Connect your cameras
Point Relay at the streams you already record. No new hardware and nothing to install on the cameras.
Dock 3 · Camera 05
Live
A product of
Every camera,
Every camera,
understood.
Relay AI reads the cameras you already have, frame by frame. It sees objects, understands how they relate, and flags every issue with the evidence. In real time, on your own servers.
No new cameras. No retraining. No facial recognition.
Every square is a camera feed. Lit squares are being read by Relay: detecting, tracking and relating objects, frame by frame. Amber means a rule was broken.
The pattern is Conway’s Game of Life, mirrored, one generation every 0.7 seconds.
How it works
Connect any camera.
Get every issue flagged.
Say what to look for
Describe it in plain words, like “every bottle needs a cap”. No labeling, no retraining.
Every bottle needs a cap
matching
Line 2 · Camera 04 · 0 retraining
Get the flag, with evidence
The frame, the boxes and the timestamp. Clear it, assign it or export it.
Missing cap · Line 2
+1 flag04:12:33
Evidence saved, frame 3,214
Live demos
Every frame checked.
Every issue flagged.
Plant 1 / Line 2 · Camera 04Live50 fps04:12:33
Aisle 7Store 12ReceivingLine 1Truck Bay
One screen for every camera. Watch detections and relations live, and clear flags as they come in, with the frame that proves each one.
6 objects2 classesframe 0000
Name it and Relay finds it. Type “bottle, bottle cap” and every match is boxed and scored, live, with no training run.
hairnet #12onperson #1frame 0000
Not just objects. How they relate: a hairnet on a worker, a cap on a bottle, a pallet on the dock.
5 tracksmasks onframe 0000
Every object keeps its ID from frame to frame, with a mask and a motion trail, so Relay knows what moved, where and when.
PPE missing · Site 2Camera 01 · evidence saved
3 people2 hard hatsframe 0000
Write the rule in plain words. Relay checks every frame and raises the flag, with the evidence, wherever your team works.
Analytics
Visibility into every
line, dock and shelf.
Every flag is attributable: to a site, a camera, a rule and a moment in time, with the frame that proves it. See where problems start, and whether they stop.
Cameras connected0
Frames checked today0
Objects tracked0
Flags raised0
Flags by rule
Last 14 days, by day
Leaderboard
Cameras ranked by flags
| # | Camera | Flags | Cleared |
|---|
Sample data for illustration
Private by design
Runs on your servers.
Video stays with you.
Relay runs on your own hardware or private cloud, next to the cameras. It looks for hairnets, hard hats and pallets, never for who people are.
Your network5 cameras · 1 Relay node
hairnetglovesFaces are never processed
PPE missingBeforeFlagAfter
Export CSV / JSONSend to ERP / WMS
Use cases
On the line. On the dock.
On the shelf. On site.
Teams hand Relay the questions they used to answer by walking the floor, then see exactly what changed.
“Is there a cap on every bottle, on every line, all shift?”
Relay: Every bottle on Line 2 is capped except one, at 04:12:33. Flag raised, evidence saved.
Line 2 · Camera 04Quality check
Manufacturing0
new cameras needed. Relay reads the streams you already record.
SetupFood & beverage
“Is everyone at Station 4 wearing a hairnet and gloves?”
Station 4 · Camera 01Hygiene check
Food safety0
retraining to add a rule. Describe it in plain words.
Rules“Is anything blocking the dock door, and where did that pallet go?”
Relay: The door is clear. Pallets #3 and #4 are on the dock, and forklift #1 is moving beside them.
Dock 3 · Camera 05Logistics
24/7
every camera, every shift. Every frame checked, not sampled.
CoverageLogistics
Retail
Construction
FAQs
Frequently Asked
Questions
What does Relay AI do?
Relay is vision AI for operations. It reads your camera feeds frame by frame, detects objects, understands how they relate (a cap on a bottle, a hairnet on a worker, a pallet on the dock) and flags every issue with the evidence, in real time.
Do we need new cameras?
No. Relay works with the cameras you already have and the streams you already record.
How do we tell it what to look for?
In plain words. Describe the object or the rule, for example “every bottle needs a cap” or “everyone on site wears a hard hat”. There is no labeling and no retraining.
Does Relay recognize faces?
No. Relay looks for objects and how they relate, like hairnets, gloves, hard hats and hi-vis vests. It does not identify faces or people.
Where does it run?
On your own servers or private cloud, next to the cameras. Video is processed where it is recorded.
What comes with a flag?
The frame, the boxes, the rule that was broken, the camera and the time. Flags can be cleared, assigned or exported as CSV or JSON to your ERP or WMS.
Which industries is it for?
Anywhere cameras watch work: manufacturing and food lines, warehouses and loading docks, retail shelves and construction sites.
Is the footage on this page real?
The videos are licensed stock footage processed by Relay for illustration. Dashboard figures are sample data. Nothing on this page describes a specific customer.
Who builds Relay AI, and how do we start?
Relay AI is built by Aether AI. Book a demo through the Aether AI contact page and the team will walk you through it on your own footage.