By Bailey Brown | Minneapolis Voice Actor
I was driving recently, half-listening to the radio, when an ad came on that stopped me mid-thought.
Not because it was good. Because it was jarring.
It was a political ad — which already puts it in its own category, production-wise. But this one had something I don’t usually hear in produced audio: a stammer. A mispronounced name. The kind of rough edges that, in any other context, would have triggered a retake.
Honestly? I didn’t love it. The imperfections yanked me out of my half-listening haze — but once I was focused on them, I stopped listening to what the ad was actually saying. The message got lost. The quirks became the whole experience.
And yet I don’t think they were accidents.
I think they were the point. Just not perfectly executed — and that distinction matters a lot.
Welcome to the Uncanny Valley of Audio
You may have heard of the Uncanny Valley — it’s a concept from robotics that describes the unsettling feeling you get when something looks almost human but not quite. A CGI face that’s technically perfect but somehow deeply wrong. A robot that moves like a person but makes your skin crawl. The closer something gets to human without arriving there, the more disturbing it becomes.
Audio is now doing something similar — except in reverse.
As AI voice gets closer to perfect, perfect is starting to feel suspicious. Listeners can’t always articulate why something sounds off, but the gut response is there: this is too smooth. Too even. Too consistent. Nobody actually talks like this.
So what’s the creative response? Reintroduce the imperfection. Put the stammer back in. Leave the mispronounced name. Let the breath land in the wrong place. Not as a mistake — as a signal.
This was made by a human. Trust it.
Political ads have always known this
Here’s what makes the political ad detail so interesting: political advertising has been playing with authenticity signaling longer than almost any other genre.
The candidate who stumbles over a word. The testimonial from a “real voter” who’s clearly not a trained spokesperson. The handheld camera footage that looks deliberately unpolished next to the glossy opposition ad. These aren’t production failures — they’re calculated choices designed to make something feel less produced, because less produced has long been a shorthand for more trustworthy.
What’s new is that this calculus is no longer deciding what level of polish resonates with what audience. It’s deciding whether or not “polish” equals human. And is the humanity of an ad now the main differentiator?
When everything can be faked, polish is no longer proof of perfecting a craft. Sometimes it’s proof of the opposite.
What this means if you’re in the business of sounding human
I’ve been a voice actor for almost a decade. In that time, the standard I was trained to meet was clean, clear, minimal breaths – if any. Sure, you might set yourself up for a “real” or “conversational” read with an improvised lead in that used unscripted cues, like stammers, sighs, grunts, etc.
But now it seems like some of those things are being asked for in the final read.
This is where the ad I heard got it wrong, I think. There’s a crucial difference between imperfection that serves the work and imperfection that hijacks it. When the stammer or the mispronounced name is all a listener can think about, the message is gone. You’ve proven you’re human and lost the room at the same time.
The goal isn’t to beat someone over the head with your humanity. It’s to let it breathe through the work in a way that feels natural rather than performed. A slight variation in pacing. A breath that happens because a real person needed air. A hesitation before a word that gives it weight. These things register subconsciously — they build trust without announcing themselves.
That distinction — between imperfection that feels earned and imperfection that feels forced — is an extraordinarily specific skill. It’s the difference between a read that makes a listener lean in and one that makes them wince. And it’s one that only a human performer who genuinely understands why the imperfection is there can execute well.
AI can produce perfect audio. It’s starting to produce imperfect audio too — the parameters can be adjusted, the smoothing dialed back. But what it can’t produce is imperfection that feels intentional without feeling calculated. The texture of a real person who is actually thinking, actually present, actually letting something land. That’s still ours.
The bigger shift happening in branding
This isn’t just a voice over story. It’s part of something broader.
Rough edges are having a moment across creative industries. Lo-fi aesthetics in music production. Deliberately unpolished social content outperforming glossy campaigns. Handwritten fonts over clean typography. Candid photography over styled shoots. The visual and sonic language of “real” is being reclaimed everywhere, precisely because so much of what used to signal quality — technical perfection — can now be generated by a machine in seconds.
Authenticity used to be a byproduct of craft. Now it’s a design decision. And the brands that are figuring this out first are the ones asking a different question than “how do we make this sound professional?” They’re asking “how do we make this sound like a human made it, on purpose, because that’s what we wanted?”
That’s a more interesting question. And it opens up a more interesting conversation about what voice over performance actually is.
What I’m watching for
I don’t think this is a passing trend. I think we’re at the beginning of a genuine recalibration in how produced audio is supposed to sound — and what signals quality, trust, and humanity to a listener who is increasingly, even subconsciously, on guard for the synthetic.
But the radio ad I heard is also a cautionary tale. Someone made a deliberate choice to dial up the “human” — and it worked as a pattern interrupt. It got my attention. What it didn’t do was hold it. The imperfection was so prominent that it became the ad, rather than something that made the ad feel more real.
The lesson isn’t “add imperfection.” It’s “understand what imperfection is doing, and don’t let it do more than it’s supposed to.”
The best human reads don’t announce themselves. They just feel right. Warm where they should be warm, urgent where they should be urgent, and human in a way that you notice only in retrospect — after the message has already landed.
That’s the standard worth chasing. Not imperfection for its own sake, but humanity in service of the work.
All those years of being trained to take the human stuff out — it turns out the human stuff was the product the whole time. We just have to be smart about how much of it we let show.
Have you noticed this in ads you’ve heard recently? I’d genuinely love to know — drop a comment or send me a note. This is the conversation I want to be having right now.
Bailey Brown is a Minneapolis-based voice actor specializing in commercial, e-learning, IVR, and corporate work. She records from a professional home studio with Source Connect Standard capability, live playback, and real-time editing. Request a free custom audition here.
Frequently asked questions
Why do some ads sound intentionally imperfect? As AI-generated audio becomes more common and more convincing, perfectly produced audio is increasingly associated with synthetic content. Some advertisers are deliberately reintroducing imperfections — stammers, natural breath patterns, slight variations in pacing — as a signal that what listeners are hearing was made by a real person. It’s a trust signal disguised as a production choice. That said, there’s a real risk of overdoing it: imperfection that’s too prominent stops serving the message and starts hijacking it. The goal is humanity that registers subconsciously, not imperfection that becomes the whole ad.
What is the Uncanny Valley in audio? The Uncanny Valley is a concept originally from robotics describing the discomfort people feel when something is almost human but not quite. In audio, the same principle is emerging: as AI voice gets closer to perfect human speech, listeners are beginning to find perfectly smooth, even audio subtly unsettling — because real humans don’t actually sound like that. The closer AI gets to human, the more human imperfection becomes a mark of authenticity.
Can AI produce intentionally imperfect audio? To some extent, yes — parameters can be adjusted to introduce variation and reduce smoothing. But there’s a meaningful difference between algorithmically generated imperfection and human imperfection that feels earned. A stammer that comes from genuine thought, a breath that happens because a real person needed air, a slight hesitation before a word that gives it weight — these have a texture that AI-generated roughness doesn’t yet replicate convincingly.
What does this mean for voice actors? It opens up a more nuanced conversation about what great voice over performance actually is. The skill isn’t just in delivering a clean, technically perfect read anymore — it’s in knowing when imperfection serves the work, and being able to execute it in a way that feels genuine rather than manufactured. That’s a distinctly human capability, and one that experienced voice actors are increasingly well-positioned to offer.
+ show Comments
- Hide Comments
add a comment