People
Everyone the app found across your library, grouped automatically. Click a name to change it.
Find every photo of you
Upload 2–3 clear photos of the same face. Averaging a few references is the single biggest accuracy win available — one photo locks in whatever lighting and angle it happened to have.
Drop reference photos here
or click to choose
Matching strictness
Lower is stricter. 0.40 is a starting guess, not an answer — run tune.py against photos you can label by hand to find the right number for your set. It usually sits tighter when everyone wears similar headwear, because that compresses the gap between “same person” and “different person”.