The Footage Catalogue

Creative agencies accumulate footage faster than they can organize it. Every shoot produces dozens of clips: some are the hero shots, most are B-roll, and plenty are unusable (accidental recordings, test clips, redundant angles).

Six months later, nobody remembers which is which. The team either re-shoots content they already own, or someone loses an afternoon scrubbing through folders.

Day one: the honest answer. Asked whether this was even possible, the agent checked its own tooling before committing, then proposed a pilot on a single folder rather than promising the whole library. Nine clips came back with descriptions, 51 unique keyword tags, and quality ratings in a filterable sheet.

Day one: the wall. Scaling up, the agent hit a hard limit: the video-analysis platform rejected any file over 50 MB, and roughly 40% of the library was above that line. Modern phone footage is large. The agent reported this as a finding, with three options and a recommendation, rather than quietly skipping the files.

Day two: the wall removed. A second processing path was built: for oversized files, the agent extracts keyframes locally and analyzes them itself. A 220 MB 4K clip that previously could not be processed at all now takes 24 seconds. There is no longer a file size limit of any kind.

The test was rigged, in the good way. The client-services partner later revealed he had deliberately chosen the hardest folder available: multiple subfolders, mixed footage types, the kind of mess the team deals with constantly. It wasn’t a friendly demo. It passed anyway.

Ask in plain language. “Find me drone shots of the marina, prime quality only,” in the team’s own Slack, in the channel they already work in. The agent returns the matching clips and their exact storage paths.

Or filter the sheet. Every clip is a row: description, keyword tags, location, shot types, quality rating, and path. Filter it like any spreadsheet.

“The description is so detailed, and to the point.”

“It was perfectly produced. Exactly what we wanted.”

“Everything about it is awesome.”

It told the truth about its limits. The size limit was surfaced as a finding with options, not buried. That’s what makes the rest of its output trustworthy.

It improved within a day. The limitation was found on day one and engineered away on day two, before the deadline that had been set for it.

The capability is permanent. The pipeline now exists. Every future folder costs zero team hours. The archive stops being a liability the moment it’s catalogued, and stays that way.

Humans stayed in the loop. Access is controlled, output is reviewed by the people who use it, and the operator oversees every run. Nothing about this is unattended.

Every service business has one of these: a valuable asset that’s technically searchable but practically invisible, because organizing it was never worth someone’s month. That’s precisely the shape of work a digital employee absorbs: high-volume, judgment-light, endlessly deferred.

The footage was the pilot. The pattern generalizes.