Choosing an AI voice for your brand: gender, tone, language

12 October 2026 · 7 min read

Brands spend weeks on a logo and thirty seconds on a voice. Yet the voice is what your audience remembers longest: it opens your YouTube videos, answers the phone, reads your training modules and signs off your TikTok clips. With a synthetic voice the question is no longer “who will record it?” but “which voice, and in what tone?”. Here is a one-hour method to decide, then never revisit it. The examples use Aievalu's French and English voices, but the grid works with any tool.

Voice is part of the identity, just like colour

A brand that speaks with three different voices depending on the channel sounds like three brands. Listeners don't analyse it, they feel it: yesterday's video and today's phone greeting don't come from the same company. The whole point of a synthetic voice is that it can be the same everywhere, without depending on an actor's diary or a studio booking. So treat the voice as a brand asset: chosen once, documented, reused.

The decision grid: who listens, where, and for what

Three questions are enough. Who are you talking to (general public, business customers, learners, callers in a hurry)? On which channel (long video, short format, audio only, phone)? To say what (inform, persuade, reassure, entertain)? The answers give you the tone before you even pick the voice.

• Inform (tutorial, training, instructions, phone menu): a neutral, even tone that doesn't tire over twenty minutes. Paul or Jane “Neutral” in English, Marie “Neutre” in French.
• Hook (TikTok, Reels, trailer): a cheerful tone that brings energy from the first second. Paul “Cheerful”, Marie “Enjouée”.
• Persuade (product walkthrough, corporate video, pitch): a confident, measured tone that doesn't oversell. Paul or Jane “Confident”.
• Reassure (customer service, hold message, follow-up): neutral and warm, never bubbly. Marie “Joyeuse” works nicely in a welcome message, less so in an outage notice.

Our format guides cover each case: short formats, YouTube, podcasts, phone systems and corporate video.

Gender and age: a decision, not a reflex

“A female voice to reassure, a male voice to sound serious” is a cliché and a poor guide. What matters is consistency with who you already are: if your founder fronts the brand on video, a female voice extends that presence; if your content is carried by a mixed team, pick whichever voice sounds best on your text and leave it there. Same logic for perceived age: a youthful voice on a retirement savings offer feels off, and so does a very measured voice on a gaming channel. The blind test below settles this better than any general rule.

Language and accent: one persona per language

A French voice won't read English text correctly, and vice versa. A bilingual brand therefore has two personas, one per language, chosen once: for example Marie in French and Paul (American accent) or Jane (British accent) in English. The English accent depends on your audience: North American customers, Paul; UK and Europe, Jane or Oliver. Both personas should share the same tone in the same context: Jane “Neutral” for the English training, Marie “Neutre” for the French version. The FAQ lists the languages available.

The same-text blind test

Write one sentence of up to 160 characters that sounds like you: your promise, your name, one number. Generate it with the free trial in three tones of the same voice, then with the candidate voice of the other gender. Play the files to two colleagues without telling them which is which and ask one question: “which one is us?”. You'll have an answer in ten minutes, and almost always the same one. That is the principle behind the samples on the Voices page: an identical text for every voice, so you compare like for like. One tip: put your brand name in the test sentence and check its pronunciation before anything else; if the engine gets it wrong, try another spelling (spaced letters, dots) and write down the one that works.

Write the voice guide: one page, no more

A choice is useless if it lives in one person's head. Write a page anyone can follow, including a contractor or an AI drafting your scripts six months from now:

• Main voice per language (name and default tone).
• Tone per context: tutorial, short format, greeting, promotion.
• Pronunciation spelling of the brand and product names (the one that passed the test).
• Number format: dates in words, prices with the currency spelled out, phone numbers in pairs.
• Banned or replaced words (badly read loanwords, internal jargon).
• Audio signature: the closing sentence, identical everywhere.

Pair it with the writing rules from our article on the seven mistakes that make an AI voice sound robotic: a good voice guide is first of all a writing guide.

Consistency over time, where preset voices shine

An actor recorded in January and called back in September doesn't sound quite the same: fatigue, microphone, room, mood. A preset voice doesn't drift: the same sentence generated a year from now will give the same result. For a brand that is a rare comfort. It comes with one constraint: if you change the voice, change it everywhere at once, and log the date in the guide.

When to use two voices

Two legitimate cases. Dialogue: a question and an answer, a customer and an agent, a heroine and a narrator; two voices, or two tones of the same voice, generated separately and assembled in the edit. And the host-plus-narrator duo of a podcast or a course, where the second voice reads the sidebars, definitions and transitions. Beyond that, every extra voice dilutes the identity.

What it costs

A complete brand kit, about fifteen messages (greeting, hold, opening hours, video intro and outro, signature, three announcements), is roughly 5,000 characters: five minutes of audio and under fifty cents with the €9 pack. The test sentence fits in the free 160-character trial, no account needed. Pricing is per character: you only pay for what is produced, and audio generated with paid credits is yours, commercial use included (see the rights FAQ). To compare French voices before you decide, our listening test of French AI voices goes through the options.