Creative professionals and hobbyists have vast photo/video libraries but struggle to find specific moments. Current search relies on manual tagging or inaccurate date-based sorting.
Build a desktop app that indexes visual content and enables semantic search ('beach sunset from last summer with dogs'). Use local ML models for privacy-focused users.
Charge photographers, videographers, and memory-keepers via subscription or one-time license.
MVP could be a simple Finder/Photos extension with basic object/scene recognition.
Risk: Cloud giants may build similar features into existing photo apps.