TTS Verbalization Guide & Specifications for fr-FR | Guides

Product: fr-FR
Release Date: October 2026
TTS Engine version: 3.4.0

Introduction

This document describes how specific series of strings are read by the TTS Engine for the FR language in the FR domain. Example outputs are in French (French as spoken in France). The readings are the engine's current output for the version above: when a reading is not what is wanted, this document still shows what the engine does.

Convention: every example table includes a Lexis voice output column. Any new row appears marked with 🆕. Missing support is shown as "unsupported". Numeric ranges use [a-b] (for example, [1-99] means any number from 1 to 99).

Readings are shown in lower case with punctuation as the engine returns it and with hyphens replaced by spaces. Two markers appear in the readings: [pause] is a 0.025 s pause the Lexis voice puts between groups of digits, and [pause 0.2 s] is a fixed pause of 0.2 s.

Numbers

Integers

Format	Example input	Lexis
`[1-9][0-9]`	`25`	vingt cinq
`[2-6]1`	`21`	vingt et un
`71`	`71`	soixante et onze
`80`	`80`	quatre vingts
`[7-9][0-9]`	`94`	quatre vingt quatorze
`[1-9][0-9]…[0-9]`	`159`	cent cinquante neuf
`100`	`100`	cent
`[2-9]00`	`200`	deux cents
`[2-9]0[1-9]`	`201`	deux cent un
`<dash>[1-9]…[0-9]`	`-58`	moins cinquante huit
`[1-9][0-9]…[0-9][0-9] (>999)`	`2824`	deux mille huit cent vingt quatre
`[1-9][0-9]{3}`	`1500`	mille cinq cents
`[1-9][0-9]{0,2}<dot>[0-9]{3}`	`2.824`	deux mille huit cent vingt quatre
`[1-9][0-9]{0,2}<space>[0-9]{3}`	`2 824`	deux huit cent vingt quatre
`1000000`	`1000000`	un million
`2000000`	`2000000`	deux millions
`1000000000`	`1000000000`	un milliard
`1<dot>000<dot>000`	`1.000.000`	un million
`1<space>000<space>000`	`1 000 000`	un zéro zéro

Allowed thousands delimiter: dot only; a space reads as separate number groups (the groups of "1 000 000" are read one by one). Max supported value: 999999999999, read as "neuf cent quatre vingt dix neuf milliards neuf cent quatre vingt dix neuf millions neuf cent quatre vingt dix neuf mille neuf cent quatre vingt dix neuf". Beyond that, digits are spelled one by one.

Ordinal numbers

Format	Example input	Lexis
`[1-9]…[0-9]{er}`	`1er`	premier
`[1-9]…[0-9]{re,ère}`	`1re`	première
`[1-9]…[0-9]{e,ème}`	`2e`	deuxième
`[1-9]…[0-9]{e,ème}`	`21e`	vingt et unième

Decimals

Decimal separator: comma. Three or more digits after the comma are spelled individually ("2,4527"); a two-digit fraction is read as a number ("2,45" is "deux virgule quarante cinq", "-25,78" is "moins vingt cinq virgule soixante dix huit"); a leading zero in the fraction is dropped ("2,05" is "deux virgule cinq").

Format	Example input	Lexis
`[1-9]…[0-9]<comma>[0-9]…[0-9]`	`2,4527`	deux virgule quatre cinq deux sept
`<dash>[1-9]…[0-9]<comma>[0-9]…[0-9]`	`-25,78`	moins vingt cinq virgule soixante dix huit

Fractions and units

Format	Example input	Lexis
`[1-99]<slash>[1-99]`	`13/16`	treize seizièmes
`[1-99]<slash>[1-99]`	`45/2`	quarante cinq demis
`[1-99]<slash>[1-99]`	`37/4`	trente sept quatrièmes

A fraction that complies with the date format, thus [1-31]<slash>[1-12], is read as a date if no supported unit is provided. Numeric dates are month first when both orders are valid, so "5/12" is read as "le douze mai" while "5/12 km" is read as "cinq douzièmes kilomètre". Fractions such as 3/4 and 1/2 without a unit are therefore read as dates ("le quatre mars", "le deux janvier"). A unit that is not in the unit table does not turn it into a fraction ("3/4 mi" is "le quatre mars mi").

