Voice Cloning for Business: Consent, Quality and Use
A practical guide to business voice cloning: when it makes sense, how consent should work, how to check quality, where to limit its use and when not to use it.
A cloned voice is a synthetic voice trained to sound like a particular person. For a business, the appeal is clear: the owner's voice greeting every caller, a consistent brand sound or a familiar voice for regular customers. But a cloned voice is also a person's identity in a form that can be misused. Handling it well is as much about trust and consent as about technology.

This guide covers when cloning makes sense, how consent should be collected and kept, how to record and check quality, where to limit use, and the situations to avoid. It finishes with how the process works in Call. For related topics, see Privacy by Design for Voice AI and AI Disclosure on Calls.
What cloning is and why businesses consider it
A voice clone is built from recordings of a person reading prepared text. The system learns the characteristics of that voice and can then speak new sentences in it. Compared with a stock synthetic voice, a clone can sound more personal, which is why small businesses, creators and brands consider it.
Reasons businesses look at cloning
- A familiar voiceThe owner or a known presenter greets callers.
- ConsistencyThe same voice at every hour, in every message.
- Time savedNo need to re-record every announcement.
- Brand soundA recognisable voice across calls and prompts.
None of these reasons requires cloning. A well-chosen stock voice often serves just as well, with fewer risks. Cloning should be a deliberate choice, not a default.
Consent: the non-negotiable part
The person whose voice is cloned must agree, clearly and knowingly. That means they know what the voice will be used for, where it will speak, who can use it and how they can withdraw. A voice is personal data in many places, and rights over a person's voice and likeness may apply as well. Rules differ widely, so treat the following as good practice and check the law where you operate.
- Get the consent in writing, or as a recorded spoken statement that the person reads in their own voice.
- State the purpose and limits: which lines, which uses, how long.
- Allow withdrawal at any time, and describe what happens to the voice when they do.
- Never clone a voice from public recordings or from someone who is not involved. Clone only your own voice or that of a person who has agreed.
- Keep a record of who consented, when and to what.
If an employee's voice is used, remember that employment does not equal consent, and that people leave. Decide in advance what happens to the voice when they do, and put it in writing. For further reading on handling voice data, see Data Retention for Calls.
A sensible process
From consent to live use
- ConsentThe person agrees and records a consent statement.
- RecordThey read the training text in a quiet room.
- CreateThe system builds the voice.
- CheckListen to samples and test on a phone.
- LimitDecide where the voice may and may not be used.
- MonitorReview calls and keep the right to remove the voice.
Recording well
Quality in, quality out. Record in a quiet, soft room, use a decent microphone at a steady distance, and read in your normal speaking voice rather than a performance voice. Redo any section with a stumble, a cough or noise. A clean recording matters more than a long one.
Checking quality
Listen on a real phone, not only through good headphones. Check names, numbers and the words your business uses. Listen for odd pauses, a flat tone and mispronunciations. Ask a few people who know the real voice what they think. If it sounds close but wrong, it may be worse than a clearly synthetic voice, because callers sense something strange without knowing why.
Quality check before use
- Heard on a phone, not only in headphones
- Business names and common words pronounced correctly
- Numbers, dates and times read clearly
- Tone suits the situation, calm rather than theatrical
- The person themselves is comfortable with how it sounds
Where it fits and where it does not
Good and poor uses of a cloned voice
| Good fit | Poor fit | |
|---|---|---|
| Greetings and fixed messages | A familiar welcome for callers | Not relevant |
| Routine information | Opening hours, directions, simple answers | Advice or negotiation in the person's name |
| Sales | Plain information the caller asked for | Pressure or urgency in a trusted person's voice |
| Sensitive moments | A neutral, clearly synthetic voice | Condolences or complaints in a borrowed voice |
The sharpest test is this: would the caller feel deceived if they found out? A friendly owner's greeting, announced as a synthetic voice, passes. A cloned voice calling customers to suggest the real person is on the line does not. Never use a voice to impersonate a real person to someone who does not know it is synthetic.
Honesty with callers
Tell callers they are speaking with an AI. Using a familiar voice makes this more important, not less. Many places have rules or guidance on disclosure of synthetic voices and automated calls, and they are changing. In the United States, for example, regulators have treated AI-generated voices in some calling contexts as artificial voices, so it is wise to check the current position with the regulator, such as the Federal Communications Commission. See also The EU AI Act and Voice Agents.
Risks to manage
- Misuse: a voice that exists can be copied. Limit who can use it and keep access controlled.
- Withdrawal: make removal possible, quick and complete.
- Drift: a voice used for one purpose gradually spreads to others. Review uses regularly.
- Reputation: an error or an odd sentence in a trusted voice reflects on the person.
- Dependence on a provider: know where voice data is held and how to delete it.
Doing it in Call
In Call, a workspace can record its own voice under Voice, AI voices, with the Clone my voice card. You enter your name, choose the language you will read in, tick the box confirming it is your own voice and that you will not clone anyone else's, and read a consent sentence aloud. You then read the training text paragraph by paragraph, redoing any you want, and create the voice with the provider left on automatic or chosen by you. Before callers hear it, listen to a sample, and test with a call to your own phone from the same screen.
When you are happy, press Use for my agent's calls, choose the conversation engine, and the card shows it as in use. You can delete the voice at any time, after which calls go back to the default voice. Only workspace managers can use or delete it, and a platform operator may review the consent recording before a clone can be used. Separately, Voice profiles record the consent of clients or staff whose voices are recognised, and a consent withdrawal deletes the voiceprint. A phone line can also be given its own voice. See Greeting Scripts for AI Phone Answering for what the voice should say.
Frequently asked questions
Is voice cloning legal?
It depends on the place and the use. Consent, data protection, publicity and consumer protection rules may all apply. Get consent, be transparent and check the rules where you operate.
Can I clone an employee's voice?
Only with their informed, written agreement, which should say how the voice is used and what happens when they leave. Employment alone is not consent.
Can I clone a celebrity or a public speaker?
No. Using someone's voice without their agreement is unethical and may be unlawful.
Will callers know the voice is synthetic?
Not by ear alone, so you must tell them. Disclosing that they are speaking with an AI is good practice and may be required.
How long a recording is needed?
It depends on the system. Quality of the recording matters more than length: a quiet room, a steady voice and clean reading.
What if the person wants their voice removed?
You should be able to delete it promptly and stop all use. Plan for this before you create the voice.