What is voice cloning?

Sample requirements have collapsed. What once needed hours of studio recording now works from minutes, and rough approximations come from far less.

Businesses want it for continuity. A practice owner whose voice everyone recognizes, saying the greeting, without recording every line.

Two obligations arrive with it. The person whose voice it is has to agree, in a way you could show someone later.

And the caller still needs to know they are speaking with an assistant. A familiar voice makes that more necessary rather than less.

Say Dr. Patel, who owns your dental practice, wants her own voice on the line because patients have heard it for 20 years. A cloning vendor asks her for about five minutes of clean audio, read from a script in a quiet room. Before she records a word, you'd want one page signed and filed.

That page should say whose voice it is, which lines it may speak, where it will be used and how she can withdraw it. Add an end date as well, because a voice outlives a job. You don't want a former partner's clone greeting your callers two years after she sold her share. Keep the signed page with your other contracts. Her original recordings deserve a locked folder too, since those five minutes are now raw material for anything somebody wants her to say.

Start with ownership. Ask who holds the voice model once it's built, and whether you can delete it and get that confirmed in writing. Then ask whether your sample trains anything beyond your own account. A careful vendor also checks that the speaker agreed. Many have the speaker read a consent line aloud inside the sample, which ties the permission to the voice itself.

Tennessee's ELVIS Act, in force since July 2024, added a person's voice to the likeness rights they can sue over. Other states have written rules of their own. You don't need to track each statute to act sensibly. Treat a staff member's voice the way you'd treat their photo on a billboard: written permission, a stated purpose and a way out.

If that sounds like more paperwork than the idea is worth, a stock voice reading your own wording gets you most of the familiarity. Your callers tend to remember what was said and how fast they got help, and both of those come from your script.

You pick from the available voices. GreetKeeper does not offer cloning of a specific person's voice.

The disclosure line is the part we would push you toward regardless of voice. The assistant can announce itself at the start, and you write that wording.

Utah requires that disclosure verbally at the start for licensed professions, including medicine, law and accountancy, so a dental or legal practice should treat it as a setting to switch on.