Skip to content

use case

verified July 2026

Ship the launch video in twenty markets at once.

A launch happens once, in every market, on the same day. The audio has always arrived market by market instead, because a meter prices each re-voiced cut by its length, so the smaller locales wait behind the three that pay for themselves. Flat pricing removes the reason they wait.

01

One launch, every market, the same day

The deliverable here is not one voiceover but the same hero video, demo and set of social cuts re-voiced into every target locale, ready together. On concurrent seats those renders run in parallel, so twenty market cuts land on the launch date rather than trickling out as each one clears a per-character line in the budget.

The knock-on effect is on the long tail of assets. A campaign is not a single video; it is the trailer, the cutdowns, the vertical edits and the paid variants, multiplied by every language. Unmetered characters mean the fiftieth fifteen-second snippet costs what the first one did, so full-catalogue localisation becomes a decision about reach rather than about invoice size.

02

What a re-voiced campaign needs

  • One cloned reference carries the presenter across 23 languages, so the hero video keeps the same brand voice in every market instead of a different stranger per locale.
  • Per-request emotion and prosody, so a product demo lands as delivered energy at the pacing the edit wants, not flat narration timed to the words.
  • First audio at about 107 ms with streaming output, so a market cut can be reviewed and re-rendered in seconds while the timeline is still open.
  • Unmetered characters, which is the whole reason the twentieth market and the paid variants get a voice at all rather than shipping as subtitles.

03

Notes — an engineer's checklist

01Do you edit or mux the audio back onto the video?

No. Gandr synthesises the re-voiced audio and carries the identity across languages; the translation, the edit and the mux stay in the pipeline you already run. Send the translated script and a reference clip and you get the localised track back.

02How many markets can I launch at once?

That is a seat question, not a per-market one. Concurrent seats render cuts in parallel, so the count of markets changes how many renders run at the same time and not what the localisation costs.

03Does the presenter's voice hold up in a language it never recorded?

That is the hard case cloning is measured on, with none of the original phonemes to lean on. One cloned reference speaks three languages on the dubbing sheet, and eight of the 23 carry published, dated specimens at /languages/, so you can judge it before committing a campaign to it.

See also — related sheets

A key and one seat to build this on — the same production API this page measures.

Request access for this use case