Every starred photo in the Canto library, given a written AI caption with its
location woven in, then written back into the library's AI Caption field. Each link below opens
that geography's review page, where any caption can be edited in place and saves live.
The bar shows how many frames in that pool carry a specific named site rather than
just the region. It varies with the metadata, not the effort: pools with GPS and place-named files
place near 100%, pools with neither are lower by design, because a guessed location is worse than none.