Troubleshooting

Why this happens: Same-PC desktop screen capture (Dxgi) is disabled in customer builds by design. On older builds, clicking START before plugging in a capture card could fall back to your Windows desktop monitor and trigger exit 1. How to fix it in 60 seconds: 1. Update to the latest build from Downloads or run /download in (the current Studio build opens a guided Setup checklist instead of crashing). 2. Connect your USB 3.0 HDMI Capture Card to a blue/teal USB 3.0 port. 3. Click 📖 SETUP GUIDE in the Studio top bar for the full cable map, select your source, and press ⚡ START. See Connection Paths in the Docs.

A black preview or NO FEED badge almost always comes down to one of four physical checks: 1. USB 3.0 Port: Plug your capture card directly into a motherboard blue/teal USB 3.0+ port (USB 2.0 ports cannot carry high-rate NV12 video). 2. PS5 HDCP Off: On PlayStation 5, go to Settings → System → HDMI → Enable HDCP and turn it Off — HDCP encryption forces a black screen on every capture card. 3. Close Exclusive Camera Apps: Close OBS Studio, Elgato 4K Capture Utility, Streamlabs, or Discord camera previews so Sentinel Studio owns the device exclusively, then click Scan Capture Cards in Studio.

If Sentinel Studio draws detection boxes and shows LOCK in the viewport, vision is working — check your physical output path: 1. Three Titan Two Cables: Confirm PROG goes to your Windows PC USB, OUTPUT goes to your Console USB, and an official controller is plugged into INPUT-A using data-capable USB cables (not charge-only cables). 2. Slot 1 Active: Press the button on your Titan Two until its LED shows Slot 1 — Slot 0 passes raw stick input only and ignores AI commands. 3. Studio Output Enabled: Confirm Titan Two is selected as the Output device in Sentinel Studio and that output is turned on. 4. ADS Gate: Hold LT / L2 in the firing range if your profile is configured to engage on ADS.

When inference latency (INF in the top bar) sits around 30 ms (~25 FPS), perception delay makes any control loop overshoot and orbit the target. 1. Switch NVIDIA GPUs to TensorRT or CUDA: Look at the provider dropdown in the Sentinel Studio top bar. If you have an NVIDIA RTX GPU and it says DirectML (Any GPU), switch it to TensorRT or CUDA and accept the one-click NVIDIA acceleration pack download — inference drops from ~32 ms down to ~4–8 ms at 60–120+ FPS. 2. Capture Resolution: Use clean 1080p60 or 1080p120 capture rather than 4K input. 3. Close GPU Hogs: Close OBS, wallpaper shaders, or browser streams while playing. 4. Tune Gently: Once INF is low, test defaults in the firing range first and adjust Tracking Sensitivity in small 10–15% steps.

You do not need to open a web browser or visit localhost:5000 anymore. Every live preview, capture selector, GPU provider switch, model marketplace card, and tuning slider is built directly into the native Sentinel Studio desktop app (SentinelStudio.exe). Click 📖 SETUP GUIDE in the Studio top bar anytime for the built-in quickstart walkthrough.

Work through these in order: 1. Extract the full ZIP first: Right-click the downloaded archive and Extract All into C:\SentinelVision — never launch SentinelStudio.exe from inside an unextracted .zip, OneDrive, or temporary folder, and keep all extracted files together. 2. Check Windows Security: Confirm Windows Defender didn't quarantine a DLL during extraction (see the antivirus item below). 3. Update GPU Drivers & VC++ Runtime: Install current NVIDIA/AMD/Intel graphics drivers and the Microsoft Visual C++ Redistributable (x64), then reboot. 4. Fresh Download: Claim a clean archive from Downloads.

1. Invalid Key: Copy your SCV-XXXX-XXXX-XXXX key directly from your account page. 2. Moving PCs / Device Limit Reached: Paid plans include self-serve hardware resets in your account page (up to 2 resets every 7 days) to clear old PC bindings after a Windows reinstall or PC upