Recognition
A resized scan image is evaluated alongside a compact description of the local sneaker catalog. The service may propose a catalog identifier or report that it could not find a confident match. The app checks any proposed identifier against its bundled allowlist and shows the result for your confirmation.
Unknown pairs
When a pair is not in the bundled catalog, you can ask for an AI-assisted addition. Gemini may propose a brand, model, nickname, colorway, style code, and explanation based on visible details. You can review and edit the result before saving it.
Research and rarity
The app can request Google Search-grounded research for release, retail, market, production, and rarity context. Returned source links are retained with the draft when available. Unsupported facts remain unknown, and weak rarity evidence falls back conservatively. Market values are estimates, can change quickly, and should not guide a financial decision without independent verification.
Collectible artwork
For an AI-assisted addition, the app can generate a clean side-profile illustration intended to fit the same collectible card system as the curated catalog. Generated art may simplify, omit, or distort materials, branding, color, and construction details. It is not a product photograph.
Physical-capture safeguard
The model checks for signs that a photo depicts a screen, print, ecommerce listing, or another reproduced image. Suspected reproductions may be blocked to protect the integrity of location-based finds. This is a probabilistic safeguard, not proof that an item is authentic or physically present.
Human control
- No recognized result is silently committed to your collection.
- You can retry, search the catalog, correct details, or stop.
- AI-assisted additions stay labeled unverified.
- Deleting a draft or custom sneaker removes its locally managed photo and art.
Processing and feedback
AI requests are processed through Firebase AI Logic and Google's Gemini services. Firebase App Check is used to protect the API. More detail appears in the Privacy Policy. If a result is unsafe, misleading, or repeatedly wrong, use Support and include the model name and what you expected—without sending sensitive information.