VocabVerse

Overview

I built VocabVerse around a learning pattern I personally used. Songs were a major part of how I first started learning another language, and they stayed useful because I actually wanted to return to them.

The practical problem is that lyrics are unstructured. You have to find the words, inspect dense lines, identify useful vocabulary and grammar, make study material, and remember what keeps repeating across songs. VocabVerse turns that interest into a study loop instead of leaving it scattered across tabs and notes.

I do not think that proves the app works for everyone. I think it works best for people who already care about the language and the songs they are studying.

The app is deployed at vocabverse.jaycapist.dev.

What I Built

The core flow is:

1. Sign in.
2. Search LRCLIB or paste lyrics manually.
3. Import and save a song.
4. Parse lyric lines.
5. Generate vocabulary, grammar, and kanji study items.
6. Inspect the song and generated cards.
7. Review in study modes.
8. Track progress and repeated patterns across songs.

The useful part is the study system itself: songs become lyric lines, lyric lines become study material, and review attempts become history instead of disappearing into a notebook or browser tab.

Study Data

LRCLIB provides lyric lookup, while manual lyric entry covers songs that are not available through lookup.

The important data model is the learning state: songs, lyrics, lyric lines and tokens, vocabulary items, grammar patterns, study items, review attempts, and user item stats. That is the part that lets the app remember what someone studied instead of making every song a one-off session.

Auth and Data

The current account system is lightweight private-beta auth. A user signs in with email and a beta access code, the backend creates or updates the user, and the app stores a signed HTTP-only session cookie. It is intentionally lighter than a full password/OAuth system while the product is still being shaped.

That still gives the project a real data model: users own songs, songs produce study items, review attempts create history, and stats make progress visible over time.

Current State

The data model and study-route foundation are real. The learning claim should stay honest: VocabVerse is not proven to work for everyone. It is built for people who have a genuine interest in the language and in the songs they are studying.

That is why the technical side matters here. The idea only works if the interest has somewhere to go: saved songs, saved items, review state, and enough structure that studying does not stay scattered.

Next Pass

The next useful media pass is compact: login, song library or song detail, primary study flow, and vocabulary/progress views if present.