Plain AI in plain English
About

In the wild · synthetic voices

Voice clone

30-second gist~30s read

A voice clone is a synthetic copy of someone's voice, made by AI from a short audio sample. Research demos have produced passable clones from as little as 3 seconds of audio. In day-to-day use, the leading commercial services usually want closer to a minute — but that's still a small ask.

That sample can come from anywhere your voice has ever been recorded: a podcast, a voicemail greeting, a TikTok, a video call, a Zoom recording your employer kept.

If you want more

The honest legitimate uses~30s

Voice cloning powers dubbing in films, restores the voices of people who've lost theirs to motor neurone disease, and lets museums "interview" historical figures. Audiobooks increasingly use cloned voices with the original author's consent. The technology itself is neutral.

What to actually do5 steps
  • Agree a family safe-word with anyone who might be impersonated.
  • If a "panicked relative" calls, hang up and call back on a number you already trust.
  • Don't trust caller ID — phone numbers can be spoofed too.
  • Ask something only the real person would know.
  • Transfer no money under time pressure. Ever.

A real example

In April 2023 an Arizona mother answered a call she was sure was her 15-year-old daughter sobbing that she'd been kidnapped, with a man's voice demanding ransom. Her daughter was safely on a ski trip. The "voice" had been cloned from public social-media clips. Police forces in the US, UK, Canada, Australia, and New Zealand have since issued public warnings.