment-endpoint>/v1/audio/speech. Find your deployment endpoint in your Simplismart dashboard.

modelstr | TTSModels

Model identifier for the backend TTS model. Set to qwen-tts to use Qwen 3 TTS.

voicestrDefault: Chelsie

Voice identifier to use for generation.

languagestrDefault: English

Language for synthesis.

leading_silenceboolDefault: True

Whether to include leading silence in the generated audio.

Additional resources

The following resources provide more information about using Simplismart with LiveKit Agents.