Multilingual voice over services: an agency workflow
Fynqcast editorial team5 min read
When a client wants the same video in six languages, the hard part isn't finding a voice for each market. It's making all six versions say the same thing, in the same tone. Multilingual voice over services usually follow one process: one approved master script, translations steered by a glossary, a native-speaker check per language and client sign-off before any audio is made. Here's that workflow, a worked example and how to run it in Fynqcast.
Why does everything start with the master script?
Every change made after translation has to be made six times. So the first round with the client covers the source text only and ends with a written yes. Before translating, check three things:
- Numbers, acronyms and symbols are written the way they're spoken: "twenty percent", not "20%".
- Brand and product names are listed separately. They don't get translated or adapted.
- There's room on the clock. A translation can run longer than the original, so for a 20-second video, write the master for about 16 to 17 seconds.
How do you keep the languages consistent?
With a short glossary, agreed once and used for every version: the terms the client actually uses, the names that stay unchanged, the words that stay in the original language and the register, formal or informal. If one market expects a formal tone and another a casual one, write it down.
The voice matters too. Pick one, then listen to it in every language before you commit.
Example: a 20-second voiceover for an optician chain
The client is invented: Lumen Opticians, a chain that now lets customers reserve eye tests online.
Master script (English, 35 words, about 16 seconds by the Studio estimate):
Squinting at your screen again? At Lumen Opticians, you can book an eye test online in a minute. Pick the store nearest you and the time that suits you. Lumen Opticians. See better, starting today.
Campaign glossary:
| item | rule |
|---|---|
| Lumen Opticians | name, never changed |
| eye test | one agreed term per language, chosen by the client |
| register | informal in every language |
| "See better, starting today" | tagline: adapt it rather than translate it word for word; approved separately |
Tracking sheet:
| version | native reviewer | text approved | audio approved |
|---|---|---|---|
| EN (master) | account lead | yes | yes |
| IT | client's local office | yes | pending |
| DE | freelance translator | in review | no |
| FR, ES, PT | client's local office | pending | no |
The sheet stops the classic mistake: generating audio for a version nobody has signed off yet.
How should revisions and client sign-off work?
- Round 1, the master. The client approves the source text.
- Round 2, the translations. A native speaker checks each version against the glossary.
- Round 3, the audio. You send the file and the subtitles, and the client listens and confirms.
A change in round 3 means regenerating that version, so close the text in round 2. If you use someone else's copy or taglines, you need their permission.
Can you use a voice actor's voice, or the client's?
Not through cloning. On Fynqcast, a cloned voice can only belong to the account holder. That person must be an adult with an identity document issued by an EU member state, pass an identity check and give consent. Recording anyone else's voice isn't allowed, even with their written consent. If the client's founder is meant to speak, they can clone their own voice from their own account and choose, when consenting, the languages it may speak. Anything made with a cloned voice opens with a short spoken notice: "Voice cloned with AI." For everything else, use the stock voices.
How to do it with Fynqcast
- In your dashboard, open the Studio tab (part of the FYNQ Studio and Publisher plans) and pick the Voiceover category.
- Under How to start, choose Translate and read, paste the master and set the Original language and the Language to read.
- Open Translation settings and add the glossary: Mandatory glossary, Proper names to keep unchanged, Words not to translate, Register and Notes for the translation.
- Click Translate. The translation comes back to you for review: have the native speaker check it, edit it and approve it.
- Choose the voice, play the sample and generate. You download an MP3 and SRT subtitles.
- Repeat from step 2 for each language. Audio is available in eight: English, Italian, French, German, Spanish, Portuguese, Chinese and Arabic.
When the video goes out, say that the voice was generated with AI: it's honest with the audience and in line with the EU's AI rules. The plans that include Studio and voice cloning are listed on the pricing page.
Read also:
Also available in: Deutsch, Español, Français, Italiano, Português