Running
Recognise running movement across a sequence of frames.
Multimodal video intelligence
NexuSec
Find the moment.
Understand the event.
Search video with an image or a description. Recognise actions across frames and bring the evidence into one clear operational picture.
Playback is unavailable. Open the video directly
An investigation starts with a description. Follow the search back to the footage.
01 / Search your video
A picture. A description.
A place to start.
Find people and vehicles
Use a reference image to search for similar appearances, or describe what you remember. Choose the cameras, recordings and dates that matter to your investigation.
- 01Choose your sources and time window
- 02Search with an image or written description
- 03Compare candidates in the original footage
Search results give operators candidates to investigate. They review the surrounding video before drawing a conclusion.
VIDEO SEARCH
Description search
Person in a grey hoodie with a large dark circular logo on the back and a white cap
Candidate appearances
Appearance candidate
Review the source.
See the full moment.
Camera context
Appearance time
Return to the original footage
VIDEO SEARCH
Description search
DPD delivery van
Candidate appearances
Return to the original footage
02 / Actions, interactions and events
See the action.
Understand the scene.
NexuSec interprets movement and behaviour across multiple frames. Action recognition adds context to what a camera sees.
Falls
Surface a possible fall for an operator to review.
Barrier climbing
Interpret movement over a barrier in suitable camera views.
Aggressive interaction
Analyse movement and interaction between people.
ACTION ANALYSIS
Source footage Interaction across framesReview event
Multi-frame analysisActivity timeline
03 / The operational picture
An event, with
the evidence attached.
Review the action alongside its location, timing and surrounding footage. Operators can return to the evidence and examine the event in context.
Supporting footage
Open the relevant scene and review the moments around the event.
Event reason
See the action or configured condition that prompted the event.
Camera
Know which camera or recording supplied the evidence.
Time
Return to the event’s position in the recording.
Location
See where the activity occurred in the camera’s view.
Multimodal AI / Contextual reasoning
Turn video and sound into a story
Combine scene detail, movement and audio in a readable event narrative. Use contextual AI reasoning to explore a timestamped sequence or ask questions about what happened, with the relevant moments ready to review.
EVENT CONTEXT
Original footageSound annotations
- 00:29Gunshot / gunfire
- 00:31Gasp
- 00:36Bang
Ask the multimodal model
“Tell the story of this recording. Combine what is visible with what is heard, and highlight the key moments.”
Video + audio Model response
Multimodal outputs
Scene narrative
The camera wearer moves through a dimly lit outdoor area beside a building. Lit windows and nearby lights provide patches of visibility as the viewpoint changes.
At 00:29, a sharp report is heard, labelled as possible gunfire. A gasp follows at 00:31, then another bang at 00:36. These sounds form a short sequence of events within the moving camera view.
The recording gives a reviewer the setting, event order and moments to inspect. The source of the sounds and their connection to anyone in view remain unclear.
Time-linked event sequence
- Outdoor approach
The moving camera shows a building and local lights. Visibility changes with the wearer’s viewpoint.
- Sharp report
Audio suggests possible gunfire. Its source is not established by the visible scene.
- Gasp
A brief vocal sound follows the report; the speaker and cause are unclear.
- Further bang
A second impact-like sound marks another moment to inspect alongside the video.
Questions about the recording
- What was happening before the first sound?
- The wearer was moving through an outdoor area beside a building. The changing viewpoint affects what is visible around the event.
- Does the recording confirm a shooting?
- The sound is consistent with possible gunfire, but that label alone does not confirm a shooting or identify who caused it.
- Which part should I review first?
- Start just before 00:29 and continue beyond 00:36 to compare the scene, the gasp and the later bang in sequence.
Text-to-speech alerts
Turn important notifications into clear spoken messages. Help operators notice priority events without watching every screen, with the supporting footage ready to review.
Distress alerts
Where audio is available, highlight possible signs of distress, such as shouting or calls for help. Link the alert to surrounding footage so operators can assess the situation and decide the response.
04 / Our training capability
Our models.
Your real conditions.
NWPRO develops and tests its own AI models for the events that matter to your operation.
We test against your camera views and real operating conditions, including the situations that can cause missed events or false alarms.
- Train and validate our own models
- Test against the selected camera views
- Review results with your operators
NWPRO / MODEL WORKFLOW
Labelled frames
Event models
Evaluate
- Detections
- Missed events
- False alarms
01 / Train02 / Validate03 / Test
Focus on events
Event detection does not require facial recognition or biometric identification. Operators review the evidence and decide the response.
NexuSec / Built around your operation
Your cameras.
A clearer understanding.
Bring us your footage, camera views and operational scenarios. We’ll show you where NexuSec can help.