Describe the voice you have in mind, from age and gender to tone, pace and mood. Voice design builds a new one to match.
Age, gender, tone, pace and mood, in your own words.
Type a sentence for the new voice to read aloud.
Listen to the result, then adjust the brief.
Voice design starts from a written description instead of a recording. Say how old the speaker sounds, how fast they talk and what mood they are in.
Concrete words work best. 'A woman in her 30s, warm and slightly husky, unhurried, smiling as she speaks' says far more than 'a nice voice'.
A designed voice is built from your description, not copied from a recording, so it belongs to no real person. Imitating a celebrity or a real person is not allowed.
Voice design suits any project that needs a distinct, consistent voice not tied to a real person.
Give each NPC, villain and narrator a voice of their own. Describe a gravelly old innkeeper or a nervous young squire and keep it for every line.
Write down the voice your brand should have, such as calm, warm and precise, and use the same designed voice in ads, explainers and onboarding.
Design a distinct voice for each main character so listeners can tell who is speaking, and pair them with a steady narrator for the prose.
Apps, kiosks and voice assistants need a voice that fits the product and belongs to no one in particular. A designed voice does both.
Cast a short film or radio play from written descriptions, then rework a role by editing its brief instead of booking another session.
Use a patient, clear voice for lessons and a livelier one for quick tips, so each type of content has a sound learners recognize.
All three end with text read aloud. They differ in where the voice comes from and what you need before you start.
| Compare | AI voice design | Voice cloning | Preset voices |
|---|---|---|---|
| Voice source | Your written description | A recording of a real voice | A ready-made library voice |
| What you need | A few lines of description | A clean clip of the speaker | Nothing, just pick one |
| Sounds like | A new voice, nobody in particular | The person in the recording | The library voice you chose |
| Permission | No real person's likeness involved | The speaker's consent is required | Follows the platform terms |
| Changing it | Edit the description, regenerate | Record a new reference clip | Switch to another voice |
| Best for | Characters and brand voices | Content in your own voice | Quick jobs where any good voice works |
The more concrete the description, the closer the first result lands. Run through these six before you generate.
Open with age and gender, such as 'a man in his 60s' or 'a teenage girl'. It sets the base that every other detail builds on.
Use words people use for voices: soft and breathy, deep and magnetic, bright and youthful, husky, crisp. One or two is plenty.
Say how fast and how lively: slow and unhurried, brisk, measured, bouncy. Pace shapes a character as much as texture does.
Name the feeling and the place: calm late-night radio, a cheerful kids' show, a tense thriller trailer. Context steers delivery.
'Booming yet whispery' or 'very old but childlike' pull in opposite directions. Pick one direction per trait, then adjust.
If a result is close but not right, edit a single word, such as warmer or slower, so you can hear what each change does.
What to know before you design your first voice.
AI voice design creates a new synthetic voice from a written description instead of a recording. You act as the voice designer: describe traits such as age, gender, texture, pace and mood, and the model generates a voice with those qualities to read any script.
Yes. In voice design mode, write a description, add a preview line or fill in an example style with one click, then sign in and generate to hear it. The preview line can be up to 500 characters at 1 credit per 20, and failed runs are refunded.
Two to four sentences is usually enough. Cover who is speaking (age and gender), the texture of the voice, the pace, and the mood or setting. Leave out what you do not care about, and avoid traits that contradict each other.
Voice cloning copies a real voice from a recording, so it needs a clip and the speaker's consent. Voice design needs no audio: it builds a new voice from words. Clone when the content must sound like you; design when you need a character or a voice that belongs to no one.
No. Voice design is for original voices. Requests to imitate a celebrity, a public figure or any specific real person are not allowed. You can describe traits, such as a deep, calm, older male voice, but not name a person.
Yes, both are natural fits. Write a short brief for each character or for your brand's tone, and keep those descriptions together. Reuse the same brief each time you generate so your lines and ads keep a consistent sound.
No. Voice design needs no microphone and no reference clip, only a written description. If the voice you want is your own, voice cloning is the better choice: upload or record a short clip of yourself instead.
Commercial use follows the terms of the plan you choose. A designed voice comes from a description rather than someone's recording, so it carries no one's likeness, but it must still never be used to impersonate anyone or mislead listeners.
Write your first description in the studio, then sign in and hear it.
