Choosing an inference mode

The INFERENCE BACKEND panel in the middle decides which Tensor Dynamics engine sees the frames. You have two options:

Option A — Tensor Cloud (easiest)

• Uses the Tensor Dynamics managed API. No GPU, no server required.
• Click "Tensor Cloud", paste your API key into the input box.
• Downsides: frames leave your network, internet required, usage costs may apply.

Option B — Tensor Engine (on-device or remote)

• Runs Tensor Dynamics Vision on your hardware. Footage stays on-prem.
• Needs a compatible GPU (on-device) or a remote GPU box.

Fastest path: Tensor Cloud → paste API key → Test → green OK → done.