The core MVP decision is intentionally simple: one songs table and one lyrics table, with no lyrics_versions for now. The system follows a strict identity model where one MusicBrainz recording ID maps to one song, and each song maps to one shared lyrics row. This keeps duplicate handling predictable and prevents multiple competing lyric versions for the same track.
This structure matches the current project philosophy: build something solid that can mostly run on its own without heavy daily moderation. MusicBrainz is used as the identity source for canonical song selection, while SyncRoman's own database handles lyrics, updates, and community quality signals. Search is designed to run on normalized fields in the local catalog so the app remains fast, controlled, and consistent.
Development is still early-stage MVP. The foundation already supports browsing, search, lyrics preview, copy, and LRC export. What is not complete yet is the full submission and contribution workflow, duplicate-prevention safeguards in editing flow, and moderation-friendly tooling. The next phase is focused on those practical improvements, followed by stronger song pages and, only later if quality is stable, possible API exposure.
SyncRoman is not trying to look like a finished startup at this stage. It is a deliberate, sustainable build: fewer tables, clear uniqueness rules, and collaborative editing of one shared lyrics entry per song. That tradeoff gives simplicity and low maintenance now, while leaving room to evolve carefully as the catalog and contributor base grow.