Web · AI

French Coach

AI speaking practice for IGCSE French, because speaking is what students practise least.

French Coach app homepage showing the headline "Practise the IGCSE French speaking exam by speaking, not by reading"

Overview

French Coach grew out of a classroom observation: students studying French for years could barely speak it under pressure, because speaking time per class averaged about sixty seconds each. The app puts speaking at the centre — record an answer, get feedback on grammar and vocabulary, and build a session history that actually remembers where you struggle.

Why I Built It

Paper 3 is a speaking exam, but classroom time for speaking is tiny. I started this project to give that missing practice a dedicated loop: not a chatbot, and not more reading. The longer account of the motivation is in Why I Started Building French Coach.

How it works

The core loop is deliberately narrow. Role-play and topic-conversation scenarios are modelled on the IGCSE Paper 3 structure. Feedback is generated from the transcript rather than from a chat interface. A per-category mastery tracker surfaces the grammar and vocabulary patterns a teacher would notice across months.

Challenges

Building it has meant working through speech recognition reliability, LLM feedback latency, and the hardest part — making the memory system useful rather than noisy.

The public preview is live, and the repository is open. Speaking features are still rough. This is in-progress work, not a finished coach.