How to Clone Your Voice With AI, Step by Step (Legally)

Here is how to clone your voice with AI: record one clean speaker, confirm consent, create the clone in a voice tool and test it with new sentences. Instant clones on ElevenLabs need about one to two minutes of clear audio, while professional clones need at least 30 minutes.

Step 1: Choose instant or professional cloning

Decide first how much time and audio you can spend. Instant cloning is quick and needs a short recording. Professional cloning produces a more consistent voice but needs far more audio and a training step. The table compares the two using ElevenLabs' documentation as of October 2026.

| Option | Audio asked for | Plan access on ElevenLabs | Good for | |---|---|---|---| | Instant Voice Cloning | About 1 to 2 minutes; little gain past about 3 minutes | Starter and up on the pricing table | A quick test or a simple, clear voice | | Professional Voice Cloning | 30 minutes minimum; closer to 2 to 3 hours recommended | Creator and up | A voice you will use often and want to sound consistent |

If your voice has an uncommon accent or a distinctive texture, ElevenLabs' instant guide suggests trying professional cloning. Free plans are not listed for instant cloning on the pricing table reviewed, so check current plan details before you start.

Step 2: Record clean audio

Vendors agree on the basics. Use one speaker, a quiet room and a steady level. ElevenLabs' instant cloning guide asks for audio "without any reverb, artifacts, or background noise of any kind," and warns that the model copies breathing, mouth clicks and noise along with your voice. The specifications below come from the vendor docs:

Level: keep volume steady, ideally between -23 dB and -18 dB RMS, with a true peak of -3 dB.
Format: MP3 at 128 kbps or higher, with 192 kbps or higher in ElevenLabs' FAQ. ElevenLabs says WAV is not needed and can cause upload problems.
Room and mic: a quiet, acoustically treated room, a pop filter, and about two fists between your mouth and the microphone, per ElevenLabs' professional cloning guide.
Delivery: keep tone, pace and energy consistent. Edit out long pauses and filler words such as "uhm," since the clone will copy them. Singing is not supported for professional cloning.

For a closer look at recording setup, read voice cloning sample audio tips.

Step 3: Confirm consent and verification

This is the step most people skip, and platforms have built it into the process. On ElevenLabs' instant flow, you must "confirm that you have the right and consent to clone the voice" before saving. Professional cloning works differently. You can clone only your own voice, even with someone else's consent, and you read verification lines once each. ElevenLabs advises that you then stop, since re-reading a line can cause a failure. After failed attempts, you wait 24 hours or contact support.

ElevenLabs says voice verification is "an ethical and legal safeguard, not a technical requirement," and that it "cannot guarantee that a recording belongs to the requester." The check confirms the person was present and taking part, not who the owner is in every case. The consent steps differ across platforms:

| Platform | Consent mechanism | Access | |---|---|---| | ElevenLabs instant | Checkbox confirming right and consent | Plan-based | | ElevenLabs professional | Voice verification by reading lines; own voice only | Creator and up | | OpenAI custom voices | Scripted consent recording plus a sample from the same person | Eligible customers only | | Google Chirp 3 | Fixed consent statement, up to 10 seconds | Allow-listed users | | Microsoft personal voice | Recorded verbal statement plus a 5 to 90 second sample | Approval required |

OpenAI's consent recording must read one approved phrase, and the guide warns that "any divergence from the script will lead to a failure." Microsoft's custom neural voice is limited access and requires written permission from the voice talent.

Step 4: Create the clone and test it

On ElevenLabs, the instant flow runs in the dashboard:

Go to Voices and click the plus icon, then select Instant Voice Clone.
Upload or record your audio.
Name and label the voice, confirm consent, and click Save voice.
Open My Voices and choose Use voice.

Keep the limits in mind. You cannot download an ElevenLabs clone as a standalone file. You also cannot change the accent or tone after creation, so you must change the samples instead. Clones vary slightly even from the same audio, so keep your originals.

Test the clone with varied text: a question, a short line, a long paragraph, and a sentence with emotion. Listen for drift in accent, pace or tone. If it sounds off, the usual fix is a cleaner or more consistent recording, not more processing.

Step 5: Use it responsibly

Label AI audio when your audience needs to know it is synthetic. OpenAI's usage policies require clear disclosure to end users that a TTS voice is AI-generated. Microsoft's terms also require disclosure. In the EU, Article 50(4) of the AI Act requires deployers of deepfake-generating systems to disclose that content is artificially generated or manipulated, with a lighter form allowed for evidently artistic, creative, satirical or fictional works.

Do not use a clone to impersonate someone else, to deceive a listener, or to make calls that a listener would believe came from the real person. Keep records of your consent and your originals. For the legal picture by country, read voice cloning laws.

How the platforms compare on languages

A voice cloned from one language can carry its accent into another. ElevenLabs' professional guide says this directly. Cartesia builds each clone from a single language, so speak in the language you want the clone to use. Google's en-US consent-based voices can synthesize several other languages, including de-DE, es-US, es-ES, fr-CA, fr-FR and pt-BR. Check the specific tool's language list before you assume a clone will work in your target language.

How OndaLabs custom voices work

OndaLabs custom voices are cloned from a recording you have permission to use, or designed from a text description. A voice clone or design costs 1,500 credits. Custom voice slots depend on your plan:

| Plan | Monthly price (USD) | Custom voice slots | |---|---|---| | Creator | $22 | 3 | | Pro | $49 | 10 | | Business | $199 | 30 |

Commercial use is included on Starter and above, and OndaLabs includes 30 prebuilt voices usable in every supported language. You can review the full plan details on the pricing page. To start, open the custom voices page.

Frequently asked questions

Can I clone my own voice for free?

The ElevenLabs pricing table reviewed on 17 October 2026 did not list instant voice cloning on its free plan. Check current plan details before you start. OndaLabs lists custom voice slots on Creator and higher plans.

Is it legal to clone my own voice?

The rules described in this series target unauthorised replicas and deceptive uses. Cloning your own voice for personal use is the main case that the platforms support, such as Apple's Personal Voice. This is general information, not legal advice, so check the law where you live.

Can I clone someone else's voice if they agree?

It depends on the platform. ElevenLabs professional cloning allows only your own voice, even with consent. OpenAI requires a consent recording from the same person whose voice is cloned, and Microsoft requires explicit written permission from the voice talent.

How long does it take to clone a voice?

Instant clones are usable right after creation, according to ElevenLabs. Professional clones need a training step that ElevenLabs estimates in hours. Microsoft says its personal voice training takes less than 5 seconds, so timing varies by platform.

Can a cloned voice speak other languages?

Some can, but accent carries over. Record in the language you plan to use, since a voice cloned from one language can carry its accent into another. Cartesia builds each clone from one language, so check each tool's language support before you rely on it.