Logging footage with custom fields like client, job number and location

StoryFolder — 2026-09-02

Part of: How to log footage without tagging every clip by hand

Last updated: 2 September 2026.

A client comes back for their fourth job and asks whether you still have the exterior from the second one. The footage exists. It is on a drive, inside a project bin, under a filename somebody chose at 11pm. What makes that question answerable next time is a shot-level database with fields shaped like your business rather than like a camera: Client and Job# typed once per import, Location, Content Type, B-ROLL and Talent Name sitting on the shots themselves, and a text field holding the folder the original lives in. Six field types cover it: Text, Number, Tag, Dropdown, Checkbox and Rating. The ones a vision model should fill for you are a small subset.

That last part is the uncomfortable half of the answer. Client, Job# and release status are the fields this query asks about, and they are the fields AI is worst at, because no frame contains them. Paste those. And StoryFolder is not an asset manager: it indexes shots inside a video you import, it does not track storage, and it will never tell you which unplugged drive the original is on.

New in 0.4. Video-level metadata fields now exist, so Client and Job# can live on the project instead of being typed on every shot. How folders, collections, quick filters and video metadata fit together.

What fields do small studios actually use to log footage?

Seven or eight, and they are shaped like the business: who the job was for, which job it was, where it was shot, what kind of material it is, whether you are cleared to reuse it, and where the original file sits.

You can read that shape straight off the field names customers invented. One account built a text field called Find in Folder, because a file path is not searchable and a typed value is. Another filled B-ROLL 828 times. A third carries 589 Animation notes, one repeated per-shot classification. A fourth tracks release-form status per shot, which is a legal fact with a deadline attached. An agency account gave up on per-shot fields altogether and encodes its whole taxonomy in board titles like Client_Footage_SELECTS_BROLL. Not one of them invented a field called objects or scene type. (StoryFolder customer data: 813 storyboards from 295 paying customers, current to 1 September 2026.)

StoryFolder's guide to cataloguing b-roll across multiple clients lays that schema out field by field, with the type and the scope for each one. This page is about the next decision: how each of those fields gets filled, and which software will fill them. Build your own custom fields and pick a field type for each. Custom fields are a Pro feature.

Which fields should be video-level and which should be shot-level?

Put a field on the video when its answer is the same for the whole import, and on the shot when it varies down the board.

Client and Job# are constants. The busiest account in that data typed Client, Job# and Title 3,852 times each; typing them once per import instead is most of the saving. It is a settings decision, not a discipline problem: a category targets either the video or the shot, and the fields inside it follow. See categories, and video-level fields and how they filter whole projects.

Location is the one people get wrong. It feels like a video-level fact and it often is, until the day you have a road trip on one board and three towns in one import. If your work is single-location per job, put it on the video. If it is not, put it on the shot and accept that it costs more to fill.

One consequence worth knowing before you design the schema: AI Autofill only offers shot-scoped fields and only writes at shot level. A video-level Client will never be autofilled, which in this case is the behaviour you want anyway.

Should AI fill client, job number, or location?

Client and Job#, no: type those once on the video, because no frame contains an external business fact. Location, sometimes: a recognisable landmark may support it and an office interior will not, so check what comes back. Content Type, B-ROLL, shot size and a usability rating are the ones worth handing over.

A vision model looks at one representative frame per shot. It can tell you what is in the picture, read text that appears on screen, and classify the shot into a category you defined. It cannot know which of your clients commissioned the job, what you numbered the invoice, or whether the person on camera signed anything.

The mechanism behind that split is the field type, and it does not apply evenly. A Dropdown answer must be one of its options, a Rating must be a whole number from 0 to 5, a Checkbox must be true or false, and a Number must be a number. All four are checked against the field's own domain before the value is written. Tag and Text fields are open. A Text field has no constraint at all, because a text field's domain is anything you could have typed into it.

Which is exactly why Job# is the dangerous one. Hold it in a Text field, switch autofill on for it, and a model can return a job number that looks like yours, is formatted like yours, and belongs to nothing. Nothing in the path catches it, and once written it looks identical to a value you typed. Every field ships with its autofill switch off, so the only way that happens is if you turn it on. Worth knowing before you do: AI Autofill is a Pro feature, and it works by sending one frame per shot to StoryFolder's server, which calls a third-party vision model.

What should custom-field footage logging software let you do?

Six things, and the shopping is mostly about which of them a tool skips.

1. Define your own fields, in more than one type, so a category is a list rather than free text.
2. Decide what each field attaches to: the whole video, or each shot inside it.
3. Fill a value across many shots in one gesture, because per-shot typing is what kills every logging scheme.
4. Fill the visual fields without typing at all, on the fields you allow and no others.
5. Search and filter on those values later, together with the spoken words.
6. Get the values back out as a spreadsheet, or as a link for someone who does not have the software.

StoryFolder's search reads two layers in one pass: the values in your Text, Number, Dropdown and Tag fields, plus the transcript, across the library. It is plain substring matching, so typing a location finds the shots where you typed it. Rating and Checkbox fields are filter chips instead. See searching your library.

The distinction to hold on to while shopping is what the software counts as a thing. A file cataloguer counts files, and answers "which drive is this on". StoryFolder counts shots inside a video you imported, and answers "which moment, in which job, and is it cleared". Those are different questions and you may well need both.

How does StoryFolder compare with Kyno, Silverstack and a spreadsheet?

Competitor capabilities above were verified against vendor material on 1 September 2026.

Treat the spreadsheet as the real competitor. Most people reading this already have one, and it fails on the same thing every time: filling it costs something on every shot, and nobody pays that twice.

How do you hand this schema to another editor?

Three routes, and none of them puts your fields back into an NLE.

Export a spreadsheet with the fields you choose, which is what a producer actually wants. Publish a password-protected link for someone without the app, remembering the published page is view-only: no commenting, no picking, and no video player when the source is a local file. Or hand over the library itself, if they are working in StoryFolder too, because the fields belong to the library rather than to any one project.

What you cannot do is send your picks back as a selects sequence. There is no FCPXML or AAF round-trip, and it is the most-requested missing piece for this audience. If the schema has to arrive inside the NLE, Kyno is the honest answer.

FAQ

Are Key Takeaways and Story fields that customers created? No. They are among the twelve seeded My Notes text fields a new library ships with, and they are built for film analysis rather than client work.

Does a Text field enforce a job-number format? No. Text is pass-through with no closed domain, which is why a Job# should be typed or pasted rather than autofilled.

Can I put one value on many shots at once? Yes. Copy and paste shot metadata across a selection, with a write mode that skips cells that already have a value.

Can I search my metadata and my transcript in the same box? Yes. Library search covers Text, Number, Dropdown and Tag field values plus the transcript, using plain substring matching.

Can a spreadsheet export contain only some of my custom fields? Yes. The spreadsheet export honours a field selection. Spreadsheet export is a Pro feature.

Can a published board play the source video? Not when the source is a local file. The published page is view-only and carries no player for local-file sources.

Will AI fill a video-level field? No. Autofill offers shot-scoped fields only, and its only write path is at shot level.

← All articles