I was in the booth last Tuesday, performing audio description for a scene where a father watches his daughter leave for college. The script called for a simple transition: “He stands alone in the doorway.” But I caught something in the actor’s face: a micro-expression of pride mixed with loss. So I let my voice carry that weight. Just a slight pause before “alone,” a gentle drop in tone.
Later, I got an email from a blind viewer: “That moment hit me right in the chest. I felt like I was standing there with him.“
That’s the difference. And that’s what we’re losing when someone says, “Let’s just use synthetic voice for audio description, it’s cheaper, faster, scalable.”
And I understand the gap is narrowing; synthetic voice is sounding “real” and picking up context and emotional cues. I still see belonging, trust, and cultural parity as the unmatchable edge.
So every time I hear “it’ll do,” I don’t think about the movie Babe. I think about that email. Because what blind audiences are really hearing is: “You don’t deserve the same artistry as everyone else.“
That’s not accessibility. That’s exclusion with an efficiency sticker slapped on top.
The Booth Reality Check
I’ve spent thousands of hours behind the microphone, and here’s what I’ve learned: audio description isn’t just reading words out loud. It’s conducting an invisible orchestra of emotion, timing, context, emotional nuance, and meaning.
Last month, I heard audio description for a thriller where the killer’s identity gets revealed through a reflection in a window. The synthetic version I heard just said: “Sarah sees the murderer in the glass.” Technically correct. Emotionally flat.
When I perform audio description, I take in context. I let the silence stretch. Pause. Let the dread build.
One version delivers information. The other creates experience.
“The text-to-speech version gave me the facts, but the human version gave me chills,” a blind viewer told me. That distinction matches what Fryer et al. (2018) found: human narration significantly boosted immersion in drama (https://www.researchgate.net/publication/318426203_Vocal_delivery_of_audio_description_by_genre_measuring_users%27_presence).
Is synthetic voice as good as human voice for audio description? Not if you want your audience to feel something they’ll remember.
For example, Rodero (2023) found that human-narrated stories significantly improved recall, engagement, and mental imagery, with less effort than synthetic voices (https://repositori.upf.edu/bitstream/handle/10230/53353/Rodero_new_synt.pdf). Studies in accessibility and media psychology consistently show that human’s natural rhythms, intentions, and pacing directly impacts emotional engagement and memory retention. AI voices, no matter how polished, currently still stumble when it comes to irony, subtext, and the kind of micro-timing that makes moments land.
But there’s something deeper here. A human voice signals investment. It tells listeners: “You mattered enough for us to bring in an artist, not just run your story through software.” And research backs this signal: Fan et al. (2025) found people consistently judged human emotional voices as more authentic and trustworthy than AI-generated ones (https://escholarship.org/uc/item/8vr8s6h8).
What Gillette Taught Me About Culture
Bear with me here, this connects. Jordan Belfort wrote an article about razors for women. https://www.wolfonwealth.com/p/gillette-lied-to-women
Gillette spent fifty years convincing an entire culture that body hair was a problem that needed solving. Through ads, language, and persistent messaging, they literally reshaped social norms.
We’re at a similar crossroads with audio description.
If we keep framing it as “compliance” or “assistive technology,” we’re positioning it as an afterthought. But if we consistently describe audio description for entertainment content as immersive storytelling, as artistry that belongs to everyone, as emotional alignment between creator and audience, we can shift the entire conversation.
I’ve seen this happen in real time. Three years ago, executives would ask me, “Do we really need professional voice talents for this?” Now they’re asking, “How do we make sure our audio description matches the quality of everything else we produce?“
That’s culture change. Slow, but real.
The Cost Question Everyone Asks
(Roy takes a deep breath, and asks people to review https://roysamuelson.com/rates) How much does human narrated audio cost vs synthetic?
While synthetic voice costs pennies and delivers instantly, here’s where the conversation usually goes off the rails when cost becomes the only metric that matters.
Synthetic expands access in markets where no budget exists, and that’s a win. But in high-stakes storytelling, cutting human narration is exclusion. And there’s a measurable cost: Mak & Zhang (2024) found human voices outperformed synthetic on actual learning outcomes in educational settings (https://www.tandfonline.com/doi/full/10.1080/00220973.2024.2446169). Saving on narration can mean losing comprehension, retention, and audience impact.
