The short answer

French is the classic case of reading far better than you speak. If you can work through a menu but lose a waiter completely, the problem is that spoken French drops, links and nasalises in ways the spelling never shows, and a tutor that hears your audio rather than a transcript can hear when you have put an English r in the middle of Paris. If you have a trip booked, Travel Mode rehearses you, sets one real task a day in the city you are in, and debriefs you afterwards.

Almost everybody who studied French at school can read it. Very few can hold a conversation, and the reason is unusually specific: written French and spoken French have drifted a long way apart, and school teaches the written one.

The problem is not vocabulary

You know more French than you can use. The difficulty is that spoken French deletes and links so aggressively that the words you know go past unrecognised. Je ne sais pas is four words on paper and roughly one syllable in a café. Il y a becomes ya. The e in petit vanishes.

So comprehension collapses first, and production follows, because it is hard to produce a rhythm you cannot hear.

Why a tutor that hears you helps

Your audio goes to the model directly rather than being transcribed into text first. Pronunciation, hesitation and pace are part of what the tutor is responding to.

That matters here because the errors that mark you survive transcription perfectly. Say tout when you meant tu and the transcript quietly fixes it. Put an English r in rouge and the text is unaffected. Drop an obligatory liaison and nothing changes on the page at all. The tutor sees a correct sentence, says nothing, and you keep the habit.

What a session is

Unscripted conversation, in real time, with something you can interrupt. Behind it sits an authored CEFR syllabus of 345 items from A1.1 to C1, of which 113 are capabilities, meaning concrete things you can do rather than grammar points. Reference cards appear beside the conversation when something is worth seeing written down, and the talking carries on around them.

Afterwards you get the transcript and a short reflection, and what you keep missing shapes the next session without you configuring anything.

Going to France

Travel Mode changes the shape of the app for a trip.

Rehearsal before you go, aimed at the boulangerie, the ticket machine and the first time somebody answers you at full speed. Then one real-world mission a day while you are there, set in the city you are in and pitched at or below your level. You rehearse it, then you close the app and go and do it.

Afterwards you tell your tutor what happened, and that debrief is the part that behaves most like a teacher. France rewards it, because the gap between the French you were taught and the French spoken at a counter in Lyon is wide, and because getting one exchange to make sense afterwards is worth more than an hour of drills.

The five French tutors

Every language has five, each from a different place and each sounding like it. You can switch whenever you like, and the app remembers where you are regardless of who you talk to.

CamilleParis
A standard Parisian accent, crisp, contemporary and clearly enunciated. The one most course audio is modelled on.
ÉtienneLyon
Standard French with a calm, even, slightly rounded delivery.
MargauxMarseille
A sunny Marseille accent, with the bright intonation and warmth of southern Provençal French.
ThibaultBordeaux
Clear, refined standard French with a soft southwestern warmth and a relaxed, unhurried tempo.
RomainStrasbourg
Crisp Alsatian French with precise consonants and a faintly clipped, melodic lilt.

What English speakers usually find hard

These are drawn from the pronunciation patterns the tutor listens for in French. Because it hears your voice rather than a transcript of it, they are things it can actually notice and work on with you.

u and ou are different words apart: The u in tu has no English equivalent: lips rounded as for oo, tongue forward as for ee. English speakers substitute ou, so tu becomes tout. It is the single most diagnostic marker of an English accent in French, and it is fixable in an afternoon once somebody points at it.
Three nasal vowels, and no n at the end of any of them: The an of dans, the on of bon and the in of vin are single nasalised vowels with no n or m sound following. English speakers reach for the consonant every time. Bon and banc differ only by lip rounding.
The r lives at the back of the throat: French r is uvular, made at the back, not with the tongue tip. Rouge has friction; the r between vowels in Paris is softer. Substituting the English r is the biggest single giveaway there is.
Most final consonants are silent, and then suddenly are not: Chat is sha, petit is puh-tee, beaucoup is boh-koo. But liaison makes a silent final consonant reappear before a vowel: les amis is lez-amis, vous avez is vouz-avez. Missing an obligatory liaison sounds wrong, and adding a forbidden one sounds worse.
Stress is on the last syllable, every time: French has no strong-weak alternation. Every syllable is equally clear and the last one in a rhythmic group is slightly longer. English speakers import English stress patterns and reduce everything else to a schwa, which is the third-biggest accent tell after the r and the u.