New

Voice Design. Live today.

You can design a Parisian French AI voice with one prompt

Describe the accent, gender, age and delivery in one sentence and Gradium returns a Parisian French voice id you call from the API or the SDK. Every Parisian French voice is built to order, so the age, gender and delivery are yours to specify rather than picked from a shelf.

Paris accent, français parisien, Île-de-France accent and parisian french all describe the same voice. Write it as Parisian French in your prompt.

Hear it

  • 0:00

    Regarde-moi ça, le vieux port est encore plein de monde ce soir, c'est incroyable.

The prompt
that produced it

A Parisian French female voice, 20 to 30: mid-high pitch, fast pacing, high energy and bright resonance. Accent detail: uvular r, schwa deletion in rapid speech, clear front rounded vowels. Ideal for advertising reads and social video.

A worked example: change any part of it and the voice changes with it. The four styles below show what moving pitch, pace, energy or purpose does to the same accent.

Pitch2/5
lowhigh
Pace3/5
slowfast
Energy3/5
calmboisterous

Parisian French
in four styles

A prompt is built from parts: the accent, the age, then pitch, pace, energy and what the voice is for. Change one part and you get a variation. These four are starting points to adapt, not finished prompts.

Pitch 3 · Pace 3 · Energy 3

Everything to the middle. Watch The reference register: it is the feature a mid-range read is most likely to soften, and the one a listener places Parisian French by.

A Parisian French male voice, 30 to 45, for customer service: easy and unforced, with mid pitch, steady natural pacing, medium energy and a natural unforced resonance. Ideal for voice agents, assistants and dialogue that has to sound unscripted.

In colour: the 9 parts that make this style.

What makes Parisian French
sound like itself

Each feature below is something the prompt has to encode, or something a careless prompt will accidentally undo.

The reference register
Parisian is the variety French media and teaching treat as standard, which is why a search for a French voice usually wants this. Naming Parisian rather than French narrows the read towards the capital and away from the south.
Schwa deletion
Unstressed schwas drop in rapid speech, so je ne sais pas compresses towards j'sais pas. It is what makes Parisian sound quick and clipped, and requesting very careful articulation removes it.
Uvular R
The r is a uvular fricative at the back of the throat. Getting this wrong is the fastest way to make French sound like a non-native speaker reading French, and it is not something the rest of the prompt can compensate for.
Front rounded vowels
The vowels in tu, peu and coeur have no English equivalent and are structural rather than decorative. A breathy texture value can smear them into one another, so keep the texture clear if the script is vowel-dense.

Design your own custom Parisian French voice
for your use case

The catalogue tags voices at the parent-accent level, so any parisian french voices roll up under Standard French. Voice Design maps this accent, so the page routes to the designer. This is the difference between a voice library and a voice model: a library can only offer you what it already holds, while Voice Design builds the accent on request.

The prompt,
and where it goes wrong

The method behind the prompt: what each clause does, what breaks it, and what it was tested on.

Questions?
Answered

Hear Parisian French
in the studio

Every voice we ship clears blind listener comparison and an ELO ranking first, because roughly one candidate in a thousand has it. For Parisian French, the feature that decides it is usually The reference register.

Paste the prompt, change the parts that matter to you, and generate. No recording, no casting, no licensing.