"DELF ORAL TRAINER" - artificial intelligence · product design · build

THE QUESTION

Can AI make speaking practice feel closer to the actual exam?

I passed DELF B1, but I wasn't satisfied with my oral performance.

The obvious answer was more conversation practice. But lessons with native speakers cost money, schedules do not always align, and there is another, very human barrier: making mistakes in front of someone can make you more cautious precisely when you need to practise speaking freely.

So I decided to build something I wanted to use myself.

WHAT I WANTED

I wasn't trying to build another general French-learning app. I wanted to solve one specific problem: oral practice for an exam.

Three things mattered to me.

• A real conversation, not a speaking exercise.I wanted to respond spontaneously and have the conversation continue based on what I said, rather than simply read prompts and record answers.

• Feedback within an exam context.Correcting every imperfect sentence would defeat the purpose. I wanted feedback on mistakes that mattered at the selected DELF level, along with stronger or more natural expressions I could actually use in the exam.

• Difficulty I could control.Passing B1 did not mean I always wanted B1 practice. I wanted to move down when I needed confidence or repetition, and move up when I wanted a challenge. So the app allows practice across A1 to B2.

WHAT I BUILT

I designed and built an AI-assisted oral trainer focused on interactive DELF speaking practice.

The app allows learners to choose their level from A1 to B2 and practise through live, responsive conversation. Instead of treating every deviation from textbook French as an error, the feedback is framed around the expectations of the selected exam level.

After practice, the learner can identify mistakes, review better ways of expressing the same idea and use that feedback in the next conversation.

I defined the product concept, learning logic, user flow and feedback requirements, and used AI-assisted development to turn them into a working web application.

Try the DELF Oral Trainer

WHY I BUILT IT

This is a small product built around a small problem. That is precisely why I wanted to make it.

AI made it possible to move quickly from “I wish this existed” to defining what it should do, testing it on myself and building something usable.

For me, that is where AI becomes particularly interesting: not as a reason to build something, but as a way to reduce the distance between identifying a problem and trying out a solution.

PROJECT AT A GLANCE

A1–B2

User-selected practice levels

INTERACTIVE

AI conversation, not static prompts

EXAM-FOCUSED

Level-specific feedback and better expressions