A New Voice Engine, and Six Canadian French Voices

The first thing a caller learns about your business is the voice that answers. In Québec, the second thing they learn is whether it speaks their French. District AI now offers a new voice engine, Cartesia Sonic, with seventeen voices: ten in English, one in Spanish, and six in Canadian French. In our own testing it began speaking about a fifth of a second after the assistant had decided what to say.

Nothing changes unless you choose it. Here is what it is, why we added it, and how to turn it on.

Six Canadian French voices

Until now a Québec business had two Canadian French voices to choose from, both on the Canadian engine. Today it has eight, and six of them are new: Étienne, a precise specialist; Félix, a helpful professional; Julien, a polished partner; Amelie, a warm concierge; Madeleine, a reliable resident; and Roxane, a problem-solver.

Cartesia classifies each voice in its library by its native accent. All six of these are classified as native Canadian French: not French from France with a Québec label on the menu, but the French a caller in Longueuil or Trois-Rivières speaks. They were made for the front desk, and each is listed in the picker with the role it was designed for, so you can choose the one that sounds like the person who would answer your phone. Set your assistant's language to French, pick a voice, and your callers are greeted in the French they use.

One thing to know before you choose. These six are synthesised in the United States, like every voice on this engine, and the engine picker says so beside the option. The two Canadian French voices on the Canadian engine are synthesised in Montréal, and that engine is still the one every new Canadian workspace starts on. Your records, your transcripts and your call handling stay where your workspace lives either way. What moves with this choice is the voice, and only the voice.

Seventeen voices, made for the phone

Cartesia builds Sonic, a text-to-speech model designed for live conversation rather than narration. The English set is Skylar, a friendly guide and the default, with Daniel, Gemma, Parker, Jacqueline, Clive, Emma, Arlo, Cora and Lucy. Ximena speaks Spanish. Each voice carries the style it was designed for, from "Customer Care Line" to "Measured Expert", so you can pick by role rather than audition all seventeen.

Why speed is the feature

On a phone line, a pause is information. Half a second of silence after a caller finishes a sentence reads as hesitation. A full second reads as a machine. So before a voice engine goes into the product, we measure how quickly it starts speaking and how cleanly it stops.

Across forty spoken replies through our own agent, Sonic returned the first audio in a median of 0.22 seconds, and 95 percent of replies had begun within 0.53 seconds. When the caller interrupted, it stopped in under a millisecond and sent nothing further. On a phone call, that second number matters as much as the first: a voice that keeps talking over a caller for even half a second is the thing people remember.

Only the voice changes

A District AI engine has three legs: speech recognition, the language model, and speech synthesis. The Cartesia engine changes the last one only. Your assistant still listens with Nova-3 and still reasons with Gemini, so everything it does today, from booking an appointment to transferring a call, works exactly as before. What changes is the voice your callers hear.

That is also why switching is low risk. Choose the engine, choose a voice, save. If you change your mind, choose another. Your greeting, your instructions and the skills you have turned on are untouched either way.

Where it runs

Sonic speech synthesis runs at Cartesia's United States endpoint. For a United States workspace, that is in region. A Canadian or European workspace can choose the engine too, and the picker says so before you choose it: speech synthesis on this engine leaves your region. The two engines that run entirely in Europe stay where they are, and so does the Canadian engine. Cartesia is listed in our sub-processor register and privacy policy since September 2026, alongside the other optional voice engines, and the residency map shows where every leg of every engine runs.

Turning it on

Open your assistant's settings in the dashboard, go to Voice Profile, choose Cartesia as the engine, then pick a voice. Call your own number and listen. Every workspace keeps the engine it has until you change it, so nothing moves on its own.

If you do not have a workspace yet, the plans are on the pricing page, and the District AI page lets you talk to the assistant in your browser first.