Metadata should outlive the app that made it

Here is the test I apply to any tool that wants to organize my footage: uninstall it, in my head, and look at what is left on the drive. If the answer is "the media, untouched, plus plain files describing it", we can talk. If the answer is "a database only the app can open", the conversation is over, because I have been to this funeral before.

The graveyard is well attended

Every editor over thirty has lost a catalog. Aperture users spent years keywording and rating their libraries; Apple discontinued it, and that work had to be migrated out through whatever exits existed, or abandoned. Final Cut Server, an asset manager wired into real facility pipelines, ended with a press release. iView MediaPro was a good cataloger that became Microsoft Expression Media, then Phase One Media Pro, then nothing: three logos on one slow grave. The pattern is not that these were bad tools. Most were good. The pattern is that cataloging tools die at a steady rate, and each one takes its database format down with it, along with every hour users spent filling it.

The footage survives these funerals fine. Video files are open enough that a clip from 2009 still plays today. It is the description of the footage, the part someone typed in by hand, that dies with the vendor.

A private database is a quiet hostage situation

The modern version of the problem is politer. Today's platforms extract real metadata for you (transcripts, tags, people, descriptions) and keep the results in their database, reachable through their app, their search box, their subscription. The compute is honestly priced and useful. The custody is the trap. Stop paying, or watch the product get acquired and sunset, and the description of your own footage stays behind on someone else's server.

I sell cloud compute myself, so this is not an anti-cloud argument. Paying to make metadata appear faster is a fair trade, and speed is worth real money on a deadline. Paying to keep access to work you already did is not a trade, it is rent. The line to hold: pay to avoid effort, never to get your own results back.

Plain files beside the footage are the contract

Exactly one storage arrangement has survived every vendor death so far: ordinary files, in open formats, sitting next to the media they describe. A sidecar file on the same drive as the clip needs no company to stay alive. It travels when the drive travels. It gets backed up when the footage gets backed up. It can be read by a script, grepped on a deadline, and understood by a stranger a decade from now.

This is the arrangement ClipLogger commits to: every confirmed decision is written to a readable JSON sidecar beside the clip, on by default, carrying the media facts, the AI evidence, and your calls, with optional XMP, Resolve CSV, FCPXML, and ALE writers for pipelines that want native shapes. The app's own index is deliberately a rebuildable cache. Delete the application and the description of your footage sits exactly where the footage does. Ten lines of script can pull every tag out of a season's sidecars with no ClipLogger installed anywhere.

What to demand from any tool you adopt

The checklist is tool-agnostic and worth applying to anything that touches your library. One: decisions must land in documented plain-text files beside the media, by default, not behind an export you must remember to run. Two: the internal database must be a cache you can delete and rebuild, never the only copy of anything. Three: exports must exist in the shapes your NLEs already read, so the metadata works even where sidecars are ignored. Four: a rename or a move must not orphan the record; ask exactly how metadata finds its file again. Five: run the delete test for real on one folder. Describe ten clips, remove the app, and look at the drive. Whatever is still readable is what you will actually own in ten years.

Ask these before the honeymoon. Every dead cataloger above would have failed the checklist on day one, in five minutes, and saved its users the years.

The description is becoming the valuable half

Cameras made capture cheap. AI is making description cheap. What stays scarce is a library where the descriptions are confirmed by a human, portable by design, and provably yours. Treat the metadata like a second negative: you would never keep the only copy of your footage inside one company's proprietary container, and the words that make the footage findable deserve the same paranoia.

Any app can help you describe footage, but only open files beside the footage guarantee the description is still yours after the app is gone.