Photograph your shelf, let AI read the spines, and every tape gets an ID, a condition and a price, so you know what you own, what it's worth, and what to sell.
VHS Box runs in your browser and on your own server. Point a webcam or phone at the shelf and work through the pile.
The multi-spine crop grabs five tapes per shot. Press Space to stage frames and Enter to send them all. Barcode scanning fills titles from UPC lookups.
A local Ollama vision model (llava) or the Claude Vision API reads each spine and cover. You get review cards to confirm, edit, or swipe away, then one tap on ✓ All.
| VHS-0001 | Akira | In Coll. |
| VHS-0003 | Ghostbusters | For Sale |
| VHS-0005 | Night of the… | For Sale |
Every tape gets a permanent VHS-XXXX ID, a condition grade, a status and a value estimate. Export CSV/JSON, print price tags, or generate eBay/Mercari sell drafts.
The launch video, cut vertical with captions for Shorts, Reels and TikTok: snap the shelf, confirm what the AI read, and every tape gets an ID, a condition and a price.
Stage webcam frames, then batch them to AI. Zoom slider, torch toggle, and crop presets for covers, a single spine, or five spines.
Fills year, label, value and genre tags for incomplete tapes, cross-checked against OMDb.
Track in collection, for sale, sold, donated, missing, or wanted. Grade great to poor, with notes and multiple photos.
Full-text search over title, label, barcode, notes and tags. Filter by status, condition, label, tag and year range.
A masonry wall of covers or a shelf of spines. Click any tape to open its details.
Copy-ready eBay and Mercari titles, descriptions and suggested prices for every tape marked for sale.
CSV (formula-injection safe), JSON round-trip with photos, printable lists and dashed-border price tags.
Optional Google sign-in and a public /c/<uuid> link per collection. Runs single-user if you skip auth.
Responsive layout, rear-camera preference, and an auto-generated LAN HTTPS cert so the camera works on your phone.
Long-press any tape for a full-screen FBI warning followed by its trailer. Some titles do more:
Express + PostgreSQL in Docker. Bring a Postgres URL (Neon works) and optionally a local Ollama or an Anthropic API key for AI scanning.
# clone and configure $ git clone https://github.com/nitsuah/vhs.git $ cd vhs $ cp .env.example .env # set DATABASE_URL + HOST_IP # start it $ docker compose -f config/docker-compose.yml up -d --build # open http://localhost:8080 # phone camera: https://<LAN-IP>:8443
llava:7b) locally or the Claude Vision API via ANTHROPIC_API_KEY./api/ca-cert once (check its SHA-256 fingerprint against the host first, see the README) and the camera works over LAN HTTPS.