Skip to content
GC SurgeDocsVideo Search
5 min read

Video Search

Video Search is GC Surge’s primary interface for discovering and reviewing alarm events. It provides a structured, filterable view of all events that have been processed and stored by the platform. Operators use Video Search for real-time triage, post-incident investigation, evidence review, and quality audits. Covers: What Video Search Does, Why It Matters, Understanding the Interface.

Video search + ZenMode.mp4

What Video Search Does

Video Search replaces unstructured video playback. Instead of rewinding through recordings to find a relevant moment, operators search by event attributes — site, detection type, time window — and see filtered, tagged event cards ready for review.

Why It Matters

Speed and accuracy in event review directly affect operational quality. Video Search gives operators precise control over which events they see, reducing the time from alarm notification to triage decision. For post-incident investigations, it provides a unified interface to find and share specific events without navigating multiple systems.

Understanding the Interface

Screenshot 2026-06-07 162346.png

Site and Device Selector

Video Search requires a site selection before events are displayed. Until a site is selected, the event list appears empty. This is by design — it prevents operators from accidentally reviewing events from the wrong site.

New operators should learn this prerequisite before their first shift — an empty event list almost always means a site has not been selected yet.

After selecting a site, you can narrow further to a specific device using the Select Device selector. For a specific investigation, scope to one site and one device first, then broaden only if you don’t find the event.

Filter Panel

Once a site is selected, the filter panel allows further narrowing:

  • Detection Category — limits results to specific event types such as motion, person detection, or vehicle detection.
  • Time Window — sets the investigation period. Default is the last 24 hours. Use a predefined window (last 5 minutes, 15 minutes, 30 minutes, 1 hour, 3 hours, 6 hours, 12 hours, 24 hours, 2 days, week, month, current shift, last shift) or pick a custom From / To range with minute precision. The picker covers the full history of stored events in your account. For long investigations, search in 24-hour chunks rather than weeks at once; large ranges return results faster when split.
  • Alarm Type — filters by NOVA99x classification: Real, False, or All. Default is All so you don’t miss anything. Start investigations at All to catch any misclassified event, then switch to Real once you’ve narrowed by time and camera to find the actual incident faster. False is mainly useful for QA and noisy-camera tuning.
  • Overlay — toggles visual layers on event media. Options are Heatmap, Mask, and History Mask.

Event Card List

A Total Events counter at the top right of the page shows the number of events matching the current filters at a glance — check it first to gauge the volume before opening individual cards. Events matching the current filters appear as cards in reverse chronological order (most recent first). Each card shows a thumbnail or snapshot, the site and camera name, the event type, and the timestamp.

Event Detail Panel

Clicking an event card opens the event detail panel:

  • Breadcrumb — shows the full path: Customer > Site > Camera > Sensor, with timestamp and classification badge (REAL ALARM or FALSE ALARM).
  • Video panels — four clips: Pre Alarm, Alarm, Post Alarm, and Preview.
  • AI overlay — the Preview panel includes a bounding box highlighting the detected subject.
  • Marking — use the thumbs up (W) or thumbs down (S) buttons to classify the event.
  • Export — copy or download the Alarm clip using the icons on the panel.
  • Close — use the X button to return to the event list.

How It Works

Use this workflow for both real-time triage and post-incident investigation:

  1. Navigate to Video Search from the sidebar.
  2. Select the site, then optionally select a device to narrow results further.
  3. Optionally apply filters: Detection Category, Time Window, Alarm Type, or Overlay to adjust visual layers on event media.
  4. Review the event card list — most recent events appear first.
  5. Click a card to open the full event detail view.
  6. Review the event, mark it using thumbs up (W) or thumbs down (S). Close the panel with X to return to the event list.
Screenshot 2026-06-07 162253.png
12_video-search-grid.png
18_video-search-quad.png

Filter Strategy

Filters are powerful but require the right approach to get useful results:

  • Start with no filters. Select your site and a time window, and review what events are there before narrowing down. Jumping straight to a specific filter can cause you to miss events that don’t match the filter but are still relevant.
  • Don’t filter by detection type during an investigation. If you’re looking into an incident, you need to see the full sequence — a motion event may have happened before the person was detected. Filtering too early hides that context.
  • Agree on a time window before reviewing as a team. If one operator searches the last hour and another searches the last two hours, they’ll see different events and may reach different conclusions about the same incident.

Best Practices

  • Review events in structured windows, not continuously. Define shift review periods — for example, review the last 4 hours every 2 hours — and stick to them. Continuous monitoring without structure reduces focus and consistency over time.
  • Use Video Search and ZenMode together. Video Search is your investigation tool. ZenMode is your active triage tool. During high-volume periods use ZenMode to reduce cognitive load. During post-incident review or investigation, switch to Video Search for full filter control.
  • Record what you find in shift handoff notes. If you spot a pattern — repeated events from the same camera at unusual hours — document it before your shift ends so the next team knows what to watch for.