Entry of
How to search a footage database and actually find the shot
Footage search boxes sit on top of catalogues built by people describing cans of film, not answering questions. Typing what you want rarely works.
ResearchMethod9 min read
Search boxes on footage libraries look like search boxes on everything else, which is the problem. They are sitting on top of catalogues built by people who were describing cans of film, not answering questions. Typing what you want rarely works. Typing what was filmed usually does.
Why keyword search fails on archive collections
Contemporary contributor libraries are indexed for search: the shooter uploads a clip and tags it with the words buyers type. Archive collections are indexed for retrieval: the record describes a production, a reel, a story number and a date, in the vocabulary of the year it was catalogued.
So a 1961 record will not say "commuters looking stressed". It will say the name of the station, the city, and possibly the newsreel story number. Search the second thing and the shot appears.
Describe the event, not the feeling
Rewrite the request into facts a cataloguer could have written down: what happened, where, when, and who was visible. Then search each fact separately before combining them.
| Weak query | Stronger query | Why it works |
|---|---|---|
| Post-war optimism | Housing construction, 1951, exterior, families | Names a filmed event rather than an interpretation |
| Busy office | Typing pool, typewriters, open plan, 1960s | Uses period objects, which cataloguers describe |
| Beautiful coastline | Aerial, cliffs, Atlantic, sunrise, locked-off | Combines shot type with a place and a time of day |
| Old computers | Mainframe, tape reels, machine room, technicians | Names the object as it was called at the time |
| Protest footage | Demonstration, city and date, banners, police line | A dated event is a catalogue record; a mood is not |
The filters that repay learning
Most of the time saved in footage search comes from filters, not from better words.
- Resolution and frame rate. Filter before you fall in love with a shot you cannot use in a 25 frames per second timeline.
- Release status. Model released, property released, or editorial use only. Filter on this the moment the project is commercial.
- License type. Royalty free, rights managed, subscription. Mixing models across one sequence complicates the clearance list later.
- Date filmed against date uploaded. These are different fields and libraries sort on the second by default, which buries older material.
- Aspect ratio and orientation. The fastest way to exclude vertical and 4:3 material, or to find only that.
- Collection or contributor. Once one good shot is found, searching the rest of that collection is more productive than any new keyword.
Shot vocabulary libraries actually index
Cataloguers and contributors both use the language of the camera. It is worth using it back.
- Framing: wide shot, medium shot, close-up, extreme close-up, insert, two-shot.
- Movement: pan, tilt, tracking, dolly, handheld, locked-off, crane, drone, aerial.
- Time: timelapse, hyperlapse, slow motion, high speed, real time.
- Viewpoint: point of view, over the shoulder, reverse angle, overhead, low angle.
- Content: establishing shot, cutaway, b-roll, archival, reenactment, stock footage.
Combining one framing word, one movement word and one subject word is the single most effective query pattern across almost every library.
Tactics that only work on archives
- Search the finding aid, not the video player. Many institutions publish catalogue records long before the material is digitised. The record proves the shot exists, which is enough to start a request.
- Search by production, series or donor. Archive material arrives in blocks. Find the block and you find twenty more shots.
- Search the trade press. Dating a production or confirming a company name in a contemporary trade journal often supplies the exact term the catalogue uses.
- Search adjacent events. If the event you want was not filmed, the visit, the parade or the inspection the same week probably was.
- Ask a human. Archive researchers know their own catalogue's blind spots. A precise written request beats another afternoon of keywords.
Keeping what you find
A search is worthless if the results cannot be handed to someone else. Every library offers a saved collection, called a bin, a lightbox or a board. Use it, and record four fields for every candidate.
- Asset identifier
- The supplier's own reference. Titles change and thumbnails are ambiguous; the identifier is what a quote will be issued against.
- Source
- Which library, and which collection inside it. Two aggregators can carry the same clip on different terms.
- Rights statement
- Copied exactly as written, with the date you read it. Statements are edited over time.
- Timecode in and out
- The section you actually want, not the whole reel. Archive quotes are frequently calculated on duration.
Comps, watermarks and the offline edit
Most libraries let you download a low resolution watermarked copy for editing before you buy. That copy is for internal work only: it must not be published, shown to a paying client as a finished piece, or posted for approval on an open link. Cut with the comp, confirm the sequence works, then license only the clips that survived the edit. Buying before the cut locks is how footage budgets get spent on shots nobody uses.
The comp and the master are not the same file in any respect that matters to the online, and the gap between them is where late problems appear. The codecs, frame rates and colour spaces involved are set out in what you actually receive, and whether it will cut.
Where to go next
If the search is turning up nothing because the material is old, where archival and newsreel footage actually lives lists the institutions and houses that hold it. When it is time to hand the search to someone else, the nine lines a footage research request needs is the format researchers reply to fastest.