A hard skeptic could argue: “Audio description isn’t supposed to be art – it’s supposed to be access.”
Yet we are discriminating against the very access that sighted people have when they show up for a movie or a tv show: that immersive experience is removed.
I was talking to a (STREAMING PLATFORM REDACTED) executive last year who put it perfectly: “We spend millions on cinematography, sound design, color grading. Then someone suggests we save a few hundred bucks by flattening the one element blind viewers depend on most? That’s like serving Dom Pérignon in a paper cup.”
Synthetic voice saves you money today. Human narration builds loyalty, credibility, and audience connection over time. Which is actually more expensive to lose?
Making Synthetic Less Awful
Yes, synthetic expands access in markets where AD wouldn’t exist otherwise. Budgets are real. Deadlines exist. If you’re stuck with synthetic voice, at least don’t embarrass yourself. Here are small changes that create big improvements:
Use hybrid narration. Save human voices for the scenes, but credit lists? That is literally reading a phone book, no emotional intent required.
Program in silence. Synthetic voices rush through everything at the same pace. Add pauses. Let grief breathe. Let tension build. Control timing. It can be the difference between robotic and passable (it’ll take much more time than a professional voice actor who’s trained to do this, but… insert shrug emoji).
Some audio description writers aren’t aware of their words pacing when it’s taken off the page. They can use synthetic voice as a way to hear their script out loud first to prevent overwritten cues. And besides, if the description sounds awkward, or rushed, it’ll also sound worse coming from AI. Test every line by speaking it yourself.
Choose expressive text-to-speech options. Use pitch shifts, volume changes, artificial breathing sounds, or take the tedious time to program emotions (this is duct tape, I talk about intention here https://roysamuelson.com/performing-audio-description-that-actually-connects-a-straightforward-guide-to-inclusive-storytelling/
or even here https://roysamuelson.com/beyond-compliance-how-audio-description-builds-loyalty-emotion-and-trust/). Push the technology as far as it’ll go toward sounding human. Audiences will be… somewhat less disappointed.
Test on real devices. Play your final mix through earbuds, car speakers, phone speakers, cheap laptop audio. If it falls apart on any of these, your audience will notice, too.
These aren’t optional nice-to-haves. They’re the baseline for not insulting your audience’s intelligence.
But in high-stakes storytelling, cost-cutting doesn’t just save dollars, it costs you trust, loyalty, and cultural credibility.
How to Actually Choose
Another question I get constantly: How to choose voice for audio description?
Here’s my decision tree:
– If you want your audience to feel something → go human
– If you just need them to know what’s happening → synthetic might work
– If you care about brand reputation → invest in human talent
– If you only care about this quarter’s budget → synthetic, but stop pretending it’s “just as good” (it’ll show)
Your choice isn’t just technical. It’s a statement about who you think deserves artistry.
What We're Really Fighting For
I was at a conference panel last month where another speaker said, “Blind audiences should be grateful for any audio description at all.” The room went quiet. Then a blind audience member stood up:
“I don’t want your gratitude scraps. I want the same storytelling experience you’re selling to everyone else.”
This echoes what Podsiadło & Chahar (2016) found: even though blind users rely on synthetic daily, they still rate human narration as more natural and preferred (https://www.isca-archive.org/interspeech_2016/podsiado16_interspeech.pdf)
Tthis is really about connection versus control.
Synthetic voice often carries an implicit message: “We’ll let you in, but only in a cheaper, flatter way.” It positions blind audiences as a problem to be managed efficiently.
Human voices sends a different signal: “You belong here completely. This artistry is yours too.”
One builds trust. The other slowly erodes it.
