00:00:00/Glossary
Searchable video
Searchable video is video that has been read by an AI so that every moment can be found with a text query, the way you'd search a document instead of rereading it. Rather than scrubbing a timeline, you type “person in a red jacket near the till” and jump to the frame. Svid is short for Searchable Video; it's the idea the whole platform is built on.
How video becomes searchable
A vision model looks at frames pulled from the video and writes a plain-English description of each one: the people in it, what they're wearing, what they're doing, what objects and hazards are present. Those descriptions are indexed. A search is then a text query over the descriptions, which is why results come back in seconds regardless of how many hours of footage sit behind them.
How it differs from motion detection
Traditional CCTV search tools index metadata (camera, timestamp, motion events), so they can tell you when something moved but not what it was. Searchable video indexes meaning. That's the difference between “show me motion on camera 4 last Tuesday” and “show me the man in the high-vis vest who left through the fire exit”.
What it enables
Once footage is searchable, watching stops being the bottleneck. Review becomes a query, incidents can be reconstructed across cameras, and the same descriptions can be monitored continuously to raise alerts the moment something matters. In Svid this continuous checking is done by checks, paid watchers over the same pipeline.
Related footage