00:00:00/Glossary
Motion gating
Motion gating is a filtering step in AI video analysis: once frames have been extracted from footage, any frame that barely differs from the last kept frame is skipped before the AI runs. The model only ever sees frames where something changed, so an hour of empty corridor costs almost nothing to analyse while an hour of busy shop floor costs the full rate.
How it works
Frames are sampled from the video at a fixed interval, then each candidate is compared with the last frame that was kept. If nothing meaningful has changed (the same shelves, the same parked van, the same pool of light on the floor), the frame is dropped and no AI call is made for it. The moment someone walks in or a door opens, the difference crosses the threshold and frames start flowing to the model again.
Motion gating vs motion detection
They sound alike but sit in different places. Motion detection is an alerting feature: a camera or recorder notices movement and flags an event or starts recording. Motion gating is a cost control inside an analysis pipeline: it decides which frames deserve AI attention at all. A gated pipeline still sees everything that happens; it just declines to pay for describing nothing.
Why it changes the economics of watching
Real CCTV is static most of the time. An overnight corridor, a quiet loading bay, a stockroom between deliveries: the honest description of most frames is that they look exactly like the previous one. Without gating, an AI would be paid to write that description thousands of times a day. With it, mostly-static footage costs a fraction of busy footage. In Svid every watch level is motion-gated, and the price quoted at signup already assumes a conservative saving from it, so the skipped frames show up as money not spent rather than a discount you have to chase.
Related footage
Glossary
Searchable video
Searchable video is footage an AI has described frame by frame, so you can query it in plain English. It's also what Svid stands for: Searchable Video.
Glossary
Vision-language model (VLM)
A vision-language model (VLM) is an AI model that looks at an image and describes what it sees in words, the engine that makes video footage searchable.