How to catalogue your video library once so you can ask for montages later

Last updated 2026-09-10 · 8 min read

What is the short version?

If you have hundreds of clips, the expensive part of every montage is finding the clips. Doing that once and keeping the answers turns "make a montage of the scrap finds" from a day's work into a sentence.

1. Go to Vizard Agent with a link to the folder.
2. Ask it to review and categorise everything first.
3. Then ask for montages by theme, as often as you like.

What do you need before you start?

Somewhere the files actually live and a sense of the themes you care about. You do not need to list them exhaustively — naming two or three examples tells Vizard Agent what kind of distinctions matter in your material, which is what the catalogue gets built around.

• A shared folder link, or the files themselves.
• What the videos are of. Your subject, in a sentence.
• The kinds of things you will ask for. Two examples is plenty.
• Whether speech matters. Usually yes.
• How many there are, roughly.

What do you type into Vizard Agent?

Ask for the catalogue as its own job. The instinct is to ask for the first montage and let the search happen inside it, which works once and then happens again from scratch the next time you ask for anything.

Prompt

Variants worth knowing:

• "Review and categorise first." Makes it its own job.
• "Then afterwards I can say..." Explains what it is for.
• "How many have you analysed?" A fair question to ask.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real library of five hundred-odd clips shared as a single folder link. The middle of this list is entirely about making a job of that size both affordable to run and recoverable when part of it fails.

1. Inspects the shared folder and works out how its media can be reached.
2. Downloads the archive and lists what is in it.
3. Measures the archive and builds a manifest without extracting everything.
4. Reads the analysis tools' instructions and prices before processing the library.
5. Checks there is workspace capacity for the extracted files.
6. Extracts the library locally for review.
7. Builds lightweight review proxies — small files that keep the audio.
8. Repairs the few proxies that failed and verifies the complete set.
9. Uploads one proxy first to confirm the format before batching.
10. Uploads the remaining five hundred, recording each source-to-URL mapping.
11. Runs speech transcription and visual shot surveys across all of them.
12. Monitors the workers without starting duplicate jobs.
13. Saves structured catalogues of what is said and what is shown.
14. Searches those catalogues when you ask for a theme.
15. Narrows the hits by requiring both the visual and the spoken evidence.

Step seven is what makes this possible at all. Analysing five hundred original files is slow and expensive; small proxies that keep the audio carry everything the catalogue needs, and the originals are only touched again when a clip is actually used.

Step nine is a habit worth copying to any batch job. One upload proves the format before five hundred of them fail the same way, which is the difference between a five-minute correction and a repeat of the whole pass.

What does the result look like?

A catalogue that knows what is in every clip — the speech, the shots, the subject — and a mapping back to your original files. Ask for a theme and you get a shortlist in moments, with the reason each clip was chosen.

The first montage takes as long as the cataloguing. Every one after that takes minutes.

When does this not work well?

The catalogue is only as good as what is visible and audible. Clips with no speech and little visual distinction are hard to categorise usefully, and a library of near-identical footage will not separate into themes however carefully it is reviewed.

• Silent, similar clips. There is little to categorise on.
• A folder that cannot be read. Some share links do not expose files.
• Very large libraries. Cataloguing has a real cost; ask first.
• Themes that need judgement. "The funny ones" is hard to index.
• Footage that keeps growing. New clips need adding to the catalogue.

How do you fix a result that came back wrong?

Say which clips should have been in the shortlist and were not. Vizard Agent keeps the catalogue it built, so a miss is a search that needs widening rather than five hundred videos that need watching all over again.

• "You missed some obvious ones." The search widened and re-run.
• "Half of those are not scrap." Both kinds of evidence required.
• "Categorise them differently." Re-indexed against your categories.
• "How many did you analyse?" A count and the method, on request.

How does Vizard Agent compare to doing it yourself?

By hand, every montage starts with scrubbing the same library again, and the knowledge gained never survives the session. Filenames and folders are the usual substitute for a catalogue, and they only record what somebody thought to write down at the time.

Common questions

Can it read a shared folder link? Often. Vizard Agent will say what it can and cannot reach.

How many videos can it handle? Hundreds. Vizard Agent prices the job before starting it.

Does it process my originals? No. Vizard Agent builds small proxies and keeps the originals aside.

Will it tell me how many it analysed? Yes, and by what method. Ask at any point.

Can I add new clips later? Yes. Vizard Agent catalogues the new ones and merges them in.

What does it index? What is spoken and what is shown in each clip, saved by Vizard Agent per file.

Can I define my own categories? Yes, and giving Vizard Agent two examples shapes the whole catalogue.

How precise are the shortlists? Vizard Agent can require both spoken and visual evidence to qualify.

Does it keep the catalogue? Yes. Vizard Agent keeps it in the project so later requests reuse it.

Can it make several montages at once? Yes. Once catalogued, each one is a query to Vizard Agent.

What if a clip is miscategorised? Say so and Vizard Agent corrects that entry.

Will the montage use the originals? Yes. Proxies are for searching, not for cutting.

Can it work from a hard drive instead? Yes, as long as Vizard Agent can reach the files from the project.

Why catalogue before asking for a video? Because the search is the expensive half, and doing it inside a montage request throws the answers away.