The Long View
Here’s what gives me hope. Three years ago, audio description performers weren’t eligible to join the Television Academy, so they couldn’t do things like vote for the primetime Emmys. Working in the performers Peer Group Executive Committee, alongside some advocates and leaders within the organization, now they are https://www.televisionacademy.com/academy/organization/peer-groups/performers
And SAG screeners now require audio description. Major streaming platforms are investing in quality standards instead of just checking compliance boxes. And credits matter https://roysamuelson.com/why-audio-description-teams-deserve-credit-and-why-it-matters-for-your-production/
The culture is shifting, but this is generational work. Every time a producer chooses human narration over synthetic, they’re reinforcing the world we’re trying to build. Every time someone defaults to “it’ll do,” they’re slowing that progress.
This isn’t about today’s project budget. It’s about what becomes normal for the next generation of storytellers and audiences.
Bottom Line
Earlier this year, I had a surprisingly open conversation with a leader in AI video: their client initially wanted synthetic voice. “It’s just a 30-second spot,” they said. But after hearing the non-human version, they admitted: “I didn’t realize how much emotion we were leaving on the table.” The AI video leader chose human voices. And it made all the difference.
That’s the thing about human connection in storytelling. You don’t know what you’re missing until you hear what’s possible.
Synthetic narration will always be faster and cheaper. But human narration creates something synthetic can’t: the feeling that someone cared enough to make this beautiful for you.
And storytelling without that care? That’s not storytelling. That’s just noise.
If you're ready to do this right:
Stop settling for “it’ll do.” Your blind audiences deserve better, your brand deserves better, and frankly, your stories deserve better.
Check out my book “A Voice Actor’s Guide to Audio Description Performance” at roysamuelson.com/book to get some context.
Sign up for my newsletter for more https://roysamuelson.com/contact/
Listen to the The ADNA Presents Podcast (on all major podcast platforms) to hear from blind professionals and more who are changing industry standards from the inside
Don’t just talk about inclusion. Build it. Starting now.
Here're some references to support the research:
1. Rodero (2023) – Human narration boosts recall, engagement, imagery
Synthetic versus human voices in audiobooks: The human emotional intimacy effect.
– Link: https://repositori.upf.edu/bitstream/handle/10230/53353/Rodero_new_synt.pdf
Found that listeners exposed to human voices reported higher enjoyment, narrative engagement, mental imagery, attention, and information recall, and with less cognitive effort, compared to synthetic narration. This is your slam dunk for “human narration makes moments stick.”
2. Fryer et al. (2018) – Human narration increases immersion in drama
Fryer, L., Cavallo, A., & Freeman, J. (2018). Vocal delivery of audio description by genre: measuring users’ presence.
– Relevance: Found that human-narrated AD significantly increased presence/immersion in drama, but differences were smaller in documentary. This is your evidence for the “if you want them to feel → go human” rule, and also your nuance: synthetic may work in purely factual contexts.
3. Fan et al. (2025) – Human voices judged more authentic and trustworthy
Fan, G., Tzirakis, P., & Schuller, B. (2025). Investigating Emotional Prosody, Authenticity, and Trust in AI vs Human Speech.
– Link: https://escholarship.org/uc/item/8vr8s6h8
– Relevance: Shows that people are better at identifying emotions in human speech and consistently judge human emotional voices as more authentic and trustworthy than AI-generated ones. This backs your “a human voice signals care and investment” line.
4. Podsiadło & Chahar (2016, Google) – Blind users critical of synthetic voices
Podsiadło, M., & Chahar, P. (2016). Text-to-Speech for Individuals with Vision Loss: A User Study.
– Link: https://www.isca-archive.org/interspeech_2016/podsiado16_interspeech.pdf
– Relevance: Found that visually impaired users rely heavily on synthetic voices, but rated them less favorably on quality, naturalness, and preference. Important for showing that while synthetic is used daily (efficiency), blind users still value natural voice quality when it’s available.
5. “Cute Computer-Synthesized Voice Hinders Learning” (2024) – Human voices improve learning
– Citation:
Mak, Y. H., & Zhang, L. (2024). Cute computer-synthesized voice hinders learning performance. The Journal of Experimental Education.
– Link: https://www.tandfonline.com/doi/full/10.1080/00220973.2024.2446169
– Relevance: Found that in educational settings, human voices outperformed synthetic voices on learning outcomes and perception of helpfulness. Great support for your “retention and comprehension” claim, and a counter to “AD is just for information, not art.”