ShotIndex
You know you have the shot. ShotIndex knows where.
ShotIndex indexes a complete video archive — NAS and loose drives — and makes it searchable by what is actually in the frame. Search the way you remember a shot ('drone, polder, sunrise'), jump to the right segment and deliver into DaVinci Resolve with proxies and the correct LUT baked in.

55 TB
Own archive scanned
3,493
Clips indexed
10,989
Local embeddings
203
Tests
What it does
Measure before guessing
ffprobe, optical flow, histograms and person detection capture camera movement, exposure, sharpness and shot size — free, in seconds, without AI.
AI descriptions per segment
Segments of 8–12 seconds get Dutch descriptions including spoken dialogue — interviews become findable by quote.
Three search modes
Broad, precise or semantic via local text embeddings — always with an explanation line showing why something matched.
Map & drone flight paths
Clips clustered by location, drone paths coloured by altitude, and manual pinning per folder — because only 3 out of 2,816 clips carried usable GPS.
Client portal
Clients view only their own material online — published per share, signed in with a 6-digit code or magic link. The separation is physical, not a WHERE clause.
Straight into your timeline
FCPXML export to Resolve or Final Cut with linked 960p proxies and the log→Rec709 LUT baked in; ProRes RAW gets white balance and an S-curve instead of a wrong log LUT.
How we built it
Node.js 22 as the engine, Electron as the window, one SQLite file as the index — local-first, copyable as a backup, with almost zero dependencies. The same engine doubles as a CLI.
AI runs on the user's own key — proxying through our server would make us a GDPR processor. Upfront cost estimates, a hard budget ceiling, resumable queues; embeddings stay local so a model switch never bills twice.
Originals are never touched and never deleted: missing files get flagged, not wiped, and clip identity hangs on a fingerprint rather than a path — renaming folders costs no analysis work.
The client portal runs on Cloudflare in the EU and publishes only what you release: a few gigabytes per share instead of a 55-terabyte archive, with a cross-leakage test over real HTTP.
Why we built it
In our own production house, over 200 terabytes of shoot days were slowly becoming unfindable. The first scan produced the hard numbers: 104,114 folders, 82 clients, 296 projects — and order numbers covered only 10% of projects, so the archive had to be unlocked by content.
So we built the footage library we were missing ourselves. When a client asked to view his footage online, the client portal followed — now a standalone product in private beta at shotindex.online.


Need something like this in your company?