Troubleshooting

If the launcher fails to open or closes immediately, work through these in order: 1. Make sure you're on a supported version of Windows and that Windows is up to date. 2. Right-click the launcher and choose Run as administrator once — some first-run setup needs elevated permissions. 3. Confirm the app wasn't quarantined by your antivirus (see the antivirus item below) and that all files finished downloading. 4. Install the latest Microsoft Visual C++ Redistributable and current GPU drivers, then reboot. 5. If it still won't launch, download a fresh copy from the Downloads page and reinstall over the top. Still stuck? Post the exact error in Discord or email support@sentinelcorevision.com.

Most activation errors come down to a mistyped key or a blocked connection. 1. Copy the key straight from your order email or the account page rather than typing it — avoid trailing spaces, and note the launcher formats it as JAYY-XXXX-XXXX-XXXX for you. 2. Confirm the PC is online; the launcher makes a short check to the license server to confirm your subscription is active. 3. Temporarily allow the app through any firewall, VPN, or antivirus that could be blocking that check, then try again. 4. If the key was recently purchased, give it a moment and retry. 5. An "already in use" or "machine limit reached" message means the key is activated on its allowed number of machines — see the next item to move it. If the key is genuinely rejected, contact support@sentinelcorevision.com with the key and we'll check its status.

Your license activates on a limited number of machines and is tied to each machine's hardware id, so a new PC — or a Windows reinstall on the same PC — can look like a new machine. To move it cleanly: on the old machine, open the launcher and use Deactivate to release that binding before you retire it. Then enter the same key on the new machine to activate. If the old machine is gone (dead drive, wiped, or you reinstalled without deactivating first), email support@sentinelcorevision.com from your purchase address and we'll release the old binding for you. Keys are meant to be moved, not shared across machines at once.

The dashboard is served locally by the engine, so it only appears once the engine is running. 1. Start the engine from the launcher first, then open http://localhost:5000 in your browser. 2. Use http://localhost:5000 or http://127.0.0.1:5000 exactly — not https://, and not an external IP. 3. If the page won't load, another program may be using port 5000; close it, or restart the engine and try again. 4. Allow the app through your firewall for local/private connections if prompted. 5. Try a different browser or a private window to rule out a cached page or extension. If it still won't open, capture what your browser shows and reach out via Discord.

A black or empty preview usually means the engine isn't receiving a video feed yet. 1. In Setup, confirm the correct capture source is selected — a capture card if your feed comes from a console or second machine, or desktop capture if you're capturing the same PC. 2. Make sure the source is actually producing video: the capture card is connected and powered, or the target window is open and on screen. 3. Close any other application that has exclusive control of the same capture device, then re-select it. 4. On laptops with two GPUs, desktop capture must run on the adapter that owns the display; if the preview is black, switch the capture source and re-check. 5. Update your GPU and capture-device drivers and reconnect the device. If the preview stays black, note your capture hardware and share it with us so we can help match the right mode.

If the engine is detecting objects in the preview but no output is reaching your device, check the output path. 1. Confirm the supported input device is plugged in, powered, and recognized by Windows before you start the engine. 2. In Setup, make sure the correct output device is selected and that output is enabled — nothing moves until output is turned on. 3. Verify the device's output mode matches what the engine expects; the wrong mode is the single most common reason output is sent but nothing responds. 4. Re-seat the cable and try a different USB port, ideally a direct port rather than a hub. 5. Restart the engine after reconnecting so it re-detects the device. If the preview shows activity but the device stays silent, share your device details in Discord and we'll walk through the output mode with you.

Performance depends on your hardware, GPU, drivers, model, configuration, and environment — there are a few reliable levers. 1. Provider: confirm the engine picked a hardware-accelerated provider. NVIDIA cards use CUDA/TensorRT and other modern GPUs use DirectML via ONNX Runtime, with automatic fallback if a provider is missing — but a missing driver can drop you to a slower path, so update your GPU drivers. 2. Capture resolution: a lower capture resolution means fewer pixels to process each frame; drop it if your source is very high resolution. 3. Model size: a smaller, speed-tuned model runs faster than a large accuracy-tuned one — switch models from the dashboard and compare. 4. Close other GPU-heavy applications so the engine isn't competing for the card. 5. Keep Windows and drivers current. Balance these three — provider, resolution, and model — to find the sweet spot for your rig.

Models load from the dashboard and can hot-swap without restarting the engine, but the file has to be in a format the engine understands. 1. Use a supported model file — the engine runs on ONNX Runtime, so the model must be a valid, non-corrupted file in the expected format. 2. Re-download the model if it may have been interrupted, and confirm the file size looks right. 3. Place it where the dashboard's model manager expects, then select it from the list. 4. After loading, check that the model's classes match what you're detecting — an unexpected class filter can make a working model look like it "sees nothing." 5. If a model loads but performs poorly, try a different size (see the performance item above). If the engine reports a format error, note the exact message and reach out via support@sentinelcorevision.com.

Some antivirus tools flag new or less-common desktop apps by default, which can quarantine files or block the license check. If you trust your download, you can allow-list it. 1. Always download from the official Downloads page so you know the file is legitimate. 2. Open your antivirus, find the quarantine or protection history, and restore the flagged file. 3. Add an exclusion for the app's install folder so it isn't re-flagged, and allow it through the firewall so the license check can reach the server. 4. Reinstall from a fresh download if any files were deleted, then reboot. 5. If you're unsure about a file, don't guess — ask in Discord or email us first and we'll confirm what's expected.

If none of the above sorts it, we're happy to help. The fastest route is our community Discord, where you can ask and search past answers. For account, licensing, or anything you'd rather keep private, email support@sentinelcorevision.com from your purchase address. It helps to include your Windows version, GPU, what you were doing, and the exact error text or a screenshot. You can also browse the docs, the FAQ, the latest notes on the changelog, or open a ticket from the support page.