Sometimes you need a voice that does not exist yet: a narrator for a fantasy audiobook, a friendly assistant for your app, a distinctive voice for your brand. With AI voice design you describe that voice in words, and the engine creates it - no recording and no real person behind it.
What is AI voice design?
Voice design generates a new synthetic voice from a written description. You describe qualities like age, pitch, accent, pace and personality; the model turns that description into a voice and speaks a preview line so you can judge it. Once you are happy, the voice is saved to your voices and works in Text to Speech like any library voice.
Voice design, voice cloning or a library voice?
| Option | Best for | What you need |
|---|---|---|
| Library voice | Getting started fast | Nothing - just pick one |
| Voice design | Characters, brand voices, a unique sound without a real person | A good written description |
| Voice cloning | Your own voice, or a voice you have permission to use | A 30-60 second clean sample and consent |
If the voice must be yours, read how to clone your voice with AI instead.
How to write a voice description that works
Vague descriptions give generic voices. The more specific you are about the sound, the closer the result. Cover these elements:
- Age range and gender presentation - "a woman in her forties", "an elderly man", "a young adult".
- Pitch and timbre - low, deep, bright, husky, breathy, warm, crisp, nasal.
- Accent - "a soft Scottish accent", "neutral American English", "a light French accent".
- Pace and rhythm - slow and deliberate, quick and energetic, relaxed with long pauses.
- Emotion and attitude - calm, confident, playful, warm, serious, mysterious.
- Context - "like a late-night radio host", "like a patient teacher explaining something".
Tip: Describe the sound, not the person's looks or job title alone. "A gruff pirate captain" tells the model less than "a rough, gravelly low voice with a West Country accent, speaking slowly and loudly".
Example voice descriptions
Documentary narrator
"A deep, calm male voice in his fifties with neutral British English, speaking slowly and clearly, warm and authoritative, like a nature documentary narrator."
Friendly app assistant
"A bright, friendly female voice in her late twenties with a neutral American accent, medium pace, smiling and helpful but not over-excited."
Fantasy villain
"A low, smooth, slightly raspy male voice, speaking slowly with long, deliberate pauses, quiet and menacing, with a hint of amusement."
Energetic ad read
"A young, energetic female voice with a crisp, clear tone, fast pace and an upbeat, confident delivery, like a radio advert."
Meditation guide
"A soft, warm, gentle voice, speaking very slowly with relaxed breathing and long pauses, soothing and quiet."
Step by step in Pynokio
- Open Voice Design.
- Write your description using the elements above.
- Enter a preview text that sounds like your real use case - a line from your script, not "hello world".
- Generate and listen. If it is close but not right, change one part of the description at a time: the accent, then the pace, then the emotion.
- Save the voice you like. It appears in your voice list for Text to Speech.
Getting consistent results
- Keep a description library. Save the descriptions that worked, so you can recreate a similar voice for a sequel or a new project.
- Use the voice in its language. A voice designed with an English description and preview performs best on English scripts.
- Tune delivery in Text to Speech. Speed, Stability and Clarity work on designed voices too. See our text to speech guide.
What not to do
Voice design is for creating new voices, not for imitating real people. Describing a specific celebrity, politician or private person in order to reproduce their voice breaks our Voice Policy and may break the law. Every generated voice also carries an inaudible watermark, so synthetic speech can be identified later.
FAQ
Do I need to record anything for voice design?
No. Voice design works from a written description only. You only need a recording if you want to clone a real voice.
Can I use a designed voice commercially?
Designed voices are used like other voices on your account, under the Pynokio Terms and your plan. Disclose AI-generated speech where your audience could mistake it for a real recording.
Why does my designed voice sound generic?
The description is probably too short. Add concrete details about pitch, accent, pace and attitude, and use a realistic preview text.
Can I describe a famous person's voice?
No. Voice design must not be used to imitate real, identifiable people.
Describe it, hear it, use it. Create a voice that is unique to your project.
