Production orale
The speaking axis: produce French out loud and get measured pronunciation feedback — the sounds, joins and rhythm that separate accented-but-correct from almost-native, on the speech runtime.
Everyone understands you. That sentence is the reason this domain is hard to start: when comprehension never fails, nothing in daily life tells you that you have been saying tu and tout with the same vowel for years, or that your liaisons land in places a native ear notices and politely ignores. Immersion rewards being understood, and being understood is a much lower bar than sounding almost-native. Without a measurement, there is no signal at all.
This domain supplies the measurement. Speech runs on the platform's speech runtime: you record, and the sounds come back scored against a native model rather than assessed by how you felt about them — the front-rounded vowels, the four nasals, the schwa, each kind of liaison and élision, then the group rhythm and melody that make a phrase sound French even when every phoneme is right. On top of the sounds sits the discourse: a monologue suivi you record and defend, graded on the same B2 rubric the examiner uses — task, coherence, range, accuracy, register — plus the written preparation that decides what you will say before the recording starts.