Person Attributes
Age range, gender, clothing, bags and accessories — structured attributes for every person in frame, in real time.
On-prem AI Box
Miss nothing.
OsonVision is a compact AI Box that turns ordinary cameras into a real-time intelligence system — person attributes, vehicle recognition and action detection, with VMS, MLOps and VLM built in.
PersonVehicleAction8 channels / box
LIVE · CH-01OsonVision AI Box · 62 FPS
Person0.97
Male · Adult · Backpack
Person0.95
Female · Adult · Handbag
Vehicle0.94
Sedan · White · 01 A 123 AB
Action: Walking· Loitering 0.08
Cheap & Competitive
An 8-channel AI Box for $1,750 — a fraction of rack-server analytics stacks, with no per-camera cloud fees.
Easy to Use
Unbox, connect cameras, and you are analysing video the same day. A clean console anyone on the team can run.
Private by Design
Every frame is processed on the box. Footage and models stay inside your network — private by default.
The hardware
One silent box. Eight cameras. Zero cloud.
Fan-quiet and palm-sized, the OsonVision AI Box runs every model on the edge. Drop it on a shelf, wire in your cameras, and keep all footage on-site.
- 8 channelsConcurrent camera streams per box
- Edge inferenceReal-time detection, fully local
- PrivateNo footage leaves your network
- Plug & playRTSP / ONVIF, running the same day
One platform
Six capabilities that usually need three vendors, unified on a single edge device.
Vehicle Recognition
Type, colour, make and licence plate. Track vehicles across channels for traffic, parking and access control.
Action Detection
Walking, running, falling, loitering and intrusion. Trigger alerts the moment behaviour crosses your rules.
VMS Built-in
A full video management system on the box — live view, recording and multi-channel playback. No extra server.
On-box MLOps
Collect, label, retrain and roll out custom models directly on the AI Box. Your data never leaves the premises.
VLM Search
Ask in plain language — "white sedan near gate after 6pm" — and the Vision-Language model finds the clip.
How it works
From camera to insight in three steps
Connect your cameras
Plug RTSP/ONVIF streams into the box — up to 8 channels per unit. No cloud account required.
Detect on the edge
The box runs detection, attributes and actions locally at real-time frame rates, fully on-premise.
Search & integrate
Review events, search with natural language, and push results to your systems over a simple API.
Channels / box
Full 8-ch set
To deploy
On-premise
Put an AI Box on your network this week
Detection, VMS, MLOps and VLM in one silent box. See it running on your own cameras.