One server. Your room. Online or fully offline.
Tessavox runs on a single server inside your perimeter and connects the field to the command center. Every number on this page is marked as measured or estimated.
Three single-server configurations on one platform
Capacity is set by CPU and memory; the GPU keeps many simultaneous rooms smooth and runs AI translation. All three share one hardware platform — add a GPU to Entry and you have Standard; upgrade CPU and memory and you have High-performance.
| EntryNo dedicated GPU | StandardMeasured configuration | High-performanceSame GPU, larger CPU and memory | |
|---|---|---|---|
| GPU | Not required (integrated graphics) | NVIDIA GeForce RTX 3090, 24 GB | NVIDIA GeForce RTX 3090, 24 GB |
| CPU | AMD Ryzen 9 7900X, 12 cores | AMD Ryzen 9 7900X, 12 cores | AMD Ryzen 9 9950X, 16 cores |
| Memory | 64 GB DDR5 | 64 GB DDR5 | 128 GB DDR5 |
| Network | 1 GbE | 1 GbE (2.5 GbE recommended) | 10 GbE |
| Storage | NVMe SSD 1 TB+ (recordings extra) | NVMe SSD 1 TB+ (recordings extra) | NVMe SSD 2 TB+ (recordings extra) |
| Measured | 39 in one room, CPU 15% 40 across 10 rooms, CPU 37% | 40 in one room, CPU 16% 40 across 10 rooms, CPU 20% | — not yet measured |
| Recommended planning capacity 1080p / 30 fps | 55 in one room 70 across up to 5 rooms | 60 in one room 90 across 8+ rooms | 70 in one room 125 across roomsEstimated — not yet measured |
| Simultaneous rooms | 1–5 recommended (all CPU-encoded) | 8 GPU-encoded; more fall back to CPU automatically | As Standard |
| Voice-only participants | 500+ | 500+ (measured: 400 at CPU 2.5%) | 500+ |
| AI translation and captions | — | Included | Included |
How we plan: a 60% CPU safety line, then about 30% more headroom because real camera video costs more to process than test patterns. Without a GPU, ten simultaneous rooms showed video jitter, so Entry is recommended for up to five. High-performance figures are estimated from roughly 1.4× multi-core CPU performance; actual capacity is confirmed by a PoC before purchase. Need more? Add servers and distribute rooms across them.
Real browsers, real video, every stream checked
End-to-end path
Every simulated participant signed in, joined a room and sent and received 1080p video over the same encrypted path a real user takes.
Quality per stream
Received frame rate, packet loss, latency, freezes and disconnects measured for every participant — not just “connected or not”.
Server monitored
CPU, memory, GPU and network logged every five seconds, with headroom kept on the test clients so they never skewed the result.
From the server room to the field
Fully offline
No internet connection needed for any function, including AI translation. Verified October 2026: a complete demonstration on an isolated network with only a local router.
Portable kit
One server and one router in a case. Power up, connect phones and laptops to the local network, and the command picture is ready. TAK map tiles are cached in advance.
Field connectivity
Pilots and remote teams connect over a mesh VPN, a 5G private network, tactical routers or private 5G — only the network layer changes.
Modular
The drone gateway and the H.323 gateway are separate modules. Start and stop them independently; a disabled module has no effect on the core system.
Scale out
Distribute rooms across servers for large sites and regional networks.
Coexists with other systems
Runs in containers alongside your other on-premises services without conflicts.
A predictable path
- Live demo60–90 minutes, online or at your site, fully offline if you prefer.
- Technical workshopArchitecture, integration points and customization with your engineers.
- PoCOn your network, with your devices — TAK server, phones, radios, drones.
- DeploymentInstallation, layouts, users and training, delivered with your integrator.
See it live — on our hardware or yours
A 60–90 minute demo: command-center layout, a drone on the TAK map, radio translation and the phone system, shown end to end. Online, or at your site fully offline.