vision-agents-plugins-cartesia 0.6.0

Cartesia TTS integration for Vision Agents

Cartesia

Cartesia is a service that provides Speech-to-Text (STT) and Text-to-Speech (TTS) capabilities. It's designed for real-time voice applications, making it ideal for voice AI agents, transcription pipelines, and conversational interfaces.

The Cartesia plugin for the Stream Python AI SDK allows you to add TTS functionality to your project.

Installation

Install the Stream Cartesia plugin with

uv add "vision-agents[cartesia]"
# or directly
uv add vision-agents-plugins-cartesia

Examples

Read on for some key details and check out our Cartesia examples to see working code samples:

• in tts.py we see a simple bot greeting users upon joining a call
• in narrator-example.py we see a well-prompted combination of a STT -> LLM -> TTS flow that leverages the powers of Cartesia's Sonic 3 model to narrate a creative story from the user's input

Initialisation

The Cartesia plugin for Stream exists in the form of the TTS class:

from vision_agents.plugins import cartesia

tts = cartesia.TTS()

To initialise without passing in the API key, make sure the `CARTESIA_API_KEY` is available as an environment variable. You can do this either by defining it in a `.env` file or exporting it directly in your terminal.

Parameters

These are the parameters available in the CartesiaTTS plugin for you to customise:

Name	Type	Default	Description
api_key	str or None	None	Your Cartesia API key. If not provided, the plugin will look for the CARTESIA_API_KEY environment variable.
model_id	str	"sonic-3"	ID of the Cartesia STT or TTS model to use. Defaults to the recently released Sonic-3
voice_id	str or None	"f9836c6e-a0bd-460e-9d3c-f7299fa60f94"	ID of the voice to use for TTS responses.
sample_rate	int	16000	Sample rate (in Hz) used for audio processing.

Functionality

Send text to convert to speech

The send_iter() method sends the text passed in for the service to synthesize and yields TTSOutputChunks containing the produced PCM audio.

async for chunk in tts.send_iter("Demo text you want AI voice to say"):
    pass