Decision Guide · Parity Endeavors
Human or Synthetic Voice for Audio Description: How to Decide
You have a release date, a budget line for Audio Description, and a vendor quote with a synthetic voice in it. It costs less and turns around faster. Someone has to sign off, and it might be you.
These are the five questions I'd want answered before that signature.
TL;DR
A synthetic voice works when the content's job is information, like a training module, a product walkthrough, a short explainer. But when the content tells a story, a human performer is the better choice. In a story, the viewer is supposed to feel the dread before the door opens. The voice has to know what the scene is doing, and a synthetic voice reads the words without knowing context or intention.
1. What Is This Content Trying to Do?
If it’s “Explain our benefits enrollment” that’s one direction. "Make the audience fall for this family before the second act" points the other.
Plenty of content can be in between those two choices, like a documentary, which carries both facts and a story at the same time. When it's mixed, decide by the scenes the audience will remember.
2. Who Heard It?
Technical review checks that the words match the script and the timing fits the pauses. It might not check whether the scene lands.
The failure I hear most often is a synthetic track that passes review yet to the people listening, still sounds off. Some tells: emphasis falling on the wrong word. Or a tense moment gets the same read as a restaurant menu.
Before you sign off, pay blind and low vision viewers from your audience to listen to the same scene performed both ways. Here's one scene described twice with the same words, once by a human performer and once by a synthetic voice: Human AD and Synth Voice Comparison Opens in a new tab..
3. Who Owns the Track?
In March 2026 the Supreme Court declined to hear Thaler v. Perlmutter. That leaves the US rule in place: copyright needs a human author.
A script your writers wrote can be protected. A performance generated by a synthetic voice has no human author behind it. How much human involvement makes AI-assisted work protectable hasn't yet been settled.
(I’m not a lawyer. Ask your counsel what this means for your titles before the track ships.)
4. What Happens Next Season?
Some questions for the vendor:
- Will the same voice be available for season two?
- Can another company use that voice on its own shows?
With a human performer, those answers live in a contract you can negotiate.
5. What Does the Audience Lose If It Goes Wrong?
Synthetic voice scales to languages and titles that didn’t before have Audio Description. That matters to the people watching them. Your decision is about this title and this audience.
Audiences who get the full story come back and tell their friends. Audiences who don't usually leave without telling you, so the loss doesn't show up in an accessibility report.
What I Can't Decide for You
Your budget, your leadership's priorities, and the pressure on your team to ship are real, and they belong in this decision.
About Me
I'm Roy Samuelson, an award-winning Audio Description producer and performer with thousands of credits across Netflix, Disney+, HBO Max, Hulu, and FX. My patent portfolio covers a way to mark, inside the Audio Description track, whether the voice is human or synthetic.
Get a Read on Your File
If you'd rather walk into the sign-off meeting with your own read of the track, send me one file. The Trust the Signal Scan is $250. Within 2 business days you get the timestamps where the voice stops carrying the meaning, what's breaking in plain language, and wording you can use with your leadership. If you go on to Trust Architecture, the $250 counts toward it.
Book the Scan ($250)