Scores

Picturesque judges photos with several independent signals, and combines them for ranking and sorting. Each comes with a detailed page.

Signal	What it measures	Needs
Technical scores	Sharpness, exposure and noise, measured from the pixels (Quality scores)	Nothing; always computed
Aesthetic score	How good the photo looks to people, from a model trained on human ratings (Aesthetic score)	The CLIP image model
Faces	Faces found, closed eyes and how sharp each face is (Faces and closed eyes)	The face models
Similarity	Whether two photos show the same scene: perceptual hashes and CLIP embeddings (Visual similarity)	Hashes always; embeddings with the CLIP model

How they are used:

• Series group photos taken close together that look alike (Series).

• The best photo of a series is ranked from sharpness (relative to the series, with faces counting), exposure, noise and the aesthetic score; closed eyes halve a photo’s score. The weights are settings.

• Every photo gets a combined score (0–100%) to compare photos across the library; Best quality first sorts by it (Combined score).

• The loupe’s Why this score list explains a photo’s score in words.

Scores are suggestions. Your own picks, rejects, stars and best-photo choices always win, and are never changed by re-analysis.