INDUSTRY
Synthetic voices are now a normal part of the market, consent rules are being written into contracts and law, and campaigns go out in more languages than ever. Here is what is changing in voice over in 2026, without the hype in either direction.
AI voices are part of the market now
The question is no longer whether synthetic voices will be used, but where. In May 2025 Audible began offering selected publishers more than 100 AI-generated voices in English, French, Spanish and Italian to turn print books into audiobooks, with AI translation into other languages announced alongside it. Text-to-speech is also widely used for phone systems and fast-turnaround internal content.
At the same time, the market for human performance has kept growing. The Audio Publishers Association reported US audiobook sales of $2.43 billion for 2025, up 9% on the year, with the number of active titles up 43%. AI-voiced audiobooks made up 0.03% of that revenue, and 16% of listeners said they had tried one.
Both things are true at once. Synthetic voice is useful for volume and speed, and it is being used. Listeners and brands are still paying for performances where the voice carries the story or the brand.
The audiobook numbers
$2.43bn, up 9%
US audiobook sales in 2025 (Audio Publishers Association). AI-voiced titles accounted for 0.03% of revenue; 16% of listeners had tried one.
Consent and pay are being written into contracts
The biggest shift for performers has been in the paperwork. In July 2025 SAG-AFTRA members ratified a new Video Game Agreement with 95.04% voting in favour, ending a year-long strike. It requires notice and express consent before a performer’s voice or likeness is used to create a digital replica, along with disclosure about how that replica is used.
In the UK, the performers’ union Equity has been negotiating AI terms for film and TV with the producers’ body Pact. In an indicative ballot that closed in December 2025, 99.6% of Equity’s film and TV members who voted said they would refuse to be digitally scanned on set to secure AI protections, and in January 2026 the two sides went back to the negotiating table.
For anyone commissioning voice work, the practical effect is that consent is becoming explicit. Contracts increasingly say what a recording can be used for, whether it can be used to train or build a synthetic voice, and what happens if it is.
The law is catching up
In the EU, the AI Act’s transparency rules for synthetic content apply from 2 August 2026. Anyone deploying AI to generate or manipulate audio that counts as a deepfake has to disclose that it is artificial. For clearly creative or fictional work, the disclosure can be made in a way that does not spoil the work. The separate requirement for providers to mark AI output in a machine-readable way has been pushed back to 2 December 2026 for systems already on the market.
In the US, the NO FAKES Act, which would give everyone a licensable right in their own voice and likeness, cleared the Senate Judiciary Committee in June 2026. An attempt to pass it by unanimous consent on 30 September was blocked over free speech concerns, so its timing is still open.
In the UK there is no specific law on voice cloning yet. A campaign called Save Our Voices Now, backed by a group of well-known actors, is asking the government to give people legal ownership of their own voice.
Key date
2 August 2026
EU AI Act transparency rules apply: AI-generated or manipulated audio that counts as a deepfake has to be disclosed as artificial.
More languages, more versions
Content now travels further and faster. Netflix said non-English titles made up more than a third of all viewing on its service in the first half of 2026, and brands are planning campaigns the same way, with several language versions from day one rather than one master and a translation later.
That means more casting per project and more attention to accent and region. A Spanish campaign may need Castilian and Latin American versions; a French one may need France, Belgium and Canada. Some of that volume is going to synthetic voices and some to human casts, and the choice usually follows budget, turnaround and how much the voice carries the brand.
What it means for a brief in 2026
A few questions are worth settling at the start of any voice project this year:
Usage and rights. Where will the recording run, for how long, and in which territories. Be clear whether the audio may be used for anything beyond the agreed usage.
Consent. If a synthetic voice is part of the plan, check that the person whose voice it is based on agreed to that use and is paid for it.
Disclosure. If the content will run in the EU and uses AI-generated audio, plan how it will be labelled.
Native voices. For each language version, decide early whether you need a native speaker of a specific region.
Neither hype nor panic
The voice market in 2026 is bigger and more varied than it was a few years ago. Synthetic voices have a place in it, human performance has a place in it, and the rules around both are getting clearer. The projects that go smoothly are the ones that decide early which they need, and get the rights right from the start.
Planning voice work for the months ahead? Send us your brief, call 0207 183 3750 or get in touch.

