Both layouts are based on finding optimal photo row breaks using a variant of Knuth & Plass.
Flex
Flex is the simpler layout of the two where all rows are approximately the same height. Think of it as Album with smarter row breaks.
Mostly noticable with panoramas and at the end of the collection (it will try to avoid photos hanging off the end). It will also add locations inline if photos have GPS coordinates and reverse geocoding is enabled.
Highlights
Highlights is a more experimental variant on Flex, but with a bit of a twist! Instead of a more or less content row height, it will try to vary it based on the "sameness" of the photos. Best used with bigger photo sizes (pictured button).
The idea is to make travel photo collections more skimmable by shrinking similar and repeating photos.
Similarity is based on the AI image embeddings, so photos close in "meaning" will be considered as semantically similar. Seems to work quite well, but you need to have AI enabled and photos scanned for it to work as expected.
TODO
[x] make it handle non-ai-indexed photos more gracefully than just skipping them
Both layouts are based on finding optimal photo row breaks using a variant of Knuth & Plass.
Flex
Flex is the simpler layout of the two where all rows are approximately the same height. Think of it as Album with smarter row breaks.
Mostly noticable with panoramas and at the end of the collection (it will try to avoid photos hanging off the end). It will also add locations inline if photos have GPS coordinates and reverse geocoding is enabled.
Highlights
Highlights is a more experimental variant on Flex, but with a bit of a twist! Instead of a more or less content row height, it will try to vary it based on the "sameness" of the photos. Best used with bigger photo sizes (pictured button).
The idea is to make travel photo collections more skimmable by shrinking similar and repeating photos.
Similarity is based on the AI image embeddings, so photos close in "meaning" will be considered as semantically similar. Seems to work quite well, but you need to have AI enabled and photos scanned for it to work as expected.
TODO