Could AI keep a brand's voice alive forever?

For 23 years - since the brand's creation - the Yorkshire-born Bafta-winning actor Sean Bean has been the voice of O2's brand campaigns, created by VCCP, lending the mobile company some of his down-to-earth grittiness and gravity.

The most recent outing saw Bean lend his voice to 'First Breath', a spot which imagines a newborn baby's future life.

In an interview earlier this year Rachel Swift, brand and marketing director at O2, told Creative Salon why Bean was so important to the brand. "It's really easy as a marketer, to want to change things. But he's so synonymous with the O2 brand, and especially when you're changing a platform, you want to keep some consistency. He does such a fantastic job, and actually, we've elevated his role in our comms to be much stronger now." Indeed such is Bean's importance that O2 also uses his voice for internal sessions.

His partnership with O2 is a powerful example of what System1 calls 'Fluent Devices' - repetitive and established elements within ads that use the positive emotion and associations they generated could be carried over from ad to ad, helping long-term brand-building. The 'fluidity' comes from the elements becoming recognised immediately by the viewer, making recognition faster and delivering positive emotion more quickly.

But what happens when an established and long-standing celebrity voiceover artist is no longer able to fulfill the role - either through ill-health or because they are reaching the end of the circle of life as Bean so eloquently describes in the O2 'First Breath' ad? Morbid perhaps, but it's something that many advertisers and agencies that rely on this form of device will have to inevitably consider at some point.

For instance, HSBC was faced with such a dilemma as its long-standing brand spokesman, Michael Gambon aged. He had been responsible for its voiceovers since 2005.

In 2018, as he reached his late 70s, Gambon was replaced by the comedian Richard Ayoade. He died just five years later.

Similarly, VCCP has used the voice of the 'I'm Alan Partridge' actor Simon Greenall for Orlov, the Russian meerkat, in its advertising for Comparethemarket.com since 2009.

Sadly, celebrities are not immortal, but perhaps death is no longer the end of their brand legacies, given the rise of AI and technology such as ElevenLabs, Synclabs, and OpenAI's GPT 4.0. Their voiceovers becoming eternal is becoming a potential reality.

By adopting AI-generated synthetic voiceovers of celebrities, advertisers could rule out the upheaval of finding and establishing the credibility a new spokesperson - thereby keeping the 'fluency' of the device that has been long-established.

After all, last year T&P created an AI clone for the (very much alive) football manager Jose Mourinho that could deliver personalised messages to fans.

But is it desirable - or indeed ethical? That is already playing out.

Just because we can, should we?

A lawsuit is set to be heard in New York, brought by two voice-over artists who allege their voices were stolen by a start-up business. They believe their voices have been used to train an AI model. Already, the potential ramifications for artists are becoming clear.

Peter Gasston, innovation lead at VCCP and faith, says the technology is already more than capable of replacing human voice-overs. “One of the ways synthetic voice can be used is to replicate the audio likeness of an existing voiceover artist, especially of a celebrity or talent strongly associated with a brand. There’s a two-fold case for this: efficiency and effectiveness.”

Recording voice talent, he points out, requires forward planning, budgets, and logistics – fine for long sessions, but cumbersome for quick-turnaround work.

“Having a synthetic model of the talent’s voice enables quick edits and quick responses,” he says. “Pickups or reactive social content become far simpler.”

Beyond speed, Gasston sees creative potential: “Imagine receiving a WhatsApp voice message from the brand, in the talent’s voice, saying your name. Or hyper-localised radio spots recorded in seconds. And synthetic voice could even open up opportunities for smaller brands, if voice talent licenses cheaper synthetic usage rights.”

But will AI voices truly replace living, breathing celebrities? Menno Kluin, global chief creative officer at Iris, is sceptical. “In most cases, the choice of celebrity voice is partially driven by the ego of the CMO or CEO - the desire to meet and work with their heroes, he says. "Yes, you could buy the rights to a great voice, but it doesn’t come with the same story.”

In short, real celebrities bring human narratives - and flaws - that AI cannot replicate. “Yes, an AI voice is scandal-proof,” Kluin says. “But it won’t give you relevance or uplift. Human flaws and uniqueness create controversy and genius – both are needed to be part of pop culture.”

He also warns that AI might not actually reinforce brand consistency as hoped. “We’ve used AI mostly as a creative exploration tool – testing voices on rough cuts to see if the investment makes sense. Often, we’ve found multiple voices in the same genre can achieve the same result.”

For Kluin, AI voices may make the most sense in functional contexts – local adaptations, fragmented formats, logistics-heavy campaigns – rather than in brand-defining work.

Bridget Fahrland, DEPT’s VP of Applied AI, believes the debate isn’t just about capability – it’s about responsibility. “There’s what you can do with AI, and then there’s what you should do. When and how brands use AI for commercial voiceovers depends on the brand, the format, and the run length,” she says.

Fahrland offers some guidelines for when synthetic influencers might - and might not - be appropriate:

• Don’t use AI voices for large-scale hero ads – emotion and intonation still lag behind human talent.

• Do use AI to edit or personalise existing human recordings for short-form or preroll.

• Don’t be gimmicky – a famous voice as a chatbot might feel off-brand.

• Do consider creating a bespoke “brand voice” from paid talent, then manipulating it with AI.

• Do use AI for instructional or educational content.

• Don’t let an AI voice run longer than a human equivalent – unless it is your brand voice.

For now, then, AI voice cloning may be more of a supplementary tool than a wholesale replacement. It offers clear efficiencies, creative flexibility and cost savings - but lacks the emotional charge, narrative richness and cultural presence of a real human voice (especially a live one).

Handled well, with fair compensation, creative control and ethical guardrails, it could extend the life - and broaden the reach - of an iconic brand voice. Handled poorly, it risks reducing a rich brand asset to a lifeless digital facsimile.

As Gasston puts it: “When done in the right way, it can be a win for both brand and talent.”

For the likes of Sean Bean or Simon Greenall - and perhaps their estates - that might just be the sound of the future.