Deploy FireVision AI
Directly On Your Local Hardware
The standalone, offline-operable computer vision appliance for organisations with strict video privacy, internal data compliance, and air-gapped security mandates. Ingest camera streams locally, run neural networks in-memory, and retain all video files entirely on your own local storage.
How the 7-Day Local Trial Works
Evaluate on-premise AI computer vision on your own PC with zero sales friction. Everything needed for setup and licensing is self-contained directly inside the application.
Download & Install
Download the official verified Windows installer directly from this page. Run the setup on your local computer or dedicated surveillance server.
Launch Local Application
Launch the desktop application. It boots instantly into the native surveillance workstation environment without requiring any online account login.
Request 7-Day Licence via Email
Open the in-app 🔑 Licence menu (or select File ➔ Licence Manager) to copy your machine's unique Hardware ID (HWID). Email your Machine ID with your organisation details to info@firevision.uk to receive your unique 7-day lic.txt licence file.
In-App Help & RTSP Setup
Press F1 or select Help ➔ User Manual inside the application to access full RTSP stream setup guides, network diagnostics, and hardware optimisation manuals packaged right in the app.
Ready to evaluate on-premise AI fire detection?
Download the verified standalone setup package and begin monitoring in minutes.
Why Critical Infrastructure Chooses FireVision Local
Engineered specifically for local authorities, defence contractors, chemical facilities, healthcare estates, and high-security sites where video feeds cannot legally or policy-wise leave the local network.
100% Local LAN Video Processing
RTSP streams connect directly from your IP cameras to the local workstation over standard Ethernet. Frames are decoded and tensor-processed in volatile workstation memory with 0 bytes ever uploaded to external cloud servers.
- Zero WAN bandwidth congestion
- Zero external inbound port forwarding
- Complete immunity from cloud service outages
Total Storage & Retention Control
Recordings and incident snapshots are written directly to your chosen local disk drive or local private cloud array. Circular in-memory ring buffers capture pre-alarm frames and encode clips only when an incident occurs, saving over 90% disk space.
- Automated 12-hour background cleaner
- One-click instant manual disk purge
- Custom per-camera retention (1–365 days)
Air-Gapped Cryptographic Security
FireVision Local operates in fully disconnected, air-gapped security environments. Licences are verified cryptographically offline using your computer's unique Hardware ID (HWID).
- Local RBAC (Site Administrator, Lead Operator, Viewer)
- Immutable user-attributed audit trail
- Lock Terminal Standby mode for control rooms
Assign Dedicated AI Architectures Per Camera
Unlike generic CCTV analytics, FireVision Local V2 features an in-memory multi-model caching engine. Assign different specialised detection architectures to different cameras simultaneously with zero reload latency.
Specialised Fire & Smoke Detection Architecture
Trained specifically on wildland-urban interface (WUI) datasets, indoor warehouse fires, chimney plumes, and waste-recycling flare-ups. Includes our proprietary Stage 2 temporal plume expansion verification to eliminate false positives from steam, mist, and headlights.
Inference Device: CUDA / DirectML Singleton (In-Memory)
Target Classes: ['fire', 'smoke']
Temporal Confirmation Window: 3 consecutive frames (0.6s)
Cooldown Protection: 45s Smart Continuity Protection
Pre-Alarm Buffer: 5.0s Circular RAM Ring Buffer
Hardware Profiler & Workstation Sizing Guide
FireVision Local V2 dynamically detects your workstation's GPU via DirectX 12 WMI and attaches to the fastest available acceleration engine without manual driver configuration.
Select Hardware Backend
NVIDIA CUDA (Dedicated GPU)
GeForce RTX 30/40 Series, Quadro, RTX A-Series
DirectML / DirectX 12 (Universal)
Intel Iris Xe, Intel Arc, AMD Radeon, Ryzen APUs
AVX2 Vectorised Multi-Threaded CPU
Intel Core i5/i7, AMD Ryzen 5/7 (No dedicated GPU)
Workstation Sizing Calculator
Adjust camera count and retention to see correlated compute & storage guidance.
Optimal balance for commercial facilities, timber yards, and waste transfer stations monitoring 3–8 cameras.
These sizing metrics are recommendations and engineering guidance. Physical hardware provisioning, operating system maintenance, and power environments are the user's responsibility. The FireVision engineering team can support and advise on bespoke new hardware builds.
Naturally, higher compute power, larger RAM, and fast NVMe/RAID storage result in faster processing throughput, lower inference latency, and quicker incident detection. All performance is also influenced by local network switch throughput and camera RTSP bitrates.
Workstation Sizing Specifications
The cards below reflect standard deployment baselines. Adjust the calculator above to highlight your matching tier.
1–2 Camera Feeds
Small office, remote test gate, workshop
- CPU: Intel Core i5 (8th Gen+) or AMD Ryzen 5
- RAM: 8 GB DDR4 Memory
- GPU: Intel Iris Xe / DirectML / CPU AVX2
- Storage: 256 GB SSD (50 GB free for event clips)
- Network: 100 Mbps Ethernet
3–8 Camera Feeds
Warehouses, distribution hubs, civic campuses
- CPU: Intel Core i7 (11th Gen+) or AMD Ryzen 7
- RAM: 16 GB DDR4/DDR5 Dual Channel
- GPU: NVIDIA RTX 3050/3060 (6-8 GB VRAM)
- Storage: 1 TB WD Purple Surveillance HDD
- Network: 1 Gbps Gigabit LAN
9–16+ Camera Grid
24/7 Security Operations Centres (SOC), industrial plants
- CPU: Intel Core i9 / AMD Ryzen 9 / Xeon
- RAM: 32 GB – 64 GB High-Bandwidth DDR5
- GPU: NVIDIA RTX 4070 / RTX A4000 (12-16 GB)
- Storage: 4 TB – 8 TB Enterprise RAID Array
- Network: Dual Gigabit LAN (Dedicated CCTV VLAN)
FireVision Local V2 vs. Cloud Platform vs. Traditional NVRs
| Feature / Dimension | 🔥 FireVision Local V2 (Standalone) | 🌐 FireVision Cloud (Online) | 📼 Traditional Hardware NVRs |
|---|---|---|---|
| Video Feed Route | 100% Local LAN (0 Cloud Uploads) | Encrypted Cloud Streaming | Local Disk DVR (No AI) |
| AI Detection Engines | In-Memory Multi-Model Singleton (CUDA / DirectML / CPU) | Cloud GPU Inference Pipeline | Basic Pixel Motion Only |
| Air-Gapped Operation | Full (Offline Cryptographic lic.txt) | Requires Active Internet | Full Offline |
| Video Storage & Purging | Flexible Local Storage / Local Private Cloud (Disk/NAS/RAID) | Cloud Object Storage | Continuous 24/7 Raw Bloat |
| False Positive Reduction | Stage 2 Temporal Plume Verification (98% reduction) | Dual-Stage Cloud Verification | High False Alarms (Spider/Rain) |
| Access Control & RBAC | Site Administrator, Lead Operator, Viewer + Lock Terminal | Multi-User Cloud Authentication | Single Master Admin Password |
Frequently Asked Questions
lic.txt) is bound to your machine's Hardware ID (HWID) and verified completely offline.Take Full Control of Your AI Surveillance Data
Download the 7-day standalone trial, install on your local workstation, and request your cryptographic licence right inside the application.