Plan your rollout
FloFrame sizer
Will your cameras work, and which edge box runs them?
First your cameras and VMS: which brands connect, and any caveats. Then a GPU, your cameras and their use cases, per camera group if you need to. The spec and its headroom update as you go.
Your cameras and VMS
First, will they work? FloFrame pulls a parallel stream, so your recording and operators are untouched.
The VMS or recorder your cameras feed today. Support: fully supported.
What we need from a camera stream
- ONVIF Profile S or a plain RTSP URL, H.264 or H.265.
- 720p or better on the analysed stream. A sub-stream is often enough.
- Analog cameras connect through the per-channel RTSP streams of a DVR or encoder.
- Video is analysed on site. Only events and sealed evidence leave the box.
- A brand that isn't listed almost certainly works too, if it speaks ONVIF or RTSP. Ask us to confirm it.
Cameras and streams
What the box has to decode: how many cameras, at what resolution and frame rate.
Decode is a fixed cost, so load grows more slowly than the frame rate.
Use cases
Use cases don't all cost the same: each carries a processing weight the model accounts for.
GPU
Any current NVIDIA card. The reference build ships with an RTX 4070.
12 GB VRAM · 200 W TDP · 1.00× the reference throughput · about 42 stream-inference units
Simulated utilization
Near capacity One FloFrame with the GeForce RTX 4070 carries 12 cameras. 36 use-case assignments, 36 of 42 units.
Your cameras and VMS
Fully compatible
3 of 3 brands fully supported Everything you chose exposes standard streams. No new cameras.
No recording platform needed: FloFrame connects to each camera directly over ONVIF or RTSP. Streams are pulled over ONVIF Profile S / RTSP (H.264 or H.265).
Brand by brand
- No VMS, direct to camerasPlatformSupported
- HikvisionSupported
- DahuaSupported
- Axis CommunicationsSupported
The line marks the validated limit (100%).
Recommended FloFrame spec
- GPU
- NVIDIA GeForce RTX 4070 (12 GB)
- CPU
- Intel Core i7 (14th gen) / Core Ultra 7
- Memory
- 32 GB DDR5
- Storage
- 512 GB NVMe SSD
- Power supply
- 550 W (80+ Gold)
- Network
- 1 GbE
Send this spec to our team
We confirm your camera models and VMS version, validate the configuration and reply with a quote-ready spec.
Assumptions behind this sizing
- Reference build: Intel Core i7 (14th gen), 32 GB RAM, 1 TB NVMe and an RTX 4070, validated at about 12 cameras with 3 to 4 use cases each (42 weighted stream-inference units).
- Use-case weights: light ×0.7, standard ×1.0, heavy ×1.3, intensive ×1.6, from model complexity.
- Resolution and analysed frame rate scale both compute and decode load.
- Storage covers the system, models and sealed event clips only (about 0.5 GB per camera per day at 1080p). Continuous recording stays on your VMS or NVR.
- 100% is the validated limit of the selected GPU. Up to 85% leaves comfortable headroom; above 100% needs a bigger GPU or another box.
- GPU factors are conservative planning estimates relative to the RTX 4070. Our solutions team validates every spec before an order.
See it on your own cameras.
A demo on your use cases, or a pilot with success criteria agreed up front.