Format	Example input	Lexis
`{vulgar fraction} <unit>`	`½ tasse`	½ tasse
`[1-99]/[1-99] <unit>`	`3/4 mi`	le quatre mars mi
`[1-31]<slash>[1-12] <unit>`	`5/12 km`	cinq douzièmes kilomètre

A vulgar fraction such as ½ is not converted: the text passes through unchanged.

Digit-by-digit spelling

The dash notation can be used to spell the digit sequence digit by digit. For Lexis, the digits are read in groups of three, with a short pause ([pause], 0.025 s) between the groups. The comma is placed where the input has a space.

Example input	Lexis
`4-2-1-4 2-3-4 7-5-3`	quatre deux un [pause] quatre, deux trois [pause] quatre, sept cinq [pause] trois

Note: this notation can be used for digit sequences with more than 2 digits. Otherwise it will be verbalized as an interval. E.g. "2-3" is "deux à trois".

16-digit card numbers

Example input	Lexis
`5-8-9-6 4-2-6-5 1-2-3-4 7-4-6-8`	cinq huit neuf [pause] six, quatre deux [pause] six cinq, un [pause] deux trois quatre, [pause] sept quatre six [pause] huit
`5-8-9-6-4-2-6-5 1-2-3-4-7-4-6-8`	cinq huit neuf [pause] six quatre deux [pause] six cinq, un [pause] deux trois quatre [pause] sept quatre six [pause] huit

Counter examples

Numbers joined by a dash more than once are read as integers and each dash is pronounced "tiret" ("10-20-30" is "dix tiret vingt tiret trente"; "4532-1234-5678" is "quatre mille cinq cent trente deux tiret mille deux cent trente quatre tiret cinq mille six cent soixante dix huit").
If only one dash is used, then the digit sequence is handled as an interval.
If no dash is used, the numbers are read as integers.

Example input	Lexis
`4532-1234-5678-9012`	quatre mille cinq cent trente deux tiret mille deux cent trente quatre tiret cinq mille six cent soixante dix huit tiret neuf mille douze
`4532 1234 5678 9012`	quatre mille cinq cent trente deux mille deux cent trente quatre cinq mille six cent soixante dix huit neuf mille douze
`4532-1234-5678`	quatre mille cinq cent trente deux tiret mille deux cent trente quatre tiret cinq mille six cent soixante dix huit
`4532-1234`	quatre mille cinq cent trente deux à mille deux cent trente quatre

Telephone numbers

Format	Example input	Lexis
`[0-9]{2} [0-9]{2} [0-9]{2} [0-9]{2} [0-9]{2}`	`01 23 45 67 89`	un vingt trois quarante cinq soixante sept quatre vingt neuf
`+[0-9]{2} [1-9] [0-9]{2} [0-9]{2} [0-9]{2} [0-9]{2}`	`+33 1 23 45 67 89`	`+33 un vingt trois quarante cinq soixante sept quatre vingt neuf`
`[0-9]{2} [0-9]{2} [0-9]{2} [0-9]{2} [0-9]{2}`	`06 12 34 56 78`	six douze trente quatre cinquante six soixante dix huit

Each pair of digits is read as a number, so a leading zero pair is read as the digit ("01" is "un", "06" is "six"); there are no commas or pauses between the pairs. The "+" is passed on and the country code is read as a number ("+ 33"). Without spaces the digits are one integer ("0123456789" is "cent vingt trois millions quatre cent cinquante six mille sept cent quatre vingt neuf").

Combination of numbers and words

Format	Example input	Lexis
`[0-9]…[0-9]<dash>{word}`	`16-bit`	seize bit
`{word}<dash>[0-9]…[0-9]`	`COVID-19`	covid dix neuf

A dash with spaces between a number and a word is read "tiret" ("16 - bit" is "seize tiret bit"), unlike "16-bit", which is "seize bit".

Ratio and intervals

Format	Example input	Lexis
`[0-9]…[0-9]<colon>[0-9]…[0-9]`	`16:9`	seize à neuf
`[0-9]…[0-9]<dash>[0-9]…[0-9]`	`16-9`	seize à neuf
`[0-9]…[0-9]<dash>[0-9]…[0-9] {word}`	`3-5 jours`	trois à cinq jours

If the ratio values overlap with the time format, the time will be announced. E.g. "24:00" is verbalized as "vingt quatre heures" while "25:00" is "vingt cinq à zéro".

