Voice and the consent line

Synthetic speech has become genuinely good, and much of what it enables is straightforwardly positive. It also contains the sharpest ethical line in this module, and the line is easy to state.

Two different things

Synthetic voice is a voice belonging to nobody — built to sound human, matching no particular person. Use it freely. It is what narrates a video, reads a document aloud, or gives a product an audio interface.

Voice cloning reproduces a specific, identifiable person from samples of their speech. This is a different act, and it requires their permission.

The distinction is not about the technology, which is similar. It is about whether there is a person on the other end of it.

Where synthetic voice earns its place

Worth stating plainly, because the cloning conversation tends to swallow the whole subject.

Accessibility. People who cannot read a screen, or read comfortably, or at all. Lrnon has read-aloud for exactly this reason.

Narration for material that would otherwise have none, because studio time was never in the budget.

Multiple languages for material that would otherwise reach one.

Voice restoration for people losing the ability to speak — banked from their own recordings, with their consent, and one of the clearest goods in the whole field.

None of these need anyone else’s voice.

The consent standard

If a voice is identifiably a real person’s, you need that person’s permission. Not implied, not assumed, not inferred from a public recording existing.

Explicit — they said yes. Informed — they know it is a synthetic reproduction, not a recording. Specific — for this use. Permission to narrate a training video is not permission to make them endorse a product.

A recording being public does not grant this. A podcast, a lecture, an interview — all publicly available, none of them consent to be synthesised.

And note that impersonating a real person can be unlawful quite apart from ethics: fraud, defamation, and in a growing number of jurisdictions specific likeness and synthetic-media provisions. “It was just a demo” is not a defence once it is distributed.

Your own voice, and the family case

Cloning your own voice is fine, and useful — narrate once, correct typos in text rather than re-recording.

Two cautions worth having in advance. A clone of your voice is a credential someone else could misuse, so treat the samples and the account like credentials. And voice authentication systems that identify you by voice are weakened by a good clone existing. Prefer another factor where your bank offers one.

The case that comes up most often, and deserves a real answer: recreating the voice of someone who has died, from old recordings. There is no clean rule. It can be tender and it can be distressing, sometimes for different people in the same family. The person cannot consent, so the question is who is being served — and if the answer is anyone other than the people grieving, it is the wrong project. Ask them first, and accept a no.

Try it now (6 minutes)

Generate a short narration with a synthetic voice — a paragraph of your own writing.

Then write down, in one sentence, the rule you will apply before ever using a real person’s voice. Having decided in advance is the whole point.

Check your understanding

1. The difference between synthetic voice and voice cloning is:

Cloning uses more computation A synthetic voice belongs to nobody; a clone reproduces a specific identifiable person Synthetic voices are lower quality

2. Does a publicly available recording of someone's voice imply consent to clone it?

Yes — public material may be used freely No — consent must be explicit, informed and specific to the use Only if the recording is more than five years old

3. One practical caution about cloning your own voice:

It degrades the original recordings It weakens voice authentication and the samples act like credentials It is prohibited by most providers

Check answers

Recap

Synthetic voice belongs to nobody and is genuinely valuable, especially for accessibility. Cloning a real person needs explicit, informed, specific consent, which a public recording does not supply — and impersonation can be unlawful regardless. Your own voice is yours to clone, with the caution that it is also a credential.