azzurra-voice is an Open State-of-the-Art Italian Text-to-Speech Model

At Cartesia, we believe that AI should be private, personal, and empathetic. As a frontier research lab based in Italy, we’re dedicated to building social agents and humanoid robots that truly understand and connect with people—not just process their commands.

Today, we’re proud to announce our first model release: azzurra-voice, a state-of-the-art text-to-speech (TTS) model trained on thousands of hours of high-quality Italian speech.

Audio samples generated by azzurra-voice and other leading TTS models.
Sentence azzurra-voice openaudio-s1-mini bark parler-tts piper-tts
La sintesi vocale è un processo complesso
Trentatré trentini entrarono a Trento, tutti e trentatré trotterellando
L'Italia è una Repubblica democratica, fondata sul lavoro. La sovranità...
Senza princìpi morali, i principi del regno persero subito il potere

Why Azzurra?

Azzurra is the beginning of a vision: to create AI that feels Italian not just in language, but in culture, warmth, and presence. We’re building an agent that doesn’t feel like a digital assistant, but like a familiar companion. One that speaks your language, respects your privacy, and understands your emotions.

What Is azzurra-voice?

azzurra-voice is a highly expressive and natural-sounding Italian TTS system designed to run efficiently while delivering top-tier speech quality. It’s been trained with a rich and diverse dataset that spans accents, intonations, and real-life conversational patterns from across Italy. Whether you’re building a smart robot, a localized voice assistant, or accessibility tools, azzurra-voice is your go-to Italian voice.

It’s free, open, and ready to use.

Why Open Weight?

We believe that personal AI should be built in the open. People should feel safe, understood, and in control of the systems they interact with. By releasing azzurra-voice, we hope to empower researchers, developers, and makers to build more inclusive, local, and human-centered AI applications.

The Azzurra Roadmap

azzurra-voice is just the beginning. Here’s what’s coming next in the Azzurra project:

  • Coming Soon: Azzurra-Brain
    A novel Italian large language model (LLM) designed to be friendly, conversational, and emotionally intelligent without the robotic “How can I help you today?” tone. This model will be tuned to sound more like a thoughtful companion than a customer service bot.

  • Later This Year: Azzurra-Pipeline
    A fully local, private, real-time conversational agent that runs on your personal computer. Azzurra-Pipeline will combine:

    • azzurra-voice (TTS)
    • azzurra-Brain (LLM)
    • A State-of-the-Art local Automatic Speech Recognition (ASR) model

The result? A private agent that listens, thinks, and speaks without sending your data to the cloud.

Try azzurra-voice Now

Ready to hear the difference? You can experience the natural and expressive quality of azzurra-voice right now. Head over to our Hugging Face to generate your own Italian audio clips and see how it can bring your projects to life.

Federico Galatolo, Cartesia CTO