Blog / Product

Introducing Multilingual Voices

A hand-drawn blue watercolour globe between two speech bubbles, one reading hello and the other hola

At Cartesia, we’ve always believed both sides have to be best in class: the model powered by the architecture underneath, and the voice that carries it to a listener.

A voice is a bundle of attributes that add up to an identity:

  • Pitch: where the voice sits and how far it moves
  • Warmth: how much air and softness it carries
  • Texture: the grain of the voice, raspy through smooth
  • Pace and emphasis: where it slows down, where it leans in

Getting all of that right in one language is hard. Keeping the same voice intact across many languages is even harder, because a voice’s identity and the language it was recorded in are tangled together.

For years, that left brands with a bad set of options. If you wanted your brand voice in five languages, you hired five voice actors and recorded each market on its own. Your brand ended up sounding like one person in English and a stranger in Japanese.

Today we’re introducing Multilingual Voices: 50+ library voices that now speak up to 25 languages natively, plus the ability to make your own voice clone multilingual. Pick one voice, keep its identity everywhere, and your customers hear the same brand wherever they are.

A language only works if it sounds native to the people who speak it, and that principle shapes how we ship. Every language and accent is scored by native speakers on a rubric, checking whether the accent holds and whether dates, numbers, and currency are read the local way.

Localize your brand voice into a new language

Many of the people we serve don’t have English as their first language, and housing is already hard to navigate. Our agents handle eligibility, income verification, and recertification across multiple languages, around the clock. Across 5M+ calls a month and half a million affordable housing units, we’re making housing more accessible by meeting people in the language they actually speak.

Tony StoyanovCTO

Far too often, we are left with only a voicemail or video of someone we love’s voice, and we wish we had so much more. Forevermore lets you upload that short audio clip and recreate the voice. Even when our families are spread across different languages, each family member can now type in their own language and hear those words spoken in that same familiar voice.

Hasan ZalkoutCTO

Instructions shouldn’t stop at a language barrier. With Cartesia, our customers generate a guide in their own language, then open it to every other language with a single click. The same voice, same guide, with nothing lost in translation.

Dovydas RemeikaCo-founder

Find the right voice for the job

The more industries and use cases we work with, the clearer it becomes: a high-quality voice isn’t automatically the right voice. Beyond the dimensions above, we score every voice against real use cases, not in a vacuum. A voice only makes sense in context. A calm, soothing voice is perfect for a meditation app and wrong for a fraud alert, and it should score lower there.

We've noticed a $2,000 charge on your account.

Soft, sleepy voice

Crisp, confident voice

With 50+ multilingual voices and growing, you can find the right voice for every use case, especially where trust and localization are crucial:

Customer support:

I can see the charge you're asking about. Let me get that refunded for you right now.

Ya veo el cargo del que me hablas. Voy a hacerte el reembolso ahora mismo.

Parker · English

Parker · Spanish

Sales:

You were looking at the Pro plan last week. Want me to walk you through it?

Você deu uma olhada no plano Pro semana passada. Quer que eu te mostre como funciona?

Skylar · English

Skylar · Portuguese

Healthcare:

Your appointment with Dr. Okafor is confirmed for Monday at 9:15.

डॉक्टर ओकाफ़ोर के साथ आपका अपॉइंटमेंट सोमवार सुबह 9:15 बजे तय हो गया है।

Henry · English

Henry · Hindi

Start with our library

To generate, send your text with a locale. Use the same voice ID with any language it supports natively:

POST /tts/bytes

Your order shipped on Tuesday.

In the Cartesia dashboard:

  1. Open the Voice Library and pick a multilingual voice, then More actionsTry in playground.
  2. Choose a language from the language menu.
  3. Type your text and hit play. Same voice, new language.

How to make your voices multilingual

Need a custom voice? Create an Instant Voice Clone from as little as ten seconds of audio:

POST /voices/clone

Then add up to 9 new accents and languages in the dashboard (Voice LibraryMore actionsManage voiceSupported languages & accentsAdd):

Or use the Add Voice Accents API, keeping the same voice_id:

PATCH /voices/{voice_id}/accents

Build voice without language barriers

We’re on a mission to make every voice and model fluent everywhere. Try Multilingual Voices for yourself, and build products people trust, in every language and locale.

Architecting AI that learns and interacts like humans.

Status