Percentage

Example input	Lexis
`15%`	`quinze pour cent`
`15 %`	`quinze pour cent`
`2,5 %`	`deux virgule cinq pour cent`

The percent sign is read "pour cent" ("15 %" is "quinze pour cent").

Time

Format	Example input	Lexis
`[00-24]<colon>[00-59]<colon>[00-59]`	`12:11:01`	douze heures onze minutes une seconde
`[00-24]<colon>[00-59]`	`23:15`	vingt trois heures et quart
`[00-24]<colon>30`	`23:30`	vingt trois heures et demie
`[00-24]<colon>45`	`23:45`	vingt quatre heures moins le quart
`[00-24]h[00-59]`	`14h30`	14h30
`[0-24] h [00-59]`	`14 h 30`	quatorze heures trente
`[00-24]<colon>00`	`24:00`	vingt quatre heures
`12<colon>00`	`12:00`	midi
`00<colon>00`	`00:00`	minuit
`[1-12] {PM,AM}`	`2 PM`	`deux heures de l’après midi`
`[01-12]<colon>[00-59] {PM,AM}`	`02:20 PM`	`deux heures vingt de l’après midi`

Quarter and half hours are read "et quart", "et demie" and "moins le quart" (with the next hour) in 24-hour times. In 12-hour times the hour is followed by the part of the day: "du matin" for AM, "de l’après midi" for 1 to 5 PM, "du soir" for 6 to 11 PM; 12 PM is "midi" and 12 AM is "minuit". With minutes, the minutes are read as a number ("08:30 PM" is "huit heures trente du soir").

Date

Format	Example input	Lexis
`[2-31] {month} [0-9]{4}`	`14 mai 2022`	le quatorze mai deux mille vingt deux
`1er {month} [0-9]{4}`	`1er juin 2020`	le premier juin deux mille vingt
`le [2-31] {month} [0-9]{4}`	`le 14 mai 2022`	le le quatorze mai deux mille vingt deux
`[00-31]/[00-12]/[0-9][0-9]`	`31/12/19`	le trente et un décembre dix neuf
`[00-31].[00-12].[0-9]{4}`	`27.05.2021`	le vingt sept mai deux mille vingt et un
`[0-9]{4}-[00-12]-[00-31]`	`2022-11-02`	le deux novembre deux mille vingt deux
`[1-31] {month}`	`21 octobre`	le vingt et un octobre
`1er {month}`	`1er mai`	premier mai
`[1-31]/[1-12] (only day-first is valid)`	`23/5`	le vingt trois mai
`[1-12]/[1-31]`	`5/12`	le douze mai
`[1-12]/[1-12]/[0-9]{4}`	`1/1/2020`	le un janvier deux mille vingt
`[00-31]/[00-12] (the date does not exist)`	`30/2`	30/2

Note: "le" is added before the day, also when it is already in the input. The day 1 is read "premier" in spelled-out dates and "un" in numeric dates; other days are cardinal numbers. Numeric slash dates are month first when both orders are valid ("5/12" is the 12th of May); if only one order is valid, that order is used. A two-digit year is read as a number ("dix neuf"); a full year is read as one number ("deux mille vingt deux").

Units

Format	Example input	Lexis
`{amount}{unit}`	`2,5GB`	2, 5gb
`{amount}<space>{unit}`	`1,5 GB`	un virgule cinq gb
`{amount}{unit}/{unit}`	`1500MB/s`	`un, cinq, zéro, [pause] zéro mb barre oblique s`
`{amount}<space>{unit}/{unit}`	`1400 MB/s`	`mille quatre cents mb barre oblique s`

In "(GB, MB/s)", keep the text, since the first sentences still match. Add: "The slash between units is read "barre oblique"."

Supported unit abbreviations

Each reading shows the singular / the plural form: the unit is read after "1" and after "2". The readings come from the French unit table of the input processor tests. A reading equal to the abbreviation means the abbreviation is not in the table: it is passed on in lower case.

