Cutting-edge research
for a safer world.
We believe our research will help protect people and property better. Our mission: to build cutting-edge AI models for safety and security.
Œil 1.1
Security vision, in real time.
The world's best vision AI for security, in real time.




How the model works
- 01The instruction
You write in one sentence what Œil should watch for.
- 02The analysis
Œil looks at the camera images, in real time.
- 03The check
It puts the scene in context and rules out false alarms.
- 04The alert
Whatever matches the instruction becomes an alert, described in one sentence.
Œil 1.1 sees more.
And makes fewer mistakes.
Œil 1.1 versus cameras' built-in AI
Intrusions detected
Higher is better100 %
82 %
80 %
79 %
77 %
75 %
Fires detected
Higher is better85 %
38 %
35 %
33 %
30 %
27 %
False alarms in one night across 10,000 cameras
Lower is better27
3,412
3,530
3,667
3,890
4,105
Night detection
Higher is better97 %
70 %
68 %
66 %
61 %
59 %
Plan 1.1
It reads floor plans like an architect.
The world's best floor plan reading model. A plan becomes a 3D site, cameras included.
Voice 1.1
A natural voice that deters.
It describes the intruder, their clothes and where they are, out loud.
- Message 1
- Message 2
How the model works
- 01The risk
Œil spots the intruder in the image and raises the alert.
- 02The description
It describes what is happening: the intruder's clothes, where they are.
- 03The speaker
The voice picks the speaker closest to the intruder.
- 04The deterrence
It speaks to them in a natural voice, on your machine, without the cloud.
Sees everything. Stays with you.
Book a demo