“We Must Not Forget That AI Is Only as Good as What It’s Seen or Heard Before”

I was listening to the radio a few days ago and a string of ads caught my attention. Each one had a distinct human voice, and each voice gave an incredibly animated or emotive performance. I instantly thought, an AI voice can’t do that. Everything I’ve heard created by AI so far hasn’t come close to what a human can do and to be quite honest it can be spotted a mile off. It’s regimented and lacking expression, uncanny to the human ear. 

 

A sensible response is that ‘AI can’t replicate human voice that well yet’.

There are companies out there making it their business to try and solve the problems of expression and intonation in computer generated speech. The solution it would seem is to no longer be reliant on clunky text to speech, but to utilise speech to speech which essentially allows any voice to be recorded and then overlaid with a computer-generated voice, be it an original character voice or maybe someone else’s actual voice, which then mimics the original delivery.

On the surface this voice manipulation seems harmless. It gets sticky quickly though when we consider that you could potentially mimic any voice to say anything you want it to without notice or warning to any potential listeners. That is morally ambiguous at best, not only in the realms of false representation but also in terms of compensation for whoever’s voice has been cloned. This is an argument being had in the music business. Perhaps there is a future meeting in the middle where voiceover artists create AI clones of themselves and license their voice per project, but I suspect that will be an opportunity that will only serve the few, not the many.

Despite the morality or legality of an AI voice, its ultimate success will hinge on whether it can replicate that human quality and whether you can really get it to do what you need it to vocally.

Moving away from voice applications we are starting to see the emergence of other AI audio post tools. On the film side tools are being developed that look to give filmmakers the ability to change an actor’s mouth movements, simplifying and helping perfect the ADR process.

Tools that use AI to help with the likes of restoration and voice and music editing are also being released and at the moment seem to be aimed at the user-generated end of the market. As professionals we’ve had the same or similar tools for years now in the form of plugins. Many of those are built using machine-learning algorithms and used by our expert engineers to fine-tune and finesse the processing to ensure the best possible sound. We will never settle for a blanket processing of audio.

For those fearful that these AI tools and those that are undoubtedly still to come pose a threat to our livelihood, we can take heart that the kind of clients that value creativity will continue to want to work with professionals - human ones - who can talk about an idea, iterate on it, and bring to it that human element and experience. We might use tools to make the process more efficient (as we already do!) but the idea and its execution will remain firmly human.

 

We must not forget that AI is only as good as what it’s seen or heard before. Computers can, I’m sure, be trained to recognise pictures, find similar pictures in a database and then apply the same sound - basic SFX tracklaying. Perhaps you could then hit the ‘Mix Like Clearcut Sound’ button. The important point here is that the computer is imitating, it is not creating.

If we do get to a point where AI is good enough to ‘create’ then we might want to ask what the ethics of that is before it’s too late. Over the pond, the Writers Guild of America is on strike and one of their demands is that writers are not forced to work with or be replaced by tools like ChatGPT. It’s a big ask but the striking members are trying to future-proof their careers and potentially the quality of our premium film and TV too. It’s the kind of solidarity that should make us all think, do we want to be working with fellow creatives or perfecting the sounds of a machine and imitating others? 

 

Our industry is built around communication, collaboration, and teamwork. To throw that all away for ease, or the click of a button, just wouldn’t sit right with me. Is that the kind of future that we would want to be a part of? Advertising is emotional, the purchases people make are often due to an emotional connection. Let’s consider what kind of emotions or connections a machine with no experiences can make.