Immergo

homepage
architecture diagram

Inspiration

Learning a new language is difficult because most apps focus on memorizing vocabulary and grammar instead of practicing real conversations. Many learners struggle when they finally need to speak in real-world situations, such as ordering food, asking for directions, or buying a ticket.

We wanted to build something that feels closer to actually traveling to another country and talking to people. Inspired by the capabilities of real-time AI conversations using the Google Gemini, we decided to create an immersive role-playing experience where users can practice speaking with an AI that behaves like a real person.

That idea became Immergo — a conversational AI language learning experience.

What it does

Immergo is an immersive language learning app that lets users practice speaking through interactive role-play missions.

Instead of passive lessons, users enter real-world scenarios such as:

Buying a bus ticket
Ordering coffee
Meeting a neighbor
Asking for directions

The AI plays a specific character (for example a bus driver or barista) and responds naturally to the user's speech.

Immergo offers two learning modes:

Teacher Mode

The AI explains phrases and grammar in the user’s native language
Provides translations and helpful guidance

Immersive Mode

A strict “No Free Rides” policy
The AI only accepts the target language
Forces users to practice real conversational fluency

The app also includes performance scoring, grading the user’s fluency as:

Tiro (Beginner)
Proficiens (Intermediate)
Peritus (Advanced)

How we built it

Immergo uses a real-time AI architecture combining modern web and AI technologies.

Frontend

Vanilla JavaScript
Vite
Web Audio API for capturing microphone input
WebSockets for real-time streaming

Backend

Python with FastAPI
AI integration through the Google Gemini Live SDK

Infrastructure

Real-time audio streaming using WebSockets
AI responses generated and returned as low-latency speech
Optional deployment via Google Cloud Run

This architecture allows natural conversations where the AI listens and responds almost instantly.

Challenges we ran into

One major challenge was handling real-time voice interaction. Streaming audio between the browser and the AI model required careful coordination to avoid delays or broken conversations.

Another challenge was designing AI personas that stay consistent within a scenario. For example, a bus driver should only respond like a bus driver and not suddenly become a teacher.

We also had to balance helpful feedback with immersion so that the learning experience still feels like a real conversation rather than a quiz.

Accomplishments that we're proud of

We are proud that Immergo successfully creates a real-time conversational language practice environment rather than a traditional lesson-based app.

Some key achievements include:

Real-time voice conversations with AI
Role-play scenarios that simulate real-world situations
Two learning modes that support both beginners and advanced learners
A clean architecture using modern web technologies

Most importantly, the experience feels like talking to a real person, which is a huge step forward compared to traditional language learning tools.