French as a working language: narrating texts full of non-French proper nouns
Institutions, NGOs and international bodies write flawless French, but their texts are packed with people and places that are not French. How a voice built for French handles a list of foreign names, and how you fix what needs fixing, once and for all.
French is a working language of the UN, the African Union, the European Union, and dozens of international organisations and NGOs. Their communications are written in impeccable French, but they talk about the whole world: a report drafted in Geneva cites Nigerian delegates, Indonesian cities, English-named programmes, and institutions whose acronyms are anything but French. Narrating those texts raises a specific question, different from the one a local news article poses: how does a voice built to read French cope with what is not French. Here is the answer, put to the test on a concrete list.
The test: a list of non-French proper nouns
Take a deliberately varied list, the kind you meet in the minutes of an international meeting: Ngozi Okonjo-Iweala, Wojciech Szczęsny, Xi Jinping, Recep Tayyip Erdoğan, Yogyakarta, Nouakchott, Ouagadougou, Reykjavik, UNHCR, USAID, GAVI. This list splits cleanly into two families of problem, and that distinction is what matters.
The first family: names written in the Latin alphabet but whose pronunciation does not follow French rules. Szczęsny, Erdoğan, Reykjavik, Yogyakarta. A French voice reads them by applying its own phonetics, so it pronounces them "the French way". The result is intelligible, a listener understands who or what is meant, but a speaker of the source language will hear an approximation. That is not a bug, it is the expected limit of any engine that reads text without knowing which language each word comes from.
The second family: names that read correctly with no adjustment at all. Nouakchott, Ouagadougou, most francophone African place names: their spelling is compatible with French and they come out clean. Xi Jinping and Ngozi Okonjo-Iweala often pass well too, because their Latin spelling lends itself to an acceptable French reading. So the real tipping point is not "foreign or not", it is "does the Latin spelling betray the pronunciation or not".
Acronyms, a case of their own
UNHCR, USAID and GAVI are not words, they are initialisms, and an initialism follows its own rule: some are spelled out letter by letter, others are said as a word. USAID is often pronounced "you-es-aid" the English way, GAVI is said as a word, UNHCR is spelled out. A French voice left to itself decides by its internal logic, and it does not guess that an acronym is English. This is exactly the subject we cover in acronyms and abbreviations in audio: the machine applies a default rule, and it is up to the editor to correct it when it gets it wrong.
What you fix, and how
The good news, for an organisation narrating a steady stream, is that these names recur. The same countries, the same institutions, the same officials come back from one document to the next. Our targeted correction lets you set, once and for all, the expected pronunciation of a word or an acronym, and the tool then respects it everywhere, across every text. So an organisation sets its recurring vocabulary upstream: the name of its director, the acronyms of its programmes, the capitals it cites most. It is the same move as for any proper noun or place name in news audio, simply applied to a broader international lexicon.
This work is done once and it holds. It does not require re-recording, it does not touch the published text, it acts on how the voice reads a given word. For a communications team producing ten documents a week, it is a few minutes of initial setup that prevent dozens of repeated errors.
What the tool does not do, and why we say so
What our tool does not do is guess, on its own, the source language of every proper noun in order to switch to the right phonetics. It reads in French, with French rules, and corrects where you tell it to. An unadjusted international text will therefore come out correct in substance, Frenchified on the foreign names you have not corrected. For many internal uses, notes, minutes, monitoring, that intelligibility is enough. For public-facing communication where a name's exact pronunciation is a matter of respect, prior setup is not an option, it is the work to be done.
We say it plainly, because it is the line our French text-to-speech page defends: we would rather read one language well and give you the means to correct the rest, than promise a perfect reading of thirty languages that nobody delivers. An honest tool shows you where it stops, and puts the correction tool in your hand.
Give your articles a voice with WeDispatch
This blog is itself voiced by WeDispatch. Curious how it sounds on your content?
Book a demo