Let’s get something straight upfront: Anyone who says AI isn’t good enough to replicate most voiceover tasks doesn’t understand, or hasn’t heard, the technology.
It’s pretty flipping good, and can maintain quality longer than it could before, (though still not with enough consistency to be on even footing with good pro voice actors.)
But, it’s a workable technology that many are integrating into their business flow, which terrifies voice over artists.
It shouldn’t.
Let’s trot out the common arguments first: AI can never be BETTER than a voice actor, only AS GOOD, and when it comes to top tier performances it still isn’t even close to being capable of offering the options and range of a great human VO, and falls way behind in trying to do so in the same period of time a well-directed person can.
It requires detailed and precise prompting that is challenging to get right and even more difficult to get right over long stretches, meaning a buyer must spend on a skilled prompter to make the thing work well at all, (when they could cut out the extra step by simply hiring human.)
It is environmentally unsustainable in its current form. This isn’t some bleeding-heart save-the-whales Greta Thunberg statement. It is LITERALLY unsustainable. It takes between one and three GALLONS of water for ChatGPT to generate a SINGLE IMAGE. It consumes unfathomable amounts of energy for simple tasks. The techbros who preach its inevitability better find a few new oceans at this rate, or AI anything will be a very short-lived proposition once people have to choose between hydration, showering, and protecting their homes from fire versus workflow simplification.
To say nothing of the legitimate concerns of many of AI’s very creators that it may, in fact, lead to the end of human civilization.
From our remote little corner of the economy, however, comes solace. AI voiceover, specifically, has two massive flaws, one which may ultimately prove its undoing when it comes to making a business case for its use.
The first flaw, as backed up by recent discussions with leadership in the online voiceover casting site space, is substantial data that now shows that even when presented with quality AI voiceover at marked-down rates right alongside humans at higher rates on the same platform, uptake of AI voices is NEAR ZERO over YEARS of testing, with substantial feedback offered that once the buyer KNOWS the voice is artificial, they no longer want it, no matter how good it may be. This dovetails with much of our own reaction to the technology in everyday life. AI-generated research papers are considered frauds or cheating. You can’t turn in AI homework. And even Google’s new Veo 3 AI video generator, while impressive, still lives firmly in the uncanny valley. I’m looking at my watch and waiting for the first fully-AI hit movie or TV series. Think I’ll be here for awhile.
More importantly for voiceover, however, is that the AI voiceover business model ONLY works financially for the companies offering these products as a software-as-a-service model, generally based on buying credits or subscriptions that increase in price as the buyer wants more functionality or variety. And THIS is the fatal flaw. Have a look through the comments on the social media channels of major AI voice companies. They are a litany of complaints about having to buy more credits, auto-renewing subscriptions, upgrade fees, and gatekeeping. The BUYERS are not happy because the entire model is built on bait-and-switch sales tactics to extract more and more rent from the user. Yet, without this deeply frustrating model, one we can all relate to in other parts of our lives and which makes us hate the corporations who employ it, these firms, (which are almost universally still losing money and have uncertain hopes for profitability,) have zero possibility of satisfying the demands of their investors. And they are already starting to fail.
Ultimately, WE are the easy solution. Not the robots. And more and more buyers are starting to realize this, which is why there has been continued growth in the voiceover industry at large, and far less erosion in certain segments than expected.
AI voiceover is unlikely to simply disappear. It’s a reality. But buyers of human voices are consistently rejecting it, (the buyers of AI voice are turning out to be a parallel industry, not our existing client base,) and even those who gravitate to AI voice first are starting to encounter its fatal flaws.
AI voiceover companies and major corporations want voice actors scared, divided, and willing to make bad deals for personal security. But they are the ones with the clock ticking on their viability. Not us.
Thank you for this article J. Michael. Of all the things written about AI and voiceover, this was the most well informed and calming one I have read. Actual data and research to back up all the points which is so welcomed in this day and age of shock articles.
Bada bing, bada boom, bruh
I second this 👌🏻🙌🏻😂
It’s going to be a long, bumpy road. I recently listened to an audio book, but didn’t notice til after that it was voiced by “a synthesized voice”. While I had some quibbles with the pronunciation, the overall read was fairly solid. I was low-level shocked.
I definitely think a lot of the lower-level work is gone for the foreseeable future. But I also know that no AI can currently do what I can do, and I am proceeding boldly as if I have a reason to keep making myself available. While I’m at it, I’m also working with a couple AIs to expand my offerings (using only my human voice). I look forward to seeing what comes of all this in the next few years.
You brought up solid points, J. Michael. There will always be a place for AI, but it will run its course in Voiceover, and the dangers it poses to civilization will eventually stump its progress. Your research supports this, and I have even heard some of its ethical flaws discussed at AI Conferences in Higher Education. I did not realize that AI uses huge amounts of water to generate a single image until a presenter was asked if he could show the attendees how to use AI to create an image. He flat-out refused. He said his personal ethics will not ever use it due to the environmental impact it has on our water supply. Then, just the other day, my daughter, who earned her BFA in Art and is currently completing her Master’s in Business Marketing, told me about her marketing project where she had to create original images using AI. She was pretty proud of her work until I shared this information. The realization that she contributed to losing water was a shocking and sobering discovery. My point is regular consumers do not know this. When we do, then we are not so enamored with AI.
This awareness reminds me of growing up in NY during a drought. I remember the PSA campaign employed. It was a cartoon of a child brushing his teeth with the water running while a fish was losing its water to survive—until the fish told him to please shut off his faucet. To this day, I will not run the sink water while brushing my teeth. Perhaps we should revive this PSA, but it would be a person using AI to create an image while we see our water sources draining.
Your article brought awareness that goes beyond the concerns of AI taking certain jobs away, thereby only affecting certain people, to the possibility that AI can potentially destroy the resources we all need to survive.
THIS! Is FANTASTIC! So many excellent points I’d not considered, nor been aware of. Thanks for sharing this incredibly informative, enlightening piece!