Rei-Lingo

Inspiration

I wanted to learn English ten times faster, but conventional study methods were difficult for me to continue. I enjoy reading stories, so I began exploring whether a manga-style experience could make language learning more engaging and easier to return to every day.

The original idea was designed for my own English-learning problem. For this public Build Week submission, I present a Japanese-learning version so that judges and learners outside Japan can understand and experience the concept directly. The story also provides a small window into Japanese language and culture.

What it does

Rei-Lingo is an experimental language-learning experience built around an interactive vertical manga.

The current prototype demonstrates the first episode through Scene 2. The learner swipes through the story to reveal the comic panels, dialogue balloons, and sound effects step by step. Synthetic voice is used to support the dialogue and connect written language with pronunciation.

This submission is a focused vertical slice, not a complete language-learning course. It demonstrates the interaction pattern and presentation style that the larger system is intended to use.

How we built it

I used GPT-5.6 and Codex throughout the planning and implementation process. They helped me structure the learning experience, design the story flow, implement the swipe interaction, refine the dialogue presentation, investigate bugs, and improve the consistency of the visual assets.

I also prepared character and asset references before generating the story images. This helped keep the characters, locations, and props more consistent across the scenes.

Challenges we ran into

The main challenge was creating a readable manga-like learning experience while keeping the implementation small enough to complete. A second challenge was visual consistency: generated images can change character details, clothing, lighting, and props between scenes.

Another challenge was presenting the project honestly as an early prototype. The full learning system is larger than the current demo, so the submission focuses on one working slice instead of claiming unfinished features as complete.

Accomplishments that we're proud of

We created a working vertical manga prototype in which the learner controls the pace by swiping. Panels, dialogue balloons, sound effects, and synthetic voice are presented in sequence rather than all at once.

We are also proud of using character sheets and asset references to improve visual continuity in the generated story images.

What we learned

I learned that a language-learning product needs more than translated text. Story pacing, visual context, pronunciation, learner control, and the amount of assistance shown at each moment all affect whether the experience feels useful.

I also learned that AI-assisted development still requires careful verification. GPT-5.6 and Codex can accelerate planning and implementation, but the result must be tested against the actual behavior of the application.

What's next

Planned future work includes completing the rest of the first episode, adding branching choices, free-form answers, delayed review, and additional story episodes.

The visual presentation is also intended to evolve from static panels into a restrained interactive motion-comic experience, while preserving readability and learner control. Future story worlds, such as medieval fantasy and science fiction, should be added through external content files rather than major changes to the core player.

The long-term goal is to support both Japanese learning for people around the world and English learning for Japanese speakers.

Built With

• cloudflare
• codex
• gpt-5.6
• html5
• next.js
• openai
• react
• speech
• typescript
• web