Voice over and artificial intelligence: where are we heading?

Voice over and artificial intelligence: it's THE topic of the moment. With the advent of ChatGPT, MidJourney and others, it's Fear in the City, Deep Impact, Armageddon, Titanic, Don't Look Up, Giant Meatballs, in short, you get the picture. So, is AI the big bad wolf or a damp squib?

Spoiler alert: no, I don't have a crystal ball. What I do have, however, is the fine profession of French voice actor since 1994, and I'm constantly keeping abreast of the various trends, technological advances and other aspects of our profession. Both in France and abroad, working in French (of course!) and also recording voice overs in English with French accent. I discuss this periodically with my sound engineer pals working in commercial studios, with my agents, and of course with my voice over colleagues, whether they work in a commercial studio, from their private studio or a mix of the two.

We're reading a lot at the moment on the subject of 'voice over and artificial intelligence': that it's going to take our jobs, that it's really too unfair (more on the subject of unfairness in the profession in the article Calimero Syndrome), that robots are evil. So yes, in the long term, voice-overs for projects that don't need humanity, such as internal e-learning modules, will be done by computer-generated voices - and this has already been going on for a while.

This knee-jerk reaction - and a natural one at that - is very reminiscent of the one that appeared when the internet first came into being (and the appearance of private studios that accompanied it) among some of the old guard of voice over talent: it had to be prevented at all costs, it was going to kill the profession, blah blah blah. What these naysayers refused to see was that the internet had also created a new, gigantic source of work, and had also given access to clients from all over the world. I'm talking about the thorny issue of remote recordings and private studios here.

It's true that the web has opened up the profession to many people who previously had no access to it, and many of these people, lacking the necessary talent and skills, have found nothing else but to sell themselves at rock-bottom prices. What we often forget is that this mainly concerned jobs that didn't exist before the internet: e-learning, motion design, web spots etc. What's also often forgotten is that even before the advent of the web, some voiceover actors weren't necessarily very good, but as voiceover was their preserve, producers didn't really have much choice...

Obviously, it's much more convenient to see ourselves as victims and blame AI, the internet, mean clients who don't want to give us work because they're mean and colleagues who get the jobs because they're mean too, rather than questioning our lack of investment in our profession. No training (because everyone's talking, right? I talk about the good, the bad and the ugly voice over trainings here), an amateur website that's 20 years old (when it exists - I talk about the importance of voice over websites here), a non-existent marketing, no effort to be more visible... but no, the amount of work is meagre due to mean people. Oh, and AI of course.

Anything new is frightening, and provokes a Pavlovian reaction of rejection. It's part of the defence mechanisms we mammals are equipped with to protect ourselves from danger. I talk about the 'it was better before' syndrome in the article aptly entitled Voice Over: Was It Really Better Before?
But before we jump off the cliffs of Étretat shouting in an overplayed theatrical voice "we're damned", let's ask ourselves the following question:

Who's interested in computer-generated voices?

Answer: two types of customers.
- The former produce internal educational content that can do without the humanity of in-house e-learning. If you were a producer, you'd be doing the same thing. Why use a Rolls when a Dacia will do?
- The second: those who were already booking these mediocre price breakers, who cast by price, who don't understand that a poor-quality voiceover takes away engagement rather than adding value. And that's the only good thing about AI: it's cheaper. After all, TTS is nothing new. And even if, at the time, computer-generated voices were created by concatenation and the result was pretty rotten, that didn't stop some people attracted by the cheap prices, brother, from substituting them for voice-over actors.

It's a fact that we voice over professionals have already been replaced for some time by these people who don't know how to work, who read aloud while mumbling their lines, content with a pittance as payment. The fear of being replaced by AI replaces the veterans' fear of being replaced by newcomers who are not actors. But in the end, the cream always rises to the surface.

Clients who understand the difference between cost and investment are not deaf: they hear that even with the most recent advances, computer-generated voices are soulless, don't connect with the audience, and do a disservice to their project. They know that a professional voice-over actor brings real added value, which translates into better sales.

Some people, frightened by the progress made in recent times, suggest that AI will soon be capable of generating voices as 'human' as humans. These people have probably never worked in a directed session on TV commercials that cost hundreds of thousands of euros to produce: they would realise that the extremely specific demands of a good producer require a finely tuned ability to listen, an ability to translate 'creative' directions into intentions that connect with the audience.

Saying 'I love you', for example, can mean 'I love you', 'I hate you', 'Pass me the salt', 'Leave me alone', 'Daddy, can you buy me a Playstation', 'See you tonight', 'Could you lend me some cash'? and thousands of other things, depending on the context and the intention responsible for differences in inflection, support, rhythm, projection - in short, extremely subtle nuances that only a human speaker (and not one who reads aloud) can translate into the right emotion for the desired purpose. Another example, which the voice-over veterans among us will have experienced: "Can you make it browner for me, but like a giraffe, with the light at the end of the tunnel?" Go teach that to the AI.

In short, the people who 'read aloud with tone' and respect punctuation like good pupils, without realising that oral transmission works differently from written transmission, will have to leave a profession in which they have never had a place: AI already does it better than they do. But voice-over actors who know how to work, who know and speak to their audience, who put humanity into their interpretation, will be much, much harder to replace. I wrote an article on this subject for humanity in interpretation, which you can find here.

Are we beating ourselves into the ground?

We also need to analyse, as individuals, our relationship to communication. Millennials, for example, are often allergic to phone calls, preferring written text messages (with few phonetically written words and lots of emojis), or voice PMs. It's our attitude that shapes how we communicate - and how we're communicated with. If we are looking for more humanity, the market will move in that direction, using professionals and, for example, specifying 'spot recorded with a human voice' to reinforce the importance the advertiser attributes to humanity.

A lot of voice-overs grumble about AI, but many use the AI filters that rejuvenate, age, turn you into a pirate, an 80s college girl etc. that hang around on social networks, or simply use MidJourney or ChatGPT, or other AI tools. When AI doesn't bother us, we're happy to use it. Ironic, isn't it? In short, where we're going is collectively our choice.

AI, a blessing in disguise?

So, voice over and artificial intelligence... conclusion? Of course, I could be wrong and it is possible that one day the human race will be enslaved by an evil supra-elite who will use AI to achieve their greedy and demonic ends. Or not - there are also promising applications of AI in medicine, surgery... The problem is never the tool but who uses it, and how it is used.

I sing appallingly, and AI can certainly sing better than me. If I compete with AI as a singer, it's obvious that I'll lose. But does AI sing as well as a real singer? No. The same goes for the voice over. AI is a karaoke of real life. Copy/paste. Lift music playing Mozart. Try being Mozart, rather than being afraid of lift music.

Well, on that happy and hopeful note (yes it is), I wish you lots of lovely contracts that will enable you to make a difference! I invite you to share this article with your colleagues if you're feeling generous, and I also invite you to share your views on the subject here and leave your comments, I'll do my best to respond!

PS: by subscribing to this voice over blog, you will have exclusive access to new articles before they are officially published. It's free so make the most of it, go ahead and subscribe now !