Voice Design · Custom AI voices from a description

AI Voice Design from a single sentence

Describe the voice you need — a gravelly game villain, a calm documentary narrator, a bright retail announcer — and audition the takes it returns in seconds. Keep the one you like, hear it confirm through the finished voice, and it joins your library for every project. No recording, no speaker, no consent form.

  • Audition before you keep
  • No recording needed
  • Speaks 80+ languages
  • Free trial, cancel anytime
Voice Design · the brief
The brief

A calm, confident woman in her forties with a light Irish accent — warm and unhurried, the voice of a nature documentary.

What it produced
  • The take we keptThe candidate we kept from the brief — “The tide goes out twice a day here, and every time it leaves something behind that wasn't there the day before.”
    Kept take
  • Maeve, savedThe finished voice reading a line it never heard — “Nobody recorded this voice. You described it, kept the take you liked, and now it can read anything you write — this line included.”
    Saved voice
  • Generated by Voice Design, unedited
  • More designed voices below
  • No sign-up to listen
Voice design demos

Don't take our word for it. Read the brief, then press play.

Every clip on this page is the Voice Design feature run for real: a written brief in, candidate takes out, and the take we kept turned into a finished voice. The brief is printed on every card; nothing was re-rolled or retouched. Headphones recommended.

  • Original generated audio
  • Scripts included
  • No sign-up to listen
Designed from a brief

Read the brief. Then hear what it produced.

Briefs a working producer might actually write — a nature documentary narrator, a late-night radio host, a retail announcer, a meditation guide, a game villain, a support agent, a picture-book storyteller — and the voice each one returned. The first card is a finished, saved voice reading a line it never heard; the rest are first takes, unedited.

What does an AI voice designed from a description actually sound like? Like this. Every card prints the brief beside the audio so you can judge how literally it was followed — and the saved voice is what your projects would use.

Maeve Finished voice
Saved voice

Nature documentary narrator

Brief: “A calm, confident woman in her forties with a light Irish accent — warm and unhurried, the voice of a nature documentary.”

Nobody recorded this voice. You described it, kept the take you liked, and now it can read anything you write — this line included.

Unmixr Voice Design
0:07
Cole Designed from a brief
Designed

Late-night radio host

Brief: “A deep, laid-back American man in his fifties, like a late-night jazz radio host — slow, smoky, and smiling.”

It's a little past midnight, the city's gone quiet, and I've got just the record for a night like this.

Unmixr Voice Design
0:08
Tess Designed from a brief
Designed

Retail ad announcer

Brief: “A bright, punchy American woman in her twenties with high energy, announcing a retail commercial.”

This weekend only: everything in store is thirty percent off. Doors open at nine — don't sleep on it.

Unmixr Voice Design
0:05
Anand Designed from a brief
Designed

Guided meditation

Brief: “A soft, slow, gentle man with an Indian English accent, guiding a meditation — calm, spacious, and kind.”

Let your shoulders drop. Notice the breath as it arrives, and let it leave without hurrying it.

Unmixr Voice Design
0:08
Malachar Designed from a brief
Designed

Fantasy game villain

Brief: “A gravelly, menacing older man with a theatrical British delivery — a fantasy game villain who enjoys every word.”

You came all this way to stop me? How touching. Sit. Watch. This is the part where it all falls apart.

Unmixr Voice Design
0:08
Harper Designed from a brief
Designed

Customer support explainer

Brief: “A warm, clear, professional Australian woman in her thirties for customer-support videos — reassuring, never rushed.”

No worries — we can fix this together. Open Settings, tap Billing, and the option is right there.

Unmixr Voice Design
0:07
Fergus Designed from a brief
Designed

Picture-book storyteller

Brief: “A playful, expressive man with a soft Scottish lilt, reading a picture book aloud to five-year-olds.”

And that, my friends, is how the very small dragon learned that a very big sneeze can be a very good thing.

Unmixr Voice Design
0:09

Candidate takes are auditions and cost nothing to reject. A designed voice is a clone of the take you keep, so the app reads one line back through the finished voice before you spend a slot on it. Briefs naming a real person are refused — describe the kind of voice, not a name.

Write your own brief
The design loop

Describe. Audition. Confirm. Keep.

The only step that takes any craft is the sentence. Takes are free to reject; only the voice you save spends a slot.

1

Write the brief

Age, gender, accent, energy, texture, and the job: “a warm, unhurried woman in her forties with a light Irish accent, for a nature documentary.” Add the line you want the takes to speak, or let them read a default.