Abbreviation	Lexis
`mm`	mm
`cm`	cm
`m`	mètre / mètres
`km`	kilomètre / kilomètres
`Km`	km
`m²`	m²
`m2`	m2
`m³`	m³
`m3`	m3
`"`	"
`in`	in
`'`	'
`ft`	ft
`yd`	yd
`mi`	mi
`mph`	mph
`MPH`	mph
`h`	heure / heures
`hr`	hr
`min`	min
`mins`	mins
`s`	seconde / secondes
`g`	g
`gm`	gm
`kg`	kg
`Kg`	kg
`oz`	oz
`lb`	lb
`lbs`	lbs
`ml`	ml
`mL`	ml
`l`	l
`L`	l
`gal`	gal
`cup`	cup
`Hz`	hz
`hz`	hz
`kHz`	khz
`MHz`	mhz
`GHz`	ghz
`W`	w
`kW`	kw
`MW`	mw
`GW`	gw
`V`	v
`mA`	ma
`A`	a
`dB`	db
`kb`	kb
`mb`	mb
`gb`	gb
`tb`	tb
`pb`	pb
`kB`	kb
`MB`	mb
`GB`	gb
`TB`	tb
`PB`	pb
`ko`	ko
`Mo`	mo
`Go`	go
`To`	to
`Po`	po
`bps`	bps
`Kbps`	kbps
`Mbps`	mbps
`Gbps`	gbps
`Tbps`	tbps
`kBps`	kbps
`MBps`	mbps
`GBps`	gbps
`TBps`	tbps
`PBps`	pbps
`°C`	not a unit: `5°C` reads cinquième c
`°c`	not a unit: `5°c` reads cinquième c
`°F`	not a unit: `5°F` reads cinquième f
`°f`	not a unit: `5°f` reads cinquième f

Note: Short abbreviations that are also words or letters (in, min, h, m, A, s, "Go" versus "go") are read as units only when they follow a number. The degree sign directly after a number is read as an ordinal indicator ("5°C" is "cinquième c").

Words

Abbreviations and spelling

Lexis passes words in capital letters through in lower case and the Lexis model reads them (e.g. ABC, JOHN). Dotted capital abbreviations lose the last dot (e.g. B.C.).

Format	Example input	Lexis
`[A-Z]…[A-Z]`	`ABC`	abc
`[A-Z]…[A-Z]`	`JOHN`	john
`[A-Z]<dot>[A-Z]<dot>`	`B.C.`	b. c
`[A-Z]&[A-Z]`	`R&D`	r et d

"&" is read "et" ("R&D" is "r et d", "chat & chien" is "chat et chien"). A single letter is passed on in lower case ("V" is "v", "W" is "w")..

Common abbreviations

Example input	Lexis
`Prof.`	prof
`Dr`	docteur
`Dre`	dre
`M.`	monsieur
`Mme`	madame
`Mlle`	mademoiselle
`etc.`	etc
`le 5 janv. 2024`	le cinq janvier deux mille vingt quatre
`No. 5`	numéro cinq
`No.`	numéro
`nº 12`	numéro douze
`Nos 3 et 4`	numéros trois et quatre
`nos clients`	nos clients
`par ex.`	par exemple
`cad`	c'est à dire

"Prof." and "Dre" are not expanded ("Pr" is "professeur"). A trailing dot is accepted ("Dr." is "docteur", "Mme." is "madame"). Abbreviated month names ("janv." to "déc.") are read as the full month. "No." and "nº" are read "numéro"; "Nos" is read "numéros" only before a number ("nos clients" is unchanged). "cad" is read "c'est à dire".

Dashed words

Example input	Lexis
`porte-monnaie`	porte monnaie
`e-mail`	e mail
`Paris - Lyon`	paris [pause 0.2 s] lyon
`Paris-Lyon`	paris lyon

URLs and e-mails

URLs

Format	Example input	Lexis
`(www)<dot>{word}<dot>{domain}`	`www.google.com/`	www, point google, point com
`http(s)://{word}<dot>{domain}`	`https://google.com/`	`h t t p s, deux points, double barre oblique google, point com`
`http(s)://www<dot>{word}<dot>{domain}`	`https://www.google.com/`	`h t t p s, deux points, double barre oblique www, point google, point com`
`(http(s)://)(www)<dot>{word}<dot>{domain}<slash>{sequence}`	`https://www.google.com/news=123`	`h t t p s, deux points, double barre oblique www, point google, point com, barre oblique news, signe égal cent vingt trois`
`(http(s)://)(www)<dot>{non-word}<dot>{domain}`	`www.123.com`	www, point cent vingt trois, point com
`(http(s)://)(www)<dot>{word}<dot>{domain}<slash>{word}#{word}`	`https://www.x.com/a#b`	h t t p s, deux points, double barre oblique www, point x, point com, barre oblique a, croisillon b

The scheme is spelled letter by letter, the colon is read "côlon" and the double slash "doublé barre oblique"; "www" is read as written, "." is "point", "/" is "barre oblique", "=" is "signe égal", "?" is "point d'interrogation" and "#" is "croisillon".

E-mails

Format	Example input	Lexis
`{word}@{word}<dot>{domain}`	`info@omilia.com`	i n f o, arobase om ilia, point com
`{word}<dot>{word}@{word}<dot>{domain}`	`jean.dupont@omilia.com`	jean, point dupont, arobase om ilia, point com
`{word}<comma>{word}@{word}<dot>{domain}`	`a,b@x.com`	a, virgule b, arobase x, point com

The part before "@" is spelled letter by letter and "@" is read "arobase". Symbols in the address are read by name: "," is "virgule", ";" is "point virgule", "&" is "et commercial" and "%" is "pour cent".

Currency

Format	Example input	Lexis
`{symbol}{amount}`	`$25,10`	vingt cinq dollars dix cents
`{amount}{symbol}`	`25,10€`	vingt cinq euros dix centimes
`{amount} {symbol}`	`25,10 €`	vingt cinq euros dix centimes
`{amount} {code}`	`25 USD`	vingt cinq dollars américains
`{amount} {quantity} {code}`	`1,3 million USD`	un virgule trois million de dollars américains
`{symbol}{amount} {quantity}`	`$1,3 million`	un virgule trois million de dollars

Note: the subdivision follows the main amount without "et". After "million/milliard" the noun is introduced by "de" with no elision ("deux millions de euros").

Currency codes and symbols

Each reading shows the singular / the plural form. The subdivision is read after the main amount when the amount has decimals ("1,01" and "2,50").

Code	Lexis	Subdivision
`USD`	dollar américain / dollars américains	cent / cents
`EUR`	euro / euros	centime / centimes
`GBP`	livre sterling / livres sterling	penny / pence
`AUD`	dollar australien / dollars australiens	cent / cents
`CAD`	dollar canadien / dollars canadiens	cent / cents
`JPY`	yen japonais / yens japonais	none: `1,50 JPY` reads un virgule cinquante yen japonais

Symbol	Lexis
`$`	dollar / dollars
`€`	euro / euros
`£`	livre / livres
`¥`🆕	yen japonais / yens japonais

Punctuation and character processing

Supported characters

Char	Description	Used as/in	Lexis
`.`	period	sentence end / thousands delimiter	✅
`,`	comma	sentence break / decimal separator; "virgule" in URLs and e-mails	✅
(space)	space	separator of number groups (not a thousands delimiter)	✅
`-`	dash	number group / minus sign / interval	✅
`@`	at sign	e-mail address ("arobase")	✅
`%`	percent	percentage value ("pour cent"); "pour cent" in URLs and e-mails	✅
`&`	ampersand	"et" between words and letters (R&D); "et commercial" in URLs and e-mails	✅
`$ € £ ¥`	currency signs	part of currency	✅
`=`	equal sign	"signe égal" in URLs, passed on elsewhere	✅
`#`	number sign	"croisillon" in URLs	✅
`; :`	semicolon, colon	replaced by a full stop (colon is kept in times, ratios and URLs; ";" is "point virgule" in URLs and e-mails)	✅
`? !`	question mark, exclamation mark	sentence end (see intonation)	✅
`/ ( )`	slash, brackets	time, ratio, URLs, fractions, dates, phone numbers; passed on elsewhere	✅
`° ² ³ ½ "`	degree, superscripts, fraction, quotes	passed on; "°" after a number is an ordinal indicator	✅
`« »`	guillemets	quotations, removed (no spoken output)	✅
`' ’`	apostrophe	elision (l', d', qu'), passed on as `'`	✅
`_`	underscore	removed	✅
`	grave accent	replaced by an apostrophe	✅

Any other character has no reading of its own: the processor passes it on unchanged (for example + * ~ ^ \ | < > [ ] { }).

Punctuation intonation

Char	Lexis effect
`.`	falling intonation
`?`	rising intonation for yes/no questions
`,`	falling intonation
`!`	a lone exclamation mark is rewritten as a full stop (falling intonation)
space / dash	short pause (a spaced dash is a fixed pause of 0.2 s)
no punctuation at the end of a sentence	a period is added automatically