00:00:00/Guide

How to make CCTV footage searchable

You make CCTV footage searchable by having a vision model read it: every frame gets a plain-English description, the descriptions are indexed, and from then on finding a moment is a text search. Setting that up means creating an account and pointing Svid at your footage, and not much else.

Searchable means something specific here. Not motion events, not a timeline you scrub, but typing “person in a red jacket near the till” and landing on the moment it happened. Svid stands for Searchable Video, and this page covers what turning your CCTV into that actually involves: what the system can ingest, what happens to the footage, where it runs and what it costs. If you're still weighing up whether search beats scrubbing, the search guide makes that case; this one is for getting set up.

What Svid can ingest

Four routes in, and you can mix them. An RTSP feed from your IP cameras or NVR is the usual choice for a permanent setup; you paste the address and the stream is captured on our side, with nothing to install. Uploaded files cover the footage you've already pulled off the recorder for an incident. A video URL works when the footage lives somewhere online (the title fills itself in), and a live phone camera at app.svid.ai/live.html covers a spot with no wiring at all, once you've signed in on the device. Live and recorded footage go through the same pipeline; the system only needs images out of the stream.

What happens to the footage

A vision model reads the frames as they arrive and writes a plain-English description of what's happening in each one: people, objects, actions, hazards. How often it looks is your choice of watch level. Precise takes a frame every second at 1080p, Standard every 5 seconds at 720p (the default), and Economy every 15 seconds at 480p, and all three are motion-gated, so frames that barely differ from the last kept one get no AI calls and a mostly-static camera costs much less. The descriptions are embedded for semantic search, and that index is what you search. There's no tagging and no training on your site; the cameras on the wall keep doing exactly what they did before.

  1. 01

    Create an account

    Sign up at app.svid.ai and answer four questions: business type, what to watch for, camera count, hours a day. The suggested checks and watch level are both editable before you pay.

  2. 02

    See the price, then pay

    The quote is based on your own usage for your cameras, minimum $9 a month, with the derivation on screen: frames per month, AI analyses, and your price. Checkout is a monthly Stripe subscription, and you land straight in the app.

  3. 03

    Connect your sources

    From the Add tab, paste RTSP addresses for the live cameras, drag in any exports, paste video URLs, or point a signed-in phone at the gap in your coverage from app.svid.ai/live.html.

  4. 04

    Search, and switch on checks if you want them

    Footage lands in a Library of nested folders (stores/london/entrance), and the search bar searches the folder you're in, subfolders included. If you also want alerts or written reports, checks are priced in the Store from your own usage.

What it costs

Search isn't an add-on; it's what the platform is. The monthly price is based on your own usage, floored at $9 a month, and it comes from your own profile: cameras, hours a day and watch level, discounted for motion gating. That's the whole model. Checks that raise alerts or write reports are priced separately in the Store, each shown as the extra it would add at your usage. The pricing section walks through the derivation.

Do I need to replace my cameras or recorder?
No. Svid reads what they already produce, RTSP feeds included. There's no proprietary hardware, and the recorder carries on recording exactly as before.
Can I make old exported footage searchable, or only live feeds?
Both. Uploads and live feeds go through the same pipeline, so an export from months ago gets described and indexed the same way as a live stream.
Does the footage have to leave my building?
In cloud mode, frames are sent to the vision model to be described, and the footage stays in your account and is never used beyond it. For sites that can't allow that, a fully on-prem appliance mode runs local models, so nothing leaves the network.
How long before I can search after connecting?
Footage is described as it arrives, and searches come back in seconds. There's no training period or tagging stage to get through first.

Point it at your footage

Create an account at app.svid.ai, answer four questions, and see the price with its working shown. Then add a clip and type your first search.

Get started

Related footage

Guide

How to search through hours of CCTV footage

Stop scrubbing. The fastest way to search hours of CCTV footage is to have AI describe every frame, then type what you're looking for in plain English.

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

RTSP

RTSP (Real Time Streaming Protocol) is the standard way IP cameras expose a live video stream so other software, like AI video analysis, can read it.

Glossary

Check (in Svid)

A check in Svid is a paid watcher over the core pipeline: alerting checks judge recent frames and raise alerts, report checks read whole videos and write documents.