The B-Roll Finder
Used by 40+ teams
2 skills from Editor OS, free. broll finds a shot that already exists and cuts it to the line it serves. It searches by what's said inside a clip, reads the captions to place the window, and strips the bars and the broadcaster's bug. motion-design builds the clips around it: captions, figures and CTAs.
Free. No account, no card.
What's Inside
- Search by the words spoken in a clip. A borrowed shot is chosen for a moment, and a moment is a line of dialogue. A video title describes the whole upload. broll reads a clip's own time-coded captions and hands back the window, which is the difference between a shot that lands on the word that earns it and one 2 seconds off.
- The bars and the bug come off before you see it. Half the recognisable beats on YouTube are 4:3 or 2.39:1 material re-uploaded inside a 16:9 box, so a naive fit scales the black up and ships it. The fetch measures the bars over the window it will actually use, takes them off, and punches out the broadcaster's corner logo.
- It refuses the stock plate. A generic clip of a server room or hands on a keyboard is the thing this register exists to avoid. broll says no and names what to look for instead, which is the whole difference between a borrowed shot and filler.
- The licence goes in a ledger, so you don't have to remember it. The ledger holds each shot's source and what it's allowed to do. A paid build won't start until it's filled in.
What You Get
- 2 skills in one zip: broll, which finds the shot, and motion-design, which builds the clips around it
- source.py, with the 3 commands broll runs: transcript, find and fetch
- The motion-design playbook for captions, figures and CTAs, plus when to pick Remotion over HyperFrames
- The search method written out, with the failure modes that cost real runs
- The licence ledger format, and the gate that reads it
- Works in Claude Code, Cowork or the Claude app
The borrowed shot is the one register in a video whose picture belongs to somebody else. A film beat, a meme everybody has already seen, a piece of archive, a real screen. Cut whole, 3 to 6 seconds, muted, under a read that keeps running.
It exists because the cheapest recognition there is arrives already recognised. A drawn frame earns recognition by rendering the real logo or the real screen. A shot everybody has watched walks in with it, and when a line lands on a frustration the viewer has lived, 3 seconds of the real thing says it faster and warmer than any diagram.
The hard part was never finding a video. It is finding the second inside it. Titles are written by whoever uploaded the clip and describe the whole thing; the transcript describes the second you're after. So the loop is 3 commands and the middle one is the one everybody skips: read the captions, place the window, then fetch it.
The zip carries a second skill, motion-design, because a borrowed shot rarely stands alone. It plans and builds the clips around it, from a caption on one line to the CTA at the end.
Both are pieces of Editor OS, and they're free so you can run them today. If they save you an hour on your first video, the rest of the system is the same idea applied to the cut, the layouts and the handoff.
How it works
Import
Add the template to your ClickUp workspace in one click.
Configure
Follow the setup guide to match your team's workflow.
Run
Start capturing, processing, and completing work in one place.
Common questions
Do I need a paid ClickUp plan?
How long does setup take?
Can I customize it for my team?
What if it doesn't work for us?
Keep your work in one system
Get a proven framework and set it up in under an hour.