Documentation

Documentation

Everything you need to install, activate, and get the most out of Sentinel Core Vision.

Sentinel Core Vision is a real-time computer-vision desktop application for Windows, paired with a local dashboard you open in your browser at http://localhost:5000. This guide walks you from a fresh install through activation, day-to-day use, and tuning for your hardware.

Performance depends on your hardware, GPU, drivers, model, configuration, and environment. The requirements and figures below are guidance, not guarantees — your results will vary from machine to machine.

On this page

1. System requirements

Sentinel Core Vision runs on Windows and uses your GPU for real-time inference. Most modern PCs meet the minimum; a recent NVIDIA GPU gives the lowest-latency path but is optional. Below is a minimum-versus-recommended breakdown.

Minimum

  • Operating system: Windows 10 or 11, 64-bit.
  • CPU: A modern quad-core processor.
  • Memory: 8 GB RAM.
  • Storage: Around 2 GB free for the application and models; more if you keep multiple models.
  • GPU: A DirectX 12 / DirectML-capable GPU (most recent AMD, Intel, and NVIDIA graphics qualify).
  • Internet: A connection is required to activate your license and to check that it stays active.

Recommended

  • Operating system: Windows 11, 64-bit, fully updated.
  • CPU: A recent six-core (or better) processor.
  • Memory: 16 GB RAM.
  • Storage: An SSD with several GB free for models and profiles.
  • GPU: A dedicated NVIDIA GPU for the CUDA / TensorRT accelerated path, with current drivers. AMD and Intel GPUs run well through DirectML.
  • Internet: A stable connection for activation and updates.

See the Downloads page for the current build and any version-specific notes.

2. Installation

Getting set up takes a few minutes. Follow these steps in order.

  1. Download the launcher. Grab the latest installer from the Downloads page.
  2. Run it. Launch the downloaded file and follow the prompts. Windows may ask you to confirm before it opens — allow it to continue.
  3. Paste your license key. When the launcher opens, enter the license key from your account. Keys use the JAYY-XXXX-XXXX-XXXX format and activate in seconds. You can find or manage your key from your account at any time.
  4. First-run setup. On first launch, the app checks that your system is ready — GPU provider, capture source, and output device — and points out anything that needs attention. Once the checks pass, the engine starts and the dashboard becomes available.

If the launcher reports a missing component or a failed check, the FAQ covers the most common first-run questions, and you can always reach us on Discord.

3. Using the dashboard

The dashboard is where you configure and observe the engine. It runs locally — nothing about your setup leaves your PC through it.

  1. Open the dashboard. With the engine running, browse to http://localhost:5000 in any modern browser on the same PC.
  2. Pick a capture source. Choose where the video feed comes from — a capture device or your desktop. The app auto-selects a sensible capture mode, and you can override it.
  3. Choose an engine. Select the processing engine that fits your use case. Each engine has its own settings and can be switched without restarting.
  4. Tune settings. Start with the simple sliders. Advanced and developer surfaces are one tap away when you want finer control, and every change applies live — there are no files to edit by hand.
  5. Watch the Vision view. The Vision view shows what the engine sees in real time, so you can confirm detections and the effect of each change as you go.

Output from the engine is sent to a supported input device you select during setup. You can save and load full profiles for different sources or configurations, and export or import a profile to back it up or move it to another machine.

4. Updating

Updates are delivered through the launcher. When a new build is available, download the latest installer from the Downloads page and run it over your existing installation — your license, profiles, and settings are preserved. Close the engine before updating so files aren't in use. Version-by-version notes are published on the Downloads page so you can see what changed before you update.

5. Uninstalling

To remove Sentinel Core Vision, close the engine and the launcher, then uninstall it the standard way from Windows Settings → Apps → Installed apps (or Programs and Features on Windows 10). If you plan to move your license to another PC, deactivate it first (see below) so the machine slot is freed. Uninstalling the app does not, on its own, release your activation.

6. Licensing & activation

What a key is. Your license key unlocks the full application. You enter it once in the launcher and it activates against your account. Every plan unlocks the complete engine and all updates.

Machine / hardware limit. A key activates on a limited number of machines. Each activation is tied to a hardware id (HWID) for the PC it runs on, which is how the limit is enforced and how the key stays yours rather than shareable.

Moving to a new PC. To move your license, deactivate it on the old machine to free the slot, then activate on the new one. If you no longer have access to the old PC — for example it's gone or won't boot — contact support and we'll release the old binding for you.

Reinstalling Windows. A clean Windows reinstall or major hardware change can produce a new hardware id, which may look like a new machine to the license server. If activation is blocked after a reinstall, reach out and we'll clear the previous binding so you can activate again.

Offline behaviour. Activation requires an internet connection, and the app checks periodically that your license is still active. Brief interruptions are tolerated, but a machine that stays offline for an extended period may be unable to confirm the license and will ask you to reconnect. Keep the PC online for activation and normal use.

You can view your key and manage activations from your account. Billing questions are answered on the Pricing page and in the FAQ.

7. Performance tips

Real-time performance is a balance between accuracy and speed, and the right settings depend on your hardware. A few levers make the biggest difference:

  • Provider selection. On NVIDIA hardware, the CUDA / TensorRT path is the fastest. AMD and Intel GPUs use DirectML. ONNX Runtime backs both, and the app auto-selects the fastest available provider with a safe fallback if one is missing — so you rarely need to choose manually, but you can.
  • Model choice. Smaller models run faster and lighter; larger models can be more accurate at a higher cost. Because models hot-swap from the dashboard, you can compare them live and keep whichever gives the best balance on your machine.
  • Capture resolution. Capturing and processing a lower resolution reduces load and latency. If you're GPU- or CPU-bound, step the capture resolution down before anything else — it's usually the largest single win.
  • Keep drivers current. Up-to-date GPU drivers often improve both stability and throughput, especially on the NVIDIA path.
  • Reduce background load. Close other GPU-heavy applications so the engine has headroom.

Tune one thing at a time and watch the Vision view — it's the quickest way to see the effect of each change.

8. Getting help

If something isn't working the way you expect:

  • Start with Troubleshooting — quick fixes for the most common issues (activation, capture, output, and performance).
  • Check the FAQ — it covers setup, hardware, licensing, and day-to-day questions.
  • Browse the Support center for every guide and channel in one place.
  • Ask on our Discord — it's the fastest way to an answer, and the community and team are active there.
  • Email [email protected] or use the contact page for account, billing, or license issues.

When you reach out, tell us your GPU, Windows version, the capture source and output device you're using, and what you were doing when the problem appeared — it helps us get you a precise answer faster.