use case
verified July 2026
Every variant of the script costs the same as the first.
An explainer is never one script. It is a hook tested three ways, a call to action tested five, and a re-record after the product name changes on Thursday. Per-character pricing taxes exactly the behaviour that makes the video work, because the meter counts every variant you were right to try.
01
Iteration is the point, not the overage
A marketing team rarely knows the winning read in advance, so it renders the alternatives and lets the numbers decide. On an unmetered seat the tenth version costs the same as the first, which means the shape of the work is set by the campaign rather than by a character budget.
First audio arrives in about 107 ms and streams gap-free, so a variant is something you hear back in the moment, not a job you queue. The practical effect is that a copy change is a small edit and a re-render, not a conversation about whether it is worth the spend.
02
The controls a script test needs
- An instant voice clone keeps one narrator across every cut, so an A/B pair differs by wording alone and the test stays clean.
- temperature sets the melodic range, so an energetic hook and a measured explanation come from the same voice without re-casting it.
- A pronunciation dictionary fixes the product name and the feature terms once, so every variant says them the same way.
- 23 languages run from the same script, so a localised cut is another render rather than another vendor and another invoice.
03
Notes — an engineer's checklist
01Can I render a dozen script variants without it changing the bill?
Yes. The seat is flat at $150 a month and nothing is metered per character or minute, so the variant count is a creative decision rather than a line item.
02How do I keep the same narrator across every A/B version?
The voice is a request parameter conditioned on one reference clone, so every variant is spoken by the same narrator and the only difference a viewer hears is the wording you are testing.
03Can I localise an explainer without re-recording it?
The same script runs across 23 languages on the same seat, so a localised cut is another render rather than another booking, and the cost does not move with the number of languages.
See also — related sheets
use case
$0
A narrator who never gets tired of take nine
Script-to-voiceover on a flat seat: re-render the whole video after an edit, in your own cloned voice, without a character count deciding how long the script runs.
who it’s for
1 unit
Gandr for agencies and integrators
Quote clients a fixed voice cost that stays quoted: seats × $150 makes margins knowable, and each client fleet is its own auditable key.
capability
0
Voice cloning from ten seconds, in the request
Zero-shot voice cloning: a ten-second reference rides inside each request, no training job, and the identity holds across 23 languages.
lab report
5,000 min
Per-character pricing, run against flat seats
Four voice workloads run through a $30–$100 per-million-character meter and through flat seats, with the break-even stated and the cases where the meter wins.
A key and one seat to build this on — the same production API this page measures.
Request access for this use case