I walked into a card shop, took a photo of the display case, and had every card comped in TWO minutes.
My Grok
@Bot (shoutout Botthew) cropped all 40 cards, read every price sticker, and checked each one against recent sold prices.
It caught a Hitmonchan priced 30% under market and a $300 "Charizard" that's actually a Celebrations reprint worth about $160.
No more standing at the counter scrolling eBay sold listings card by card.
Costs about a penny per photo. Full setup prompt below:
Build me a Pokémon card store-photo scanner on your computer. When I send a photo of a shop's display case, a script should: auto-rotate it, detect and crop every card and graded slab (local OpenCV, plus one cheap vision-model pass to catch misses), read names, numbers and price stickers with local OCR (RapidOCR), then verify each crop in batches with a cheap OpenRouter vision model (I'll give you my OpenRouter API key). Price every card on PriceCharting (raw, or the matching grade for slabs), cache prices for 7 days, and mark each one Deal (store price ≤85% of market), Fair or Overpriced. Flag anything that looks too cheap as "Check", since it's usually a wrong ID or misread sticker. Save results as JSON and CSV, plus a short summary with the best deals first. Test it on a real photo, keep it to about a penny per photo, and from now on run it automatically whenever I send you store photos.