Press'd is a music rating app for people who take albums seriously. Rate every song, track your listening, get AI-powered predictions of what you'll love next, and see what your friends are into.
▶ Try it live · Sign in with Google — no setup required.
Think Letterboxd, but for albums — with sharper scoring and a personal recommendation engine built in.
You add an album, listen through it track by track, and rate each song as you go. Press'd combines those song scores with a few album-level judgment calls (production, replayability, and more) into a single, consistent score. Over time you build a complete, searchable record of your taste — and Press'd learns it well enough to predict how you'll rate things you haven't heard yet.
It's also social: follow friends, watch what they're rating, like their reviews, and send albums straight to their queue.
- Search and add any album from a catalog of millions — album art, tracklist, and release year fill in automatically.
- Listen-through rating. Songs unlock one at a time in track order, so you rate the way you actually listen — top to bottom, no jumping ahead.
- Score every song from 0–10.
- Album-level factors. After the songs, rate the album on Theme, Replay Value, Production, and Distinctness.
- Live score. A running score card updates with every rating so you can see the album's number take shape in real time. Songs are flagged as bangs 🟢 or skips 🔴, and a clean album with zero skips earns a "No skips" badge.
- Shareable rating cards. Every finished album generates a polished, downloadable report you can share (including straight to iMessage).
- Personal predictions. Press'd trains a model on your own ratings and predicts a score for every album in your queue — so your "To Listen" list is automatically sorted best-first.
- Explained guesses. Predictions come with reasoning on an album's likely theme, replay value, and distinctness.
- Audio-aware. Predictions factor in real audio characteristics (tempo, key, energy, danceability, loudness) analyzed from the music itself.
- Discography discovery. On any artist's page, see the releases you haven't logged yet.
- Library organized into To Listen → Listening → Rated, so nothing gets lost mid-listen.
- Leaderboards for your top albums and songs, fully sortable and searchable.
- Stats dashboards — score distributions, genre breakdowns, and how your taste breaks down across your whole catalog.
- Baseball-style artist stats for the stat-heads: consistency, songs-above-replacement, bang %, skip %, and more.
- Invite friends with a one-tap link (or text it).
- Friend requests and a friends list — view any friend's full library and stats.
- Activity feed of everything your friends are rating, with likes on reviews.
- Recommend albums to a friend and they land right in their To Listen queue, tagged as coming from you.
Works great on mobile and desktop.
Press'd is a full-stack app:
- Frontend — React 19 + TypeScript, Vite, Tailwind, TanStack Query/Table, Recharts, Google OAuth. Deployed on Vercel.
- Backend — FastAPI + SQLModel on PostgreSQL. Deployed on Render.
- Machine learning — a gradient-boosted song-score model (retrained on your ratings), an LLM-based theme/distinctness predictor with retrieval over an album-analysis corpus, and Essentia for audio feature extraction. A nightly worker keeps everyone's predictions fresh.
- Auth — Sign in with Google (JWT sessions).
Backend
pip install -r requirements.txt
uvicorn backend.main:app --reloadFrontend
cd frontend
npm install
npm run devCopy the relevant environment variables (database URL, Google OAuth client ID, Spotify credentials, etc.) into .env / frontend/.env.local before starting.
Built by Jack Kalsched & Claude. 🎶