Skip to main content

​
Introduction

CARTER is the unhinged meme mode of Cartesia — showing what’s possible when you remove all guardrails from Cartesia’s Sonic model. This guide shows you how to build voice AI with personality, not corporate speak.

​
Why Build with Cartesia?

Real Emotions

Express genuine emotions through voice with fine-grained control

Ultra-Low Latency

Sub-6ms response times for natural conversations

Production Ready

Enterprise-grade reliability and scalability

Simple Integration

Clean APIs and SDKs for rapid development

​
Cartesia Sonic Model

The Sonic model powers CARTER’s voice capabilities:
  • Emotional Expression: Control tone, pitch, and emotion in real-time
  • Multiple Voices: Choose from stable voices or emotive character voices
  • Low Latency: 6ms average response time
  • High Quality: Natural-sounding speech with proper pronunciation
  • Streaming Support: Real-time audio generation
Sonic is Cartesia’s latest text-to-speech model, offering unprecedented emotional range and responsiveness.

​
Getting Started

1

Sign Up for Cartesia

Get your API key at cartesia.ai
2

Install the SDK

npm install @cartesia/cartesia-js
# or
pip install cartesia
3

Make Your First Request

import Cartesia from '@cartesia/cartesia-js';

const cartesia = new Cartesia({
  apiKey: process.env.CARTESIA_API_KEY,
});

const response = await cartesia.tts.generate({
  model: 'sonic',
  voice: 'your-voice-id',
  transcript: 'Hello from CARTER!',
  outputFormat: 'mp3',
});

​
Core Capabilities

​
Text-to-Speech

Generate natural-sounding speech with emotional control: 
import cartesia

client = cartesia.Cartesia(api_key="your-api-key")

# Generate speech with emotion
output = client.tts.sse(
    model_id="sonic",
    transcript="I'm excited about this!",
    voice_id="voice-id",
    _experimental_voice_controls={
        "emotion": ["positivity:high", "curiosity"]
    }
)

​
Streaming Voice

Real-time voice generation for conversational AI: 
const stream = await cartesia.tts.stream({
  model: 'sonic',
  voice: voiceId,
  transcript: 'Streaming response...',
  outputFormat: 'pcm_16000',
});

for await (const chunk of stream) {
  // Process audio chunks in real-time
  playAudio(chunk);
}

​
Voice Cloning

Clone voices for custom characters: 
# Upload voice samples
voice = client.voices.create(
    name="Custom Voice",
    description="My custom voice",
    audio_files=[
        "sample1.wav",
        "sample2.wav", 
        "sample3.wav"
    ]
)

​
Integration Patterns

CARTER uses several key patterns you can implement:
Maintain persistent connections for low-latency streaming
const ws = cartesia.tts.websocket({
  model: 'sonic',
  voice: voiceId,
  outputFormat: 'pcm_16000',
});

ws.on('message', (audio) => {
  playAudioChunk(audio);
});
Dynamically adjust voice emotions
output = client.tts.sse(
    model_id="sonic",
    transcript="Amazing!",
    voice_id=voice_id,
    _experimental_voice_controls={
        "emotion": ["positivity:highest", "surprise"],
        "speed": "fast"
    }
)
Maintain conversation context for natural flow
const context = cartesia.contexts.create();

// Each message builds on context
await cartesia.tts.generate({
  model: 'sonic',
  voice: voiceId,
  transcript: message,
  contextId: context.id,
});

​
Next Steps

Integration Guide

Step-by-step integration

Voice API Reference

Full API documentation

Code Examples

Working code samples

Cartesia Docs

Official Cartesia documentation

​
Resources

Remember to keep your API keys secure and never expose them in client-side code.