English life

Use gpt-5.6-luna to generate a study path according to the user's level.
Onboarding: promote the English Life app
Input user name and choose level
Vocab learned by the user, user can delete
Profile screen
Detail vocabulary, user can regenerate the meaning and example of this vocab
Tap the situation, show the title of this situation
Generate background situation
Input character name
Pick character style, tap button to use gpt-image-2 model generate character
A speaking situation screen where the character and the user converse with each other using the gpt-realtime-2.1-mini model.
Situation mission words
Character screen

Inspiration

Most English-learning apps rely on vocabulary lists or scripted dialogues. We wanted to make practice feel like an adventure instead of a classroom.

English Life imagines that the learner has just moved to a new English-speaking town. Every conversation has a purpose: meeting people, settling into a home, navigating daily life, and gradually becoming more confident in English.

What it does

English Life is a voice-first, AI-powered English-learning game where users role-play as a newcomer in a foreign town.

Players follow a personalized study path of five chapters and fifty real-life situations, such as:

Meeting a landlord
Asking for help in an apartment
Talking to neighbors
Ordering at a coffee shop
Shopping for essentials
Asking for directions
Joining community activities
Handling everyday services

For every situation, AI creates a level-appropriate welcome, mission keywords, character context, and conversation guidance. Users speak naturally with an AI character in a generated scene, while the app transcribes their speech and marks mission keywords as completed in real time.

Completing a mission unlocks the next situation, awards progress, and helps build the learner’s character book and vocabulary library.

How we built it

SwiftUI and MVVM for the native iOS application
OpenAI Realtime API for live, turn-based voice conversations
OpenAI transcription within the Realtime session for learner speech transcripts
gpt-5.6-luna through Chat Completions for study-path generation, mission guidance, vocabulary definitions, and examples
gpt-image-2 for custom character portraits and situation backgrounds
iOS Vision foreground extraction to remove portrait backgrounds after image generation
Structured JSON prompting for stable study paths, mission keywords, and vocabulary data
Semantic VAD and playback/input gating for natural conversation turns
Local progress, generated-media, mission-guidance, character, and vocabulary caching

Challenges we ran into

The biggest challenge was making conversations feel natural while keeping them focused on the learner’s current mission.

We also had to:

Keep each AI character in role and relevant to the exact situation.
Prevent Realtime voice turn-taking issues, such as a character responding before the learner speaks.
Match spoken mission phrases reliably despite punctuation, casing, accents, and multi-word expressions.
Generate characters and locations without regenerating assets that already exist.
Persist AI-generated paths and mission data so learners resume exactly where they left off.
Balance AI creativity with a clear educational objective.

Accomplishments that we're proud of

Created a voice-first role-playing learning experience instead of a traditional lesson flow.
Built personalized five-chapter study paths with fifty progressive situations.
Connected live speech transcripts directly to mission completion and map progression.
Generated and cached reusable AI characters and location backgrounds.
Built a character book, profile dashboard, and vocabulary library around learner progress.
Designed an immersive town-based progression system that makes speaking practice feel purposeful.

What we learned

This project showed us that AI can make language learning feel much more human when the learner has a role, a clear objective, and immediate feedback through progress.

We learned how structured prompting, realtime transcription, voice interaction, generated visuals, and persistent game progression can work together to create a more engaging language-learning experience than simple question-and-answer practice.

What's next for English Life

Our roadmap includes:

A larger explorable town with more locations, story arcs, and recurring NPCs.
Personalized future missions based on completed transcripts and learner weaknesses.
Pronunciation and grammar feedback after speaking turns.
Long-term memory so AI characters remember past conversations.
Daily missions, achievements, and story-driven quests.
Support for additional languages beyond English.
Secure server-side issuance of short-lived Realtime credentials for production.