LabelScan · Vinyl & 78 Value Spotter (MVP)

On-device OCR + heuristics to identify collectible record labels (classical LP grails & key 78 rpm series). Nothing uploads—runs entirely in your browser.

Features

• Camera or photo upload
• OCR via tesseract.js (on-device)
• Heuristic matching for valuable labels (RCA Living Stereo Shaded Dog, Mercury Living Presence, Decca SXL ffss, DG Tulip, London Blueback, etc. + Paramount/Okeh/Vocalion for 78s)
• One-click links to eBay search, eBay Sold, Discogs for comp checks

One‑Click Deploy with Vercel

After you push this repo to your own GitHub account, you can deploy with a single click:

Replace <YOUR_USER> with your GitHub username. If you haven’t pushed yet, follow the Deploy to Vercel (Step‑by‑Step) section below first.

---

Local Development

1. Prereqs
   • Node.js LTS (>= 18) and npm (or pnpm/yarn)
2. Install & run

   npm install
   npm run dev

3. Visit http://localhost:3000. Grant camera permission when prompted (Chrome/Edge allow camera on http://localhost).

If the camera preview is black: click Start camera after granting permission, or switch to Upload mode and test with a label photo.

---

Deploy to Vercel (Step‑by‑Step)

A) Push this project to GitHub

git init
git add -A
git commit -m "Initial commit: LabelScan MVP"
git branch -M main
# Create a new empty repo on your GitHub account named labelscan-mvp, then:
git remote add origin https://github.com/<YOUR_USER>/labelscan-mvp.git
git push -u origin main

B) One‑click deploy

1. Go to https://vercel.com/new
2. Click Continue with GitHub and authorize if needed.
3. Select your labelscan-mvp repo.
4. Framework is Next.js (auto-detected). No env vars needed.
5. Click Deploy.

When deploy finishes, Vercel will give you a live URL like https://labelscan-mvp.vercel.app. Open it on your phone, grant camera permission, and test.

Tip: iOS Safari requires a user gesture to start the camera; use the Start camera button.

---

Notes & Roadmap

• Tesseract will download the English traineddata on first use (CDN). Keep a good connection the first time.
• Add a small CNN logo classifier (TensorFlow.js) to complement OCR for tough fonts.
• Extract catalog numbers (LSC/SXL/SAX/14000‑D) to auto‑build precise eBay/Discogs queries.
• Add condition checklist and confidence score.
• Batch mode for rapid estate-sale triage.

MIT License