Reading percentages and rates aloud in French: « pour cent », the point, the space
3,5 %, +0,25 point, de 2 à 4 %, ‰: a percentage hides a decimal, a unit, a typographic space, and a confusion between a point and a percent. The forms that trip up a French text-to-speech engine, and what proper preparation does about them.
A percentage looks like a number with a symbol tacked on, and that is precisely why it reads badly. It actually bundles a decimal number, a unit that comes after the figure in French, a typographic space many tools ignore, and a distinction almost nobody models: the one between a percentage and a percentage point. « 3,5 % » reads « trois virgule cinq pour cent », not with « pourcent » run together, and certainly not « trois pour cent cinq ». On an economics page, an earnings story or a table of rates, these forms appear in every paragraph, and a wrong reading makes the figure useless, because a rate a listener cannot retain is a rate that was never communicated.
The table of rate forms that fool the machine
| Written form | Expected reading | Common error |
|---|---|---|
| 3,5 % | trois virgule cinq pour cent | « pourcent » run together |
| 0,25 % | zéro virgule vingt-cinq pour cent | percent split from the number |
| 100 % | cent pour cent | « cent pourcent » |
| -2 % | moins deux pour cent | « tiret deux pour cent » |
| +0,5 point | plus zéro virgule cinq point | read as « 0,5 pour cent » |
| de 2 à 4 % | de deux à quatre pour cent | « deux tiret quatre » |
| 20 % de TVA | vingt pour cent de TVA | « T, V, A » left unresolved |
| 4,5 ‰ | quatre virgule cinq pour mille | read as « pour cent » |
| taux à 2,5 % | taux à deux virgule cinq pour cent | « deux point cinq » |
The right-hand column is not a hypothesis. These are the readings you hear from voices that predict sound without explicit rules for French punctuation and units. Each row is a two-minute test you can run against the tool you are evaluating, ours included.
Why a rate resists a naive reading
The first trap is the decimal comma. In French, the fractional part of a percentage reads « virgule » followed by the rest: « 3,5 % » becomes « trois virgule cinq pour cent ». The danger is cutting in the wrong place and saying « trois pour cent cinq », separating the symbol from the number it belongs to. This is the same mechanism described in reading numbers aloud in French: the symbol is voiced only once the whole number is reassembled, never in the middle.
The second trap is the space. French typography puts a non-breaking space between the number and the percent sign (« 3,5 % »), and that space is not a pause: it must not produce silence, nor make the engine treat the figure and the symbol as two separate items. A tool that reads the space as a sentence boundary chops the reading at the worst possible spot.
The third trap, and the costliest one on economic copy, is the confusion between a percent and a percentage point. « A rate moving from 2 % to 2,5 % gains 0,5 point », not « 0,5 % ». The two phrasings mean different things, and an engine that renders « point » as « pour cent » corrupts the meaning while the voice betrays nothing. The ‰ sign, finally, reads « pour mille », not « pour cent »: it belongs to the same family of signs as the mathematical symbols that must be expanded rather than spelled.
What happens before the sound comes out
None of this depends on the timbre. It depends on normalisation that recognises a percentage, reassembles the decimal number, places « pour cent » after the figure, absorbs the typographic space and keeps a point distinct from a percent. That is upstream rule work, exactly the logic behind the French text-to-speech page: a beautiful voice fed raw text says « trois pour cent cinq »; an ordinary voice, properly prepared, says « trois virgule cinq pour cent ».
What WeDispatch does, without overpromising
Percentages with decimals, the percent sign placed after the number, the typographic space absorbed without a stray silence, per-mille expanded: these frequent forms are handled by the normalisation applied to every article, in both reading modes. That covers an earnings report, an economic note or a pricing sheet.
One case no machine can settle on its own: percentage point versus percent. When copy writes « point » where it meant « pour cent », or the reverse, the tool reads what is written, faithfully, and it is for the newsroom to fix the ambiguity at the source. For a phrasing that recurs and must always read the same way, the pronunciation lexicon pins the intended form with nothing to regenerate. We would rather say it that way than promise an infallible reading: the tool applies safe rules to the frequent cases and leaves the questions of meaning to you.
The test to run before you choose
Take a real numeric passage from your site, a results paragraph or a line from a rate table, not a fabricated example. Listen for three things: does a decimal percentage come out as one block, « trois virgule cinq pour cent »? Does the space before the sign create a silence? Is a « point » clearly distinct from a « pour cent »? If you could copy the rate by ear without a repeat, the tool holds. If not, look at the text preparation, not the voice.
Give your articles a voice with WeDispatch
This blog is itself voiced by WeDispatch. Curious how it sounds on your content?
Book a demo