More than 80 actors and creative-industry figures have signed an open letter to Prime Minister Andy Burnham asking for a law that would make your voice something you legally own. We think they're right — and that the line the campaign is actually drawing gets lost every time this story is filed under "actors versus AI." The line isn't synthetic versus human. It's consented versus taken. Those are different arguments, and only one of them is winnable.
What was actually asked for
The campaign is called Save Our Voices Now. Its ask is narrow and legal rather than rhetorical: legislation giving people ownership of their own voice, with control over how it can be replicated by AI systems. A Gov.uk petition runs alongside the letter, asking for voice ownership to be a statutory right in Britain.
The signatories reported so far include Hugh Bonneville, Nicola Coughlan, Matt Lucas, Luke Evans, Jen Brister, Siobhán McSweeney and Pearl Mackie. Co-founder Peter Caulfield has framed the urgency in terms of how little material a clone now needs — three seconds, on the right system. The campaign also cites its own figure that 28% of UK adults report having been targeted by a voice-cloning scam, which is a claim about fraud rather than about publishing, but it explains why this has political traction at all.
For a precedent, campaigners point to Denmark, where a copyright amendment would give individuals rights over AI reproductions of their voice and likeness — treating the self as something with an owner rather than something merely defamed.
Ready to try it on your own book?
Start your first chapter free →The detail almost everyone skipped
Save Our Voices Now was co-founded by a working audiobook narrator, Alice Sockett, alongside actor Peter Caulfield. Sockett has described unauthorised cloning as "an existential threat to our entire industry."
Read that again, because it isn't incidental. The campaign that pulled in Bridgerton and Downton names started in the audiobook booth. Ask why, and the whole fight becomes legible.
A published audiobook is the single best recording of a professional voice that exists anywhere. Ten to fifteen hours of one person, one microphone, one treated room, consistent level, no music bed, no crosstalk, no crowd. Cleanly enunciated, emotionally varied, and covering nearly every phoneme in the language. Film gives you a voice buried under a score. Podcasts give you a voice in a kitchen. An audiobook gives you a studio-grade corpus — and unlike a session file, it is published, catalogued, and purchasable by anyone with a credit card.
Narrators are first in line because their finished work is also, by an unlucky accident of format, the highest-quality training material on the market. They didn't wander into this argument. They were standing where it landed.
"Regulate AI" is the wrong ask. "Show consent" is the right one.
The framing that gets this campaign dismissed is the one where it becomes a referendum on synthetic voice. That argument can't be won, and it isn't the one the letter makes. The letter is about ownership: whose voice is it, who said yes, and what happens when nobody asked.
That distinction matters commercially, not just morally. A voice built for a character in a novel and a voice lifted from a working narrator's back catalogue are not the same product, the same risk, or the same ethics — even when a listener can't tell them apart in a waveform. Collapsing them into "AI voices" flatters the second and slanders the first.
It also points at the part of this that doesn't need to wait for Parliament. Statute is slow, jurisdictional, and — as Denmark's own timeline shows — subject to elections. Provenance is not. Whether a voice in a production was consented to is a fact that exists at the moment the voice is made, and any tool that makes voices can require it, record it, and refuse without it. Legislation would set the floor. Nothing stops a product from being above the floor today.
What that looks like in a tool, concretely
We build dramatized audiobooks with full casts, so we're on the hook for this question every time someone opens a project. Two things about how it works:
The default path doesn't clone anyone. Character voices in Midsummerr are designed. You write a description of how a character should sound — age, texture, temperament — and generate a voice to match it, then regenerate until it's right. The cast that ends up on your book was made for your book. It isn't assembled from anyone's published narration.
Cloning exists, and it's gated. On hybrid-mode projects you can clone a voice for a specific character — most often your own, for author narration. Before that runs, the modal requires an explicit confirmation that you own the rights to the sample and have any consent required. It isn't a checkbox you can skip past: the submit button stays disabled until it's ticked. It's per-character and opt-in, not a global "clone anything" switch. We wrote about the workflow when it shipped, in bring your own voice and in the author-narration guide.
That's a rights decision wearing a sound-design costume, and it belongs at the point of use — in front of the person clicking the button, not buried in terms nobody opens.
What this changes for a book you're producing now
Three practical things.
Know which of the two things you're doing. Designing a cast is a creative decision. Cloning a real person's voice is a rights decision that happens to sound like a creative one. If the voice belongs to someone — you, a co-author, a friend with a great read — get it in writing, and keep it. That record is the entire asset if anyone ever asks.
Don't assume cloning solves distribution. A cloned voice is a synthetic voice, including when it's yours and you consented. ACX does not accept third-party AI narration unless otherwise authorized, so if your plan is to go wide through Audible, cloning your own voice isn't the shortcut it looks like. Plan the distribution first and the voice second.
Expect disclosure to become table stakes. Whichever way the UK bill goes, the direction of travel is toward provenance being something you can be asked for. Productions that can answer cleanly won't notice. Productions that can't will.
The read
The easy version of this story is actors versus robots, and it's wrong in a way that helps nobody — least of all the narrators who started the campaign. The real fight is about a boundary that's been trivially cheap to cross and has had no consequence attached to it: taking a voice that was published, not offered.
The campaign started in the audiobook booth because that is where the boundary is thinnest. The rest of the industry should take the point on its merits rather than on who's shouting: consent is the line, provenance is the evidence, and neither of those is an argument against making audio. It's an argument about asking.
You can hear what a designed cast sounds like on our sample productions — every voice on them was made for the character, not borrowed from a person.
