AI voice design
Create a voice from a description
Write who should speak, in a sentence, and Voxmint turns it into a voice. No microphone, no recording, and no real person behind it.
What is AI voice design?
AI voice design creates a new synthetic voice from a written description instead of a recording. You describe the speaker (age, gender, accent, tone and pace) and the model generates speech in a voice that matches. Voxmint designs voices with the open VoxCPM2 model, and most of the ready-made voices in its catalogue began as one such line.
One line in, one voice out
Each of these catalogue voices was made from the description above its sample.
“a deep male documentary narrator, late forties, slow and deliberate, warm and authoritative”
“an elderly woman with a gentle, slightly raspy storyteller's voice, kind and unhurried”
“an intense male voice, thirties, urgent and clipped, like a thriller trailer”
“A woman in her late thirties with a refined British Received Pronunciation accent, a smooth, medium-low voice with a velvety texture, reading at a measured, unhurried pace with warm and intelligent poise, like a literary audiobook narrator”
Female · adult · British RP
“It was the sort of evening that made you forgive the whole of February. One by one, the lamps came on along the river, and nobody was in any hurry to go home.”
“A young woman in her late twenties, a native Vietnamese speaker from Hanoi with a Northern accent, a clear, medium-high voice with a soft, bright tone, speaking at a smooth, even pace with gentle, polite warmth, like a television presenter”
Female · young adult · Northern (Hanoi)
“Xin chào quý vị. Hà Nội những ngày cuối thu, trời se lạnh, hoa sữa nồng nàn trên khắp các con phố cổ.”
Good evening, everyone. In Hanoi’s last days of autumn the air turns chilly, and the scent of milkwood flowers fills the streets of the Old Quarter.
“a Hong Kong woman of about thirty speaking Cantonese, a clear, soft voice, natural pace, warm and poised, like a radio host”粤语，三十岁左右的香港女性，声音清亮柔和，语速自然，亲切大方，像电台主持人
Female · adult · Cantonese (Hong Kong)
“各位聽眾，早晨！今日天氣好好，陽光充足，不過下晝可能有驟雨，出門口記得帶定把遮啊。”
Good morning, listeners! It’s a lovely sunny day today, but there may be showers this afternoon, so remember to take an umbrella when you head out.
How it works
Step 1
Describe the voice
In the studio, choose Voice design and write a line: who is speaking, how old they sound, their accent, mood and pace.
Step 2
Generate your script
Type or paste the text and generate. The description is free; only the text you voice counts as characters.
Step 3
Keep the one you like
Save the description under Voices to reuse it. Set a seed in the settings to get the same take of that voice every time.
Writing a good description
Name the qualities you can hear. Five details are usually enough:
- Age and gender
- "a woman in her late fifties", "a young man, early twenties"
- Accent or region
- "a Scottish accent from Edinburgh", "a native Vietnamese speaker from Hanoi"
- Pitch and texture
- "a deep, velvety voice", "bright and slightly raspy"
- Pace and delivery
- "slow and deliberate", "quick and upbeat, like a radio host"
- Role
- "a documentary narrator", "a patient teacher", "a thriller trailer voice"
Starting points
The studio offers these as one-click presets. Change any word.
- A warm male voice, mid-thirties, calm and measured
- A bright young woman, cheerful and energetic
- A deep documentary narrator, slow and deliberate
- An elderly storyteller, gentle and a little raspy
- A crisp newsreader, neutral and professional
Voices are synthetic only: a description that names a real or famous person is refused. Read the voice policy.
Questions
- Is voice design the same as voice cloning?
- No. Cloning copies a real person from a recording. Voice design starts from words only, so the voice belongs to no one. Voxmint does not offer cloning.
- Can I describe a celebrity or a real person?
- No. A description that names a real or famous person is refused. Describe the qualities instead: age, accent, mood and pace. See the voice policy.
- Will the voice sound the same every time?
- Leave the seed blank and each take is a fresh variation on the description. Set a seed in the studio settings and the same description gives the same voice, which is how the catalogue voices stay consistent.
- Does the description cost credits?
- No. Only the characters of the text you turn into speech are counted. The description, a style instruction and the prompt preview are free.
- Which languages can a designed voice speak?
- The same ones as the rest of Voxmint. Write the description in English and the script in the language you need; the 24 languages with ready-made voices are the best tested.
- Can I use a designed voice commercially?
- Yes, on any paid plan or with a credit pack. The Free plan is for personal use.