2

Audition the takes

Candidate voices come back in seconds, each a different reading of the same brief. Nothing is saved yet; if none fit, sharpen the description and design again.

3

Hear the finished voice

Pick a take and Unmixr turns it into a reusable voice, then reads one line back through that finished model — so what you approve is the voice your projects will actually use.

4

Name it and use it anywhere

Save the voice to your library with a “Designed” badge. From then on it is castable in every studio, attachable to a character, and directable line by line like any other voice.

What you get

A voice you can keep, cast, and direct.

A designed voice is not a clip — it is a library voice like any other, with the one difference that nobody had to record it.

A sentence is the whole input

No microphone, no reference clip, no speaker to book. The brief can run to a paragraph if you want, but the ones on this page are a sentence each — and every take was the first round.

Takes to compare, not one guess

Every brief returns several voices that fit it differently. Casting is comparison, and comparing is how a producer actually chooses — the same brief, different readings, pick one.

Confirmed before it is kept

A designed voice is a clone of the take you choose, and extreme timbres can drift. The app reads a line through the finished voice before you save, so a slot is never spent on a surprise.

A voice, not a one-off clip

Saved voices live in your library with the ones you cloned and the ones we ship. Attach one to a character and a game villain or recurring narrator sounds the same in episode forty as in episode one.

Speaks 80+ languages

Designed voices are built on the same model as cloning, so a voice briefed in English can read Spanish, Hindi, or Japanese — with the accent you described carrying across, the way a bilingual speaker's does.

Original by construction

Briefs that ask for a specific real person are refused before anything is generated. What you save is a voice that never existed, which is exactly what makes it safe to publish and sell.

Design or clone?

Invent a voice, or reproduce one.

Both end as a library voice you can cast anywhere. The difference is what you start with — and whether a real person is involved.

Question Voice Design Voice Cloning
Start from A written description of the voice you want. A 30–80 second recording of a real speaker.
The result An original voice that has never existed. A reproduction of that specific person's voice.
Consent None needed — no person is involved. Briefs naming a real person are refused. Required: your own voice, or written permission from the speaker.
Best for Characters, house voices, narrators nobody has to be, casts of many. You, an instructor, an approved brand spokesperson.
Auditioning Several takes per brief; reject freely, refine the brief, go again. One clone per recording; re-record for a different result.
After saving Identical: library voice, castable everywhere, 80+ languages, directable per line. Identical: library voice, castable everywhere, 80+ languages, directable per line.
Limits 5 designs a day and 1 kept voice free; 10 a day and 5 kept on Starter; 30 a day, unlimited kept on Pro and Business. 5 clone profiles on Starter; unlimited on Pro and Business.

Limits reflect the current plan matrix; check the pricing page before you buy. A saved designed voice and a saved clone behave identically in every studio.

Where designed voices go to work

The voices teams design here.

Built for game and animation studios, audiobook and podcast producers, brand and L&D teams, and anyone who needs a voice nobody has to be.

Game and animation characters

A cast of dozens — villains, shopkeepers, companions — each designed from a line of character notes and locked to the character for every future line.

Audiobook narrators and casts

Design the narrator your readers hear in their head, then a distinct voice per recurring character, without hiring a cast.

See the audiobook workflow →

A house voice for the brand

When there is no spokesperson to clone, design the voice the brand should have — and every team member casts the same one.

Have a real voice? Clone it instead →

Instructors for a course

A calm, clear teaching voice for a curriculum nobody on the team wants to narrate — consistent from module one to forty.

See course narration →

Podcast hosts and guests

Two hosts with real contrast in pace and register, designed to sound good together, for a dialogue show you script.

See dialogue podcasts →

Ads, IVR, and product voices

A punchy announcer for the campaign, a reassuring voice for the support line, a distinct voice for the app — each an original you own the rights to use.

100+ Languages across voiceover, transcription, subtitles, and dubbing
80+ Languages a designed voice can read, with the accent you briefed
500M+ Minutes of audio and video generated by creators worldwide
FAQ

Voice design: your questions, answered

Straight answers about designing a voice from a description.

What is AI voice design?

Voice design creates a brand-new synthetic voice from a written description instead of a recording. You describe the voice you want — age, gender, accent, tone, energy, the job it is for — and Unmixr generates a set of candidate takes that match the brief. Pick the one you like, hear it confirm through the finished model, then name it and save it to your library, where it works like any other voice in narration, dialogue, video, and dubbing projects.

How is voice design different from voice cloning?

Cloning reproduces a voice that already exists from a short recording, so it needs a real speaker and their consent. Voice design invents a voice that has never existed from a description, so there is no recording, no speaker, and no consent question. Use cloning when the voice has to be a specific person — you, an instructor, an approved brand narrator. Use design when you need a voice that fits a role — a character, a house style, a narrator nobody has to be — or when you have no one to record.

How do I write a good voice brief?

Name the things a casting director would: age range, gender, accent or region, energy, texture (warm, gravelly, bright, breathy), pace, and the job — “a late-night radio host” says more than three adjectives. One clear sentence usually lands on the first round; the briefs on this page are printed beside the voices they produced so you can judge how literally each was followed. If the takes miss, sharpen the description and design again — rejecting takes costs nothing.

Can I design a voice that sounds like a famous person?

No. Briefs that ask for a specific real person — by name, by “sounds exactly like”, or by title — are refused before anything is generated, the same standard the cloning consent gate applies. Describe the kind of voice you want instead. A designed voice is an original, which is also what keeps it safe to publish commercially.

What happens between choosing a take and saving the voice?

The take you keep is turned into a reusable voice model, and the app reads one line back through that finished model before you save. That step exists because the saved voice is a clone of a short clip and unusual timbres can drift — what you approve is the voice as it will actually sound in your projects, not the audition. Rejecting it at this point costs nothing; only saving spends a slot.

How many voices can I design and keep?

Designing is metered per day and keeping is metered by slots. A free account can run 5 designs a day and keep 1 voice; Starter allows 10 designs a day and keeps up to 5; Pro and Business allow 30 designs a day with unlimited kept voices.

Does a designed voice cost credits to use?

Designing and saving a voice does not spend credits — it spends the daily design allowance and, on save, a kept-voice slot. Generating speech with the saved voice costs the same as any other voice: one credit per character, from the same pool that covers voiceover, dubbing, and transcription.

Where can I use a designed voice?

Everywhere a voice can be chosen: Quick Voiceover, Narration Studio, Dialogue and Story projects, Course and Training videos, Read Aloud, and Dubbing Studio. It appears in your Voice Library with a “Designed” badge and can be attached to a saved character so a recurring narrator or game character keeps the same voice and delivery settings across every project.

Can a designed voice speak other languages?

Yes. A designed voice is built on the same voice model as cloning, which generates speech in 80+ languages, so a voice designed from an English brief can read Spanish, Hindi, or Japanese text. Design the brief in the language and accent you want the voice to have; the accent carries into other languages the way a real bilingual speaker's does.

Can I direct how a designed voice performs a line?

Yes. Pace, pauses, emphasis, and pronunciation rules apply to designed voices exactly as they do to any other voice, and plain-language delivery directions work in the studios that support them. The brief sets who the voice is; direction sets how it reads a given line.

Can I use designed voices commercially?

Yes. On any purchased plan, every audio file you generate with a designed voice is 100% cleared for commercial use — games, ads, courses, audiobooks, podcasts, client work, and anything else you publish or sell. Because the voice is an original rather than a copy of a person, there is no likeness consent to obtain.

Is there a free trial, and does it need a credit card?

Yes to both. The 7-day trial includes 20,000 credits and starts with a payment method on file; cancel before it ends and nothing is charged. A free account with 5,000 credits is also available, and it can design voices right away — 5 designs a day, one kept voice — so you can hear your brief before deciding anything. Plans start at $19/month for Starter with 400K credits every month; Pro is $49/month and Business $99/month.

How long does it take to design a voice?

Seconds. Candidate takes come back almost immediately, the confirmation read takes a few seconds more, and saving is instant. Most of the time goes into listening — and into the second brief, once the first round has shown you what to ask for more precisely.

Plans from $19/month — every paid plan includes commercial rights for the voices you design.

Free account: 5 designs a day, 1 kept voice · Starter: 10 a day, 5 kept · Pro and Business: 30 a day, unlimited kept.

Free trial, cancel anytime · Compare plans

The voice you need is one sentence away

Write the brief, audition the takes, and keep the one that sounds like the voice in your head — then put it to work in any project.

Design a Voice Free

Free trial, cancel anytime · Commercial rights on every paid plan · Designed voices speak 80+ languages

Still have a question?

If you have a specific use case or need help configuring a custom workflow, our team is always here for you. We would love to listen to your needs!