The third take is where a per-character bill stops looking cheap.
Metered pricing looks cheap when you price one clean pass. Audiobook chapters and podcast episodes rarely work that way. Set your finished minutes and the number of drafts you really make, and see what a per-character meter counts against you.
Cost calculator
Audio that ships, not audio you generate. 10 to 600 minutes.
This is the preview tax. You hear a paragraph, notice the emphasis lands in the wrong place, revise a phrase, compare another voice, and listen again. On a metered plan those checks can draw down the same allowance as the take you keep, so you pay for characters you throw away.
There is no universal multiplier. Character dialogue, technical pronunciation, sensitive training, and audio timed to picture usually need more passes than a straightforward explainer. Use your own history rather than a tidy usage example.
Your provider rates
We do not publish other companies' rates. They change, and a stale number inside a calculator is worse than no number. Open the current pricing page of the service you use and enter what it charges. Leave a field blank and its row stays blank.
USD per 1,000 characters, from their current pricing page
USD per 1,000 characters, from their current pricing page
USD per month, flat, from their current pricing page. Check what sits outside the plan, including editing, mastering, and export.
What a meter counts
- Words a month
- 36,000
- Characters in one clean pass
- 198,000
- Characters a meter would count
- 396,000
- Of that, generated and discarded
- 198,000
Pace is held at 150 words a minute, which is how a 10-minute Vois script runs. We estimate 5.5 characters per word, including the space.
| Pricing model | How the bill moves with iteration | Estimated monthly cost |
|---|---|---|
| Credit-based voice platform | Each preview and revised section consumes the available generation | Add a rateYour rate, applied to every billed character |
| Usage-priced API | More text and more requests increase spend | Add a rateYour rate, applied to every billed character |
| Unlimited cloud plan | Volume does not require a per-character forecast | Add a rateThe monthly fee you entered |
| Vois Subscriber | Revision volume does not create per-character charges | $29.00flat, unlimited generation |
| Vois Pro | Revision volume does not create per-character charges | $49.00flat, unlimited generation |
The Vois rows are one flat monthly price with unlimited generation, and they do not move when you drag either slider. See current plans for what each one includes, or download Vois and run a real chapter through it.
The report, before you trade an email for it.
Eight pages, US Letter. The 2026 pricing landscape, per workload cost tables, and the break even arithmetic behind the calculator above.


The 2026 AI voice cost report
The full model behind this calculator, worked examples for an audiobook, a podcast season, and a monthly video schedule, plus a provider audit worksheet you fill in with the rates from your own pricing pages. We do not publish rate tables, because a published price goes stale within a month.
We will email the download link plus three short follow-ups, and Studio Notes twice a week. One click unsubscribes from everything.

Methodology, and where this is soft
A calculator that hides its assumptions is a sales tool. These are ours.
Where the numbers come from
Two of them are ours and two are yours. We hold pace at 150 words a finished minute, which matches the roughly 1,500 words in a 10-minute Vois script, and we estimate 5.5 characters per word including the trailing space. That second figure is an approximation, not a measurement. Real character counts move with language, formatting, and writing style, so count one representative finished script if you want the exact number. Every provider rate is the one you typed in.
Why we do not list competitor prices
Because we would get them wrong. Plans, character allowances, and overage terms change often enough that any table we published would mislead somebody within a month. The honest version is the one above: read the current pricing page of the service you use and enter the rate. The only prices we state are our own, and they come straight from the Vois pricing page.
What the draft multiplier assumes
It applies your draft count to the whole script, which is the pessimistic reading. Plenty of revisions touch one line or one section instead, so treat the metered rows as a ceiling rather than a forecast. Pull the slider back to 1 to see the tidy version that a pricing page assumes, then set it to the number your last project actually needed.
What this model leaves out
Plenty. Metered services differ on what they count, so a failed render, a voice comparison, or an export may or may not draw down your allowance. This page also prices generation only. It ignores your time, the rework a fragmented workflow creates, and any separate editing or mastering tool you still need after the file lands, which are often the larger numbers. Overages usually arrive in your busiest month, so run the calculation on your heaviest month rather than your quietest one.