AI dubbing against a voice actor: what the money buys

The money question has an obvious answer, so let's get it out of the way and spend the rest of the page on the question that's actually hard.

Professional voiceover is priced per finished minute, and published rate guides put narration work in the low hundreds of dollars per finished minute. Treat any single figure you see with suspicion, including that one: the guides disagree with each other, and real rates swing on usage rights, market, union status, revision terms and how long the licence runs. Check a current rate guide directly rather than trusting a number in a blog post, this one included.

Our dubbing is EUR 1.67 to EUR 2.00 per minute depending on package.

That's roughly two orders of magnitude. Which means cost isn't a comparison, it's a different category of purchase, and anyone framing this as "AI is 40% cheaper" is selling you something.

The useful question is what the money was buying, and whether you need it.

What a voice actor gives you that this doesn't

Five things, and none of them sentimental.

Direction. You can say "that read is too cheerful for a bereavement policy" and get a different take. There is no equivalent instruction anywhere in an automated pipeline. Ours accepts a file and a target language, and that's the entire interface.

Interpretation. A person reads meaning, not words. They know which clause carries the weight in a sentence, when a pause does more than an adverb, and when the script is wrong and should be changed.

Revisions. A line that came out badly can be re-recorded. With us it can't: dubbing runs automatically end to end and nothing is edited by hand afterwards, which our support page states plainly. A single bad line stays bad or you re-run the whole file.

Accountability. Someone is responsible for the read. For content with legal or regulatory exposure, having a human in the chain is sometimes the requirement rather than a preference.

Native judgement. A voice actor who lives in the target market notices when a translated phrase is technically correct and culturally wrong. Nothing in an automated pipeline is going to tell you your slogan is unfortunate in Brazilian Portuguese.

When a human is still the right call

Conditions, not a ranking.

Anything brand-critical and short. A 60 second ad that will run for a year and be heard a hundred thousand times. Per-minute cost is irrelevant at that length, and the downside of a mediocre read is enormous.

Anything with legal exposure. Regulated financial content, medical information, safety instructions. The cost of a subtly wrong emphasis is not measured in euros.

Anything heard repeatedly. A course introduction played at the start of twelve modules. Small irritations grate on the twelfth hearing in a way they never do on the first.

When your source recording is difficult. Overlapping speakers, heavy music, poor recording. An automated pipeline degrades on bad input, while a human just re-performs the script. If your source is a mess and re-recording isn't possible, that's an argument for a human rather than for better prompts.

Character work. Anything performed rather than narrated.

When the comparison is fictional

Here's the part the voice acting industry and the AI industry both tend to skip: in a lot of cases the alternative to AI dubbing is not a voice actor. It's nothing.

Forty hours of course content into six languages, at professional narration rates, is a project with a budget approval, a timeline in months, and a business case. Most organisations with forty hours of course content never localize it, not because they compared options, but because the number ended the conversation before it started.

Same for a back catalogue of 300 videos. Same for internal communications. Same for a solo creator with an audience in three countries.

In those cases the choice isn't between a good dub and a great one. It's between a decent dub and monolingual content, and the honest framing is that AI dubbing made a project exist that otherwise wouldn't have.

The costs that aren't in either rate

Two things that shift the comparison once you look past the per-minute figure.

Casting and coordination. A voice actor engagement has overhead the rate doesn't show: finding them, auditioning, briefing, scheduling, reviewing, and doing all of it again per language. Six languages means six of those relationships, or an agency margin to avoid them. That coordination is often what stalls multi-language projects, more than the invoice does.

Re-runs. Edit your source and both routes cost again. But re-running a dub is minutes and a few credits, while re-booking six voice actors is a project. If your content changes often, that asymmetry outweighs the initial rate.

On our side the arithmetic is exact, and you want it before comparing anything: minutes rounded up per file with a one credit floor, one job per language, billing settled against measured audio, and credits that don't expire. What AI dubbing actually costs works through the rules that move the number.

A reasonable hybrid

The framing as an either/or is mostly a marketing artefact. Plenty of sensible operations split it by content tier.

Human voice for the top of the funnel: the ad, the homepage video, the flagship course introduction. Automated dubbing for the depth: the 40 modules behind the introduction, the support library, the back catalogue, the internal material.

That puts a human where their judgement shows, and gives you coverage everywhere else, which beats either extreme at the same budget.

For planning the automated side, dubbing marketing video from English to German and from English to Spanish are the common starting points, and the use case index covers the 32 languages we support.

FAQ

How much does a voice actor cost per minute?

Published rate guides place professional narration in the low hundreds of dollars per finished minute, but the guides disagree and actual rates depend heavily on usage rights, market, union status and revision terms. Check a current rate guide directly rather than relying on a quoted figure.

Is AI dubbing cheaper than a voice actor?

By roughly two orders of magnitude, which makes the price comparison largely beside the point. The useful question is whether you need direction, interpretation, revisions and accountability, because those are what the higher price buys.

When should I use a human voice actor instead?

Short brand-critical content, anything with legal or regulatory exposure, anything a listener hears repeatedly, character performance, and cases where your source recording is too poor for automated processing to work from.

Can I get a line re-recorded if the AI dub is wrong?

No. Dubbing runs automatically end to end with no manual editing stage, so an individual line can't be fixed on request. The options are re-running the whole file or accepting it, which is the flexibility a voice actor's fee is paying for.