XMP Tools

A creative team receives several hundred images from a shoot. An agent calls Read Metadata on each one and gets back the caption, keyword list, credit line, rating and camera model without opening a single file by hand. It routes five-star images into the selects folder, files everything else, and writes the caption and credit into the asset record in Airtable so the copywriter working two weeks later has the photographer's own words rather than a filename. Language alternatives resolve to their default value while keeping every translation, so a multilingual caption set survives the read intact.

Before assets are pushed to the public library, an agent calls Build XMP Packet once with the campaign's copyright statement, usage terms, creator list and keyword set. Titles, descriptions, rights statements and usage terms are written as language alternatives, creators as an ordered sequence and keywords as an unordered bag, exactly as the specification requires, and anything outside the built-in list goes in as a custom qualified property. That single packet is then fed to Write Metadata for every asset in the batch, in Merge mode so the photographer's existing capture data is preserved and only the rights properties change. The stamped files are published through Cloudinary.

An agent handling incoming user-submitted photography calls Inspect Media File first, which reports the container format, file size, pixel dimensions, whether an XMP packet is present and how large it is, and whether this connector can write back into that container, without parsing anything. Files in a container that cannot be written to are branched to a re-export step. Files with no packet skip the read entirely. Files with a packet go on to Read Metadata, and the rating and colour label decide which review queue they land in. One cheap call decides the shape of everything downstream.