série 0
Why this drill?

1–4 to answer · Enter to continue

The path

Ten units. Each adds a single new idea, and each has a short lesson before the drill.

Progress

The words that carry a sentence

Readiness by unit

The gate is 90% accuracy over 30 answers and a median under 3 seconds. Both, because right-but-slow means you are still computing.

Costing you the most

This drill's items

Guide

The whole system, in one box

Silence, read off the last letter.

Ends inSilentExample
a vowelnothingami — a-MEE
-e1petite — puh-TEET
-es2tables — TABL
-ent (verb)3parlent — paRL
c r f lnothingavec — a-VEK
n or mnothing; the vowel nasalisesbon — boñ
any other consonant1petit — puh-TEE

One hole: a verb infinitive or agent noun in -er silences the r — parler, boulanger — while short words keep it: mer, cher.

Liaison. A silent final consonant speaks before a vowel, as /z/ (s, x, z), /t/ (d, t), /n/ (n) or /v/ (f, in two expressions). Blocked by an aspirate h and after et.

Nothing in this app hand-writes a silent count or a liaison. Both are derived, and the test suite asserts the derivation against every word and every pair — so the drill answer and the explanation cannot disagree. Genuine exceptions carry a visible flag; the suite enforces that too.

Reading the respelling

IPA is in every panel. The drill answers use a respelling, because four IPA strings side by side are unreadable to a beginner.

oo = /u/ (ou) · ay = /e/ (é) · eh = /ɛ/ (è) · oh = /o/ (au, eau) · uh = /ə/ · ew = /y/ (u) · eu = /ø/

añ iñ oñ uñ = the four nasal vowels — the ñ marks nasality and is never a separate n sound.

zh = /ʒ/ (j, ge) · sh = /ʃ/ (ch) · R = the uvular r, made at the back of the throat.

CAPITALS mark the stressed syllable. French stress is weak and falls at the end of a phrase rather than being a property of the word — it is a reading aid here, never a rule. This is the opposite of Spanish, where stress is the rule.

An honest verdict on the difficulty

The US Foreign Service Institute puts French in its easiest group for English speakers — roughly 600–750 class hours to professional working proficiency, the same band as Spanish and against about 1,100 for Thai.

Reading French is easier than it looks and harder than Spanish. Spelling-to-sound is regular once you know the groups; sound-to-spelling is genuinely hard, which is why dictation is a school subject in France and why native speakers misspell constantly.

Three things are genuinely difficult, and no app finishes them for you:

  • Listening. Liaison and elision erase the gaps between words, so spoken French arrives as one continuous stream. You can read fluently and still not hear where a word ends.
  • Gender. Every noun has one, it is mostly unpredictable, and it propagates through articles and adjectives.
  • The /y/ and /u/ splittu versus tout, rue versus roue. English has no /y/, and collapsing it changes words rather than just the accent.

What this app cannot do: teach you to speak, to listen to real speech at real speed, or to hold a conversation. It teaches you to look at written French and produce the right sounds. That is a real skill and a small part of the whole.

About the voice

Looking for a French voice…

Audio uses your device's own text-to-speech, switched to Spanish. If the speaker buttons are silent, no French voice is installed:

  • iOS: Settings → Accessibility → Spoken Content → Voices → French
  • Android: Settings → General management → Text-to-speech → install French
  • macOS: System Settings → Accessibility → Spoken Content → System Voice → Manage Voices

Every word also links out to Forvo, where the recordings are human.

Reference tables

Unit 1 — Ce qui ne se prononce pas
Unit 2 — Les voyelles composées
Unit 3 — Les voyelles nasales
Unit 4 — c g s et les accents
Unit 5 — La liaison
Unit 6 — Phrases
Unit 7 — Les nombres
Les plus courants
Unit 9 — Phrases qui se combinent

Glossary

On the frequency tiers

Words are marked very common, everyday or less common, and Révision draws tier 1 about 2.6× as often. Inside a single unit there is deliberately no bias — coverage matters more there.

These tiers are a considered estimate, not a corpus count. Real ranked frequency data would improve the weighting, and that is a known gap rather than a hidden one.

Switch drill

Drill

Unit

Settings

Round length

Progress code

Your progress saves automatically on this device. This code is for carrying it to another one — copy it here, paste it there.

Start over

Erases everything saved on this device. Copy your code first if you might want it back.