Cartesia

Cartesia is a text-to-speech plugin. It turns the LLM’s reply into audio the caller hears.

Setup

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

export CARTESIA_API_KEY=<key>

Usage

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

from zeroruntime.plugins import CartesiaTTS

tts = CartesiaTTS()

# Pipeline(tts=tts, ...)

import { CartesiaTTS } from '@zeroruntime/js-sdk/plugins';

const tts = CartesiaTTS();

// Pipeline({ tts, ... })

Parameters

Constructor parameters for CartesiaTTS. The Python and Node JS SDKs share these field names.

Parameter	Type	Default	Description
`api_key`	`str`	`None`	Cartesia API key. Falls back to the `CARTESIA_API_KEY` environment variable when unset.
`model`	`str`	`"sonic-2"`	Cartesia voice model.
`voice_id`	`str \| list[float]`	`"f8f5f1b2-f02d-4d8e-a40d-fd850a487b3d"`	Voice to speak with: a voice ID or an embedding vector.
`language`	`str`	`"en"`	Language of the spoken output.
`generation_config`	`GenerationConfig`	`None`	Fine-grained synthesis settings such as speed, volume, and emotion.
`pronunciation_dict_id`	`str`	`None`	ID of a custom pronunciation dictionary to apply.
`max_buffer_delay_ms`	`int`	`None`	Maximum time to buffer text before synthesizing, trading latency for smoother audio.
`enable_word_timestamps`	`bool`	`False`	Return per-word timing alongside the audio.

generation_config can be passed as a dictionary with Cartesia’s supported voice-shaping fields:

speed: 0.6 to 1.5 (1.0 = default speed)
volume: 0.5 to 2.0 (1.0 = default volume)
emotion: one of "neutral", "calm", "angry", "content", or "sad"

from zeroruntime.plugins import CartesiaTTS

tts = CartesiaTTS(
    generation_config={
        "speed": 0.9,
        "volume": 1.2,
        "emotion": "calm",
    }
)

Import paths

SDK	Import	Constructor
Python	`from zeroruntime.plugins import CartesiaTTS`	`CartesiaTTS()`
Node JS	`import { CartesiaTTS } from '@zeroruntime/js-sdk/plugins'`	`CartesiaTTS()`

Spoken replies support interruptions. When the caller talks over the agent, the runtime stops the audio and listens again.

Plugins

Inference Gateway

Speech-to-text

LLM

Text-to-speech

Realtime models

Avatars

Turn detection

Voice activity detection

Noise cancellation

Setup

Usage

Parameters

Import paths

​Setup

​Usage

​Parameters

​Import paths

Setup

Usage

Parameters

Import paths