English Book a Demo

AI Video Analytics

MultiFusion Analytics Worker
AI video analytics on-premise

MultiFusion Analytics Worker runs the AI on your own hardware and ships with its base models — no extra model licence and no cloud. It detects and tracks people and vehicles, reads number plates and powers image search.

Available nowPricing: contact us
Book a Demo Contact Sales
MultiFusion Server app — analytics channels per camera with model, worker and status (test data) test data

Film · 2:36

MultiFusion AI — 17 agents, one Analytics Worker

Music: "Voltaic" by Kevin MacLeod (incompetech.com), CC BY 4.0. Fictional plates and illustrative records.

Technology partners

Hardware the Analytics Worker runs on

Apple silicon, CPU and NVIDIA GPUs on Windows today. Packs for NVIDIA on Linux, Intel and AMD are being added — each line says where it stands.

NVIDIA logo

Member · NVIDIA Partner Network

GPU acceleration for the Analytics Worker — on Windows today, Linux next.

  • Available Analytics Worker on NVIDIA GPUs on Windows — CUDA 12 and TensorRT 10 (measured on a GeForce RTX 5050: detector in 11.5 ms, 87 fps)
  • Available Driver: Windows 528.33 or later
  • Coming soon Analytics Worker on NVIDIA GPUs on Linux (driver 525.60.13 or later)
Intel logo

Member · Intel Partner Alliance

Intel Core and Xeon servers today; OpenVINO acceleration next.

  • Available MultiFusion Server on Intel Core and Xeon, Windows and Linux
  • Coming soon Analytics Worker with OpenVINO on Intel CPUs and iGPUs
Microsoft logo

Member · Microsoft AI Cloud Partner Program · Microsoft for Startups Founders Hub

ONNX Runtime inside every worker; Windows from desktop to server room.

  • Available ONNX Runtime under every Analytics Worker runtime
  • Available MultiFusion Server on Windows 10, 11 and Server 2016+, 64-bit
AMD logo

Hardware support

AMD Ryzen and EPYC machines run MultiFusion Server today.

  • Available MultiFusion Server on AMD Ryzen and EPYC, Windows and Linux
  • Coming soon Analytics Worker CPU runtime on Ryzen and EPYC

NVIDIA, CUDA, TensorRT, Intel, Xeon, OpenVINO, Microsoft, Windows, AMD, Ryzen and EPYC are trademarks of their respective owners, and their logos belong to them. Freaquer is a member of the NVIDIA Partner Network, the Intel Partner Alliance and the Microsoft AI Cloud Partner Program; membership does not mean that these companies endorse MultiFusion. AMD is listed for hardware support only.

What Analytics Worker does

Bundled base models: Sentinel (detection), Scout and Lexa (plate reading), Recall (image search)Available
Detection, tracking, number-plate reads and “find similar” image searchAvailable
Signed model packs with canary, promote and rollbackAvailable
Built-in benchmark that reports the worker's capacityAvailable
Low-disk mode and a GPU temperature guardAvailable
Apple silicon and CPU; setup wizard in English and HindiAvailable
Analytics channels: intrusion, line crossing, loitering and crowd rulesAvailable

Sentinel: 97 fps on Apple M1 Max (Core ML) — MEASURED. Measured figures carry the MEASURED label; we publish no unmeasured capacity numbers.

Where it is used

Frequently asked questions

Do the AI models cost extra?

No. The base models ship with the worker.

Does any video leave my site for AI?

No. Analytics run on the worker you install.

Which hardware is supported?

Apple silicon and CPU today. Ask us about other accelerators.