AI camera box

One box per site. Three levels of AI.

Every Prism site starts with the Prism Box. It records every camera, gives you live view, runs the gate lines and spots possible incidents with a small AI model on the box. Higher tiers add bigger AI models — either on Prism’s own model server, or on a more powerful box at your site.

THE NODE
The Prism Box

Lives at your gym. Records, gives live view, runs the gate lines and flags possible incidents with a compact on-box model.

TIER 2 ADDS
Prism’s model server

A much larger model re-checks each gate incident clip. If the server can’t be reached, the box carries on by itself.

TIER 3 ADDS
A Pro on-site box

Intel Arc or NVIDIA graphics run the large model live, in the building, for busier sites with more cameras.

The tiers

Pick the level your doors need.

Start with Essentials. Move up when a gate gets busy or the site grows — the box you already have stays the heart of it.

01AVAILABLE NOW

Essentials

Recording, live view and gate rules on one small box at your site.

HARDWARE

Prism Box — Intel N100/N150 mini PC, 16 GB RAM, 512 GB SSD, Intel graphics.

  • 24/7 recording of every camera, with live view in the app
  • Gate lines drawn per lane: unscanned entry, tailgating, and wrong-way entry through the exit
  • Overhead people counting, and people-inside-now on the dashboard
  • After-hours presence alerts
  • Alerts to Telegram
  • Every rule is a per-site setting

Uses a compact detector on the box. Good for straightforward gates; busy scenes produce some false alarms, which staff review in the app. Supports about 2–4 cameras.

02PILOT · ROLLING OUT

Verified

The same box, with every gate incident re-checked by a much larger model.

HARDWARE

The same Prism Box, plus Prism’s own model server (a monthly service).

  • Everything in Essentials
  • Every gate incident clip is re-counted by a much larger, state-of-the-art detection model (RF-DETR class) with a proper multi-person tracker
  • Who crossed which lane, and which way — checked against the door scans, for far fewer false alarms
  • Optional AI review of incidents: reception left unattended, a customer waiting, staff on the phone, incident summaries
  • AI review only ever sees still images, with every face blurred on Prism’s own server first — never live video

If the model server can’t be reached, the site keeps working exactly as Essentials.

03ON REQUEST · 4+ CAMERAS

Pro on-site

A more powerful box that runs the large model live, in the building.

HARDWARE

A more powerful on-site AI box with Intel Arc graphics or an NVIDIA graphics card.

  • Everything in Essentials, with the large model running live on site
  • Accurate counting and tracking through crowds
  • The large face-recognition model
  • Matching people across cameras

For sites with four or more cameras. The AI work happens in the building — nothing leaves it.

Specs

Side by side.

 01Essentials02Verified03Pro on-site
AvailabilityAvailable nowPilot, rolling outOn request
On-site hardwarePrism Box: Intel N100/N150, 16 GB RAM, 512 GB SSD, Intel graphicsSame Prism BoxPro AI box: Intel Arc or NVIDIA graphics
Where the AI runsOn the box (compact detector)On the box, plus Prism’s model serverOn the Pro box, live
Gate-incident checkCompact detectorLarge model (RF-DETR class) + multi-person trackerLarge model, live
Checked against door scansYesYes, per lane and directionYes
Recording & live view24/7 · in the app24/7 · in the app24/7 · in the app
People countingOverhead counterOverhead counterCounting and tracking through crowds
Large face-recognition model——Yes
Matching people across cameras——Yes
AI incident review—Optional · face-blurred stills only—
CamerasAbout 2–4About 2–4 (same box)4 or more
If the model server is unreachableNot usedKeeps working as EssentialsNot needed
Measured, not promised

Why the big model doesn’t run live on the Prism Box.

On the Prism Box, the compact detector takes about 15 ms per check. The state-of-the-art model takes about 186 ms on the same box — too slow to run live there. That is why it runs on Prism’s model server (Verified) or on a Pro box (Pro on-site). In a test on real gate footage, the larger model found the person far more often:

TIME PER CHECK ON THE PRISM BOX
~15 ms
compact detector
~186 ms
state-of-the-art model
PERSON FOUND · TURNSTILE
12/15
compact detector
15/15
larger model
PERSON FOUND · OVERHEAD COUNTER
9/15
compact detector
14/15
larger model

Frames with the person correctly found, out of 15, from one test on real gate footage.

Privacy

Your footage stays yours.

Video stays put

Video and recordings stay on the box at your site and on Prism’s own infrastructure.

Faces blurred first

Any AI review by an outside model only ever sees still images, with every face blurred — never live video.

Pro keeps it in the building

On a Pro on-site box the large models run at your gym. Nothing leaves the building for AI.

Book a walkthroughRead it in the manualInstall the box