AI is killing voiceover.

Or is it?

An email to George kicks off a much bigger conversation this week about what artificial intelligence is actually doing to voiceover and audio production right now, rather than what everyone thinks it might do sometime in the future.

There's no question some areas have already been hit. IVR, phone prompts, repetitive information reads and other work where consistency matters more than performance are obvious targets. In some cases, AI may simply be a faster and more practical way of doing the job.

But what happens when performance, judgement, direction, taste and creativity matter?

Robbo talks about putting a real commercial script through ElevenLabs and trying repeatedly to coax a usable performance from it. The result was nowhere near the performance that ultimately went to air. The guys also discuss an experiment where someone reportedly spent hours directing an AI voice to achieve something a good human performer could deliver in minutes.

That leads to the bigger question.

Are we actually replacing creative people, or are we mostly automating the work that required very little creative interpretation in the first place?

The conversation moves through advertising, radio, post production and sound design, including the danger of companies judging AI purely by the immediate saving on a spreadsheet. Robbo raises the possibility that reductions in quality may take years to become obvious in audience behaviour, much like cost cutting in radio can take time to show up in ratings.

There's also another side to AI that rarely gets discussed.

Could voice actors eventually license their own synthetic voices?

Could agents represent both the performer and an authorised AI version of that performer?

And if they do, how do you protect the voice once it leaves your control?

That brings the conversation to contracts, watermarking, voice cloning, the story of the original Siri voice, unauthorised voice use, and why performers need to understand exactly what they are signing before handing over the raw material required to recreate their voice.

AI is clearly changing the business.

The interesting question may not be whether humans disappear, but which jobs change, which jobs vanish, and which ones become even more dependent on human judgement, performance and taste.

• Why some voice actors are already feeling the effects of AI

• IVR and repetitive reads as obvious targets for synthetic voices

• Why low cost advertising is adopting AI faster than major creative campaigns

• Robbo's experiment directing an ElevenLabs voice

• Why directing AI can sometimes take longer than directing a performer

• Whether technological change actually destroys industries or reshapes them

• AI generated sound design and the continuing importance of human taste

• AI in radio traffic, weather and news

• Whether audiences should be told when AI is being used

• Short term savings versus long term damage to content quality

• Licensing your own AI voice

• Contracts, ownership and voice cloning

• The story of Siri's original voice

• What happens when a cloned voice escapes into the wild

After finishing Part One with bikes, the team takes a fairly sharp turn into microphone physics.

What makes a microphone directional? How does an interference tube reject sound? Why can the Sennheiser MKH 416 sound fantastic in one booth and thoroughly ordinary in another?

Prompted by tours of the Neumann factory, George, Robert, and AP explore shotgun microphones, cardioid patterns, off axis coloration, acoustic treatment and the physics that manufacturers use to control what a microphone hears.

The discussion includes the Sennheiser MKH 416, MKH 8060, RØDE NTG5, Earthworks Ethos, electret capsules, true condensers and MEMS microphones.

There is even a detour into the oldest captured human voice before everything eventually winds up on Electric Avenue.

What actually makes a microphone directional?

In Part One of this seriously nerdy microphone discussion, the team dives into interference tubes, polar patterns, room reflections and the physics behind shotgun microphones.

Prompted by a tour of Neumann's anechoic chamber, George, Robert, AP and Robbo examine whether microphones labelled hypercardioid are always what they claim to be. They explore how the Sennheiser MKH 416 achieves its familiar focused sound, why that same microphone can sound horrible in the wrong booth, and what manufacturers such as RØDE and Sennheiser have done differently with newer designs.

There is also a surprisingly revealing look at acoustic foam, low frequency room modes, the enormous wavelength of a 20 Hz signal, and how combining microphone capsules creates different polar patterns.

Yes, it gets technical. No, the maths does not completely derail the show. Mostly.

This is Part One of a two part episode.

Thanks to Tri Booth and Austrian Audio for supporting The Pro Audio Suite.

Recorded using Source Connect.
Edited by Andrew Peters.
Mixed by Robbo.

Find out more at The Pro Audio Suite.