AI Voice Cloning - Clone Your Voice in Minutes
Online AI voice cloning: clone your voice from a few clean samples or design a new character voice from text. Reuse voices across languages and projects.
最終更新: 2026-10-18
VoiceSmiths AI voice cloning copies your own voice from 1–3 minutes of clean audio, and if you have no recording, it can design a brand-new character voice from a one-line description. Cloning costs 30 credits and designing costs 20; either way the voice lands in your library and works across languages and projects. Only clone your own voice, or one you have written permission to use.
At a glance
| Voice cloning | Voice design | |
|---|---|---|
| Input | 1–3 minutes of clean audio | A one-line text description |
| You get | Your own voice | A new voice that belongs to no real person |
| Cost | 30 credits | 20 credits |
| Speech afterwards | 7 credits per 100 characters, like any voice | 7 credits per 100 characters, like any voice |
| Storage | Saved to your library; delete anytime | Preview candidates, save the one you like |
| Best for | Voiceovers, courses and audiobooks in your own voice | Animated characters, brand personas, multi-character stories |
For a first try, the $3.90 trial pack is enough: 150 credits that never expire cover one clone with plenty left to hear it read a script.
Two ways to get a voice with AI voice cloning
Voice cloning: replicate a real voice
AI voice cloning is simple: upload a few clean samples (1–3 minutes, no music, no reverb) and get a reusable voice within a minute. From then on, everything — shorts, audiobooks, courses — can be generated in your own voice, even when you're traveling, sick, or simply out of studio time.
Voice design: invent one from words
No recording available? Describe the voice you want:
"A woman in her late twenties, low and calm, slightly slow pace, resolute and controlled"
The system generates candidate voices to preview; save the one you like to your library. Ideal for animated characters, brand personas and multi-character stories — one distinct voice per character, no accidental overlaps.
What a designed voice sounds like
Neither voice below belongs to a real person — both were generated from a one-line description.
"Woman in her late twenties, low and calm, slightly slow, resolute":
"Older British butler, polite and restrained, faintly wry":
A designed voice in English:
Clone or design: which one
- You want your own voice on voiceovers, courses or audiobooks: use AI voice cloning with one clean recording
- You want a voice that doesn't exist, such as an animated character or a brand persona: use voice design and describe it in a sentence
- You just need a good voice: browse the 54 curated voices and 5,000+ community voices in the voice library first; you may not need to spend credits on a clone at all
How to record a good sample
A clone is only as good as its sample. Before you hit record:
- Quiet beats gear: phone-memo quality is fine; what matters is no background music and no reverb
- 1–3 minutes is enough: cleaner beats longer
- Read in the style you'll use: record in your voiceover voice if you make voiceovers, at your teaching pace if you make courses, so the clone sounds like the finished product
- One recording per language: a voice carries the accent of its source audio, so record separately for each key market
- Clean up noise first: run a noisy take through the studio's voice isolator before cloning — quality improves noticeably
What a cloned voice can do
A cloned voice behaves like every other voice on the site, so you can pick it in any generation tool:
- Speak 70+ languages — note the accent rule: a voice carries the accent of its source recordings, so for key markets, build one voice per language
- Stay consistent long-form: dozens of audiobook chapters or hundreds of lessons without drift
- Work with text to speech: emotion tags and subtitle timing fully supported
- Power the voice changer: upload someone's dry read, keep the performance, swap in your voice
The AI voice cloning consent line (read this)
Only clone your own voice, or one you have written permission to use.
Cloning celebrities, colleagues, or any unauthorized third party violates the Terms of Service and gets the account suspended. This isn't boilerplate — a voice is biometric data. Treat someone's voice like you'd treat their passport.
Quick start
- Record 1–3 clean minutes (phone memo quality is fine — quiet matters more than gear)
- Open Voice Studio → Voices, upload and name it
- Pick your voice in any generation tool
FAQ
How much audio does AI voice cloning need? 1–3 minutes for instant cloning. Cleaner beats longer: phone-memo quality is fine as long as the room is quiet and there is no music or reverb.
How much does AI voice cloning cost? Cloning a voice is 30 credits and designing one from a description is 20; speech in it is then billed like any voice, 7 credits per 100 characters. The $3.90 trial pack has 150 credits, and they never expire. See pricing.
Can I clone someone else's voice? Only with their written permission. Cloning celebrities, colleagues or any unauthorized third party violates the Terms of Service and gets the account suspended.
Can a cloned voice speak other languages? Yes, 70+ languages. The voice keeps the accent of its source recordings, so for key markets build one voice per language.
Who owns the voice? Your samples and the cloned voice are yours; delete them anytime.
Can I use it commercially? Yes, provided the source voice is legitimately authorized, under the Terms of Service. See the AI voice generator page for commercial-use details.
Start cloning
Have a clean recording? AI voice cloning gives you your own voice in a minute. No recording yet? Design one from a sentence and hear it first.