Skip to content

FEATURES.md

Latest commit

 

History

History
115 lines (93 loc) · 5.16 KB

FEATURES.md

File metadata and controls

115 lines (93 loc) · 5.16 KB
Raw

Features

Detailed feature walkthrough for 歌詞ノート (Kashi Note). For setup and deployment see README.md and DEPLOYMENT.md.

Lyrics Reading

  • Furigana Lyrics — Paste Japanese lyrics; client-side kuromoji-es auto-converts kanji to hiragana furigana via <ruby> annotations (lazy-loaded on first use, works offline after that).
  • Reading Modes — Switch between original lyrics, furigana, and Hepburn-style romaji; the preference is remembered locally.
  • Adjustable Font Size — A−/A+ controls for comfortable reading.
  • Lyrics Dot Grid — Canvas dot-matrix background behind the lyrics with a pointer spotlight. Tunable live from the debug panel (spacing, dot size, spotlight radius, glow, magnet pull).
  • Apple-style Follow Scroll — While syncing, the active line is centered with an eased scroll animation (fast start, graceful deceleration).
  • Copy Lyrics — Strip furigana, copy clean text to clipboard.
  • Export — Download lyrics as plain text, LRC (timestamped), or HTML.
  • PiP (Picture-in-Picture) — Floating lyrics window over other apps (desktop Chrome).

Lyrics Sources

  • lrclib.net Sync — Fetch timestamped lyrics for precise per-line synchronization (with fallback providers and heuristic match scoring).
  • One-Click Import — Import lyrics for the currently playing Spotify track instantly.
  • Playlist Batch Import — Import all tracks from a Spotify playlist at once, with per-track result reporting.

Timeline Annotation Workspace

  • Mark previously untimed lyrics line by line against live Spotify progress (keyboard shortcuts: Enter mark, ↑/↓ navigate, Ctrl/Cmd+Z undo, Ctrl/Cmd+S save).
  • Save partial work at any time; reopen and continue.
  • Replay marked positions (seek Spotify to any line), clear individual timestamps, and apply global offsets (±500/±100ms quick buttons or a custom value) to shift the whole draft.
  • Live progress card, track-mismatch warning, and dirty-state leave guard.

AI Lyric Translation

  • SSE streaming — Translations stream in as the model thinks; a live reasoning panel shows the model's thought process chunk by chunk (auto-scrolls unless you scroll up to inspect).
  • Terminology glossary — Extracts names/terms from the source lyrics and injects them into the prompt for consistent translations.
  • Pluggable providers — OpenAI-compatible APIs, Anthropic Messages API, and Cloudflare Workers AI (TRANSLATION_PROVIDER).
  • Resilience — Retry with backoff on transient failures, per-error localized messages, and a daily token quota (AI_DAILY_NEURON_LIMIT, 429 / ai_quota_exceeded → quota toast).
  • Whole-song workspace — /songs/[id]/translation translates the full lyrics into the target language (SSE, max_tokens 32768).

Spotify Integration

  • OAuth 2.0 login with an optional passphrase gate.
  • Real-time sync — line-by-line auto-scroll with the currently playing track. Server mode (Node singleton poller + SSE) or client mode (browser polling), see SPOTIFY_POLL_MODE.
  • Canonical metadata — stable Track IDs, URI, album, duration, cover, and canonical title/artist persisted for exact matching; provenance (provider, match confidence, fetch time) recorded.
  • One-click import of the playing track; seeking from the timeline.

Share Cards

  • Generate a shareable canvas image (landscape 1200×630 or portrait 630×1200) with the cover art, selected lyric lines, a QR code linking back to the song, and the site name — download as PNG.
  • Select/deselect individual lines (?line= deep-links a selection).

Admin Console

/admin (admins only):

  • Users — promote/demote admins, block/unblock (with reason), delete users (self-protected).
  • Songs — toggle public/private visibility, approve or reject public-approval requests (pending queue with prominent actions), delete songs.
  • Translation service — live provider/model configuration and a connectivity test against the configured API.

Experiments

Open a song page → More menu (desktop toolbar or mobile overflow) → Experiments:

  • Capture mic spectrum — lights the dot grid's bottom rows with a live frequency wave (Web Audio AnalyserNode, fftSize 2048 ≈ 23 Hz resolution; log-spaced bins mapped across the full panel width; peak capped at 1/3 of panel height; dB-style scaling).
  • The toggle persists while the panel is closed; the mic is released only when the toggle is off or you leave the page (unmount).
  • Requires a secure context (HTTPS or localhost) and mic permission.

Platform

  • PWA — Installable on Android/iOS with offline caching and update notifications.
  • Dark / Light Theme — System-aware with manual toggle, persisted via localStorage.
  • Multi-Language UI — Japanese, English, Simplified Chinese, Traditional Chinese (auto-detected from browser).
  • Favorites & Collections — Star songs, organize into collections, filter by favorites, bulk-manage from the home page.
  • Responsive — Mobile-optimized bottom bar with a 3-dot overflow menu; desktop toolbar menu for the same actions.