Could be local only. New faces are saved as they're detected. Then later on your computer, add notes to the new faces ("John Smith, met 11/7/19 at Liz's party, programmer at Initech").

When those faces are re-detected, show the info text.

Would certainly be nice if you could do a similar thing without needing a camera on the glasses at all though. How can we recognize a person reliably without seeing them? Maybe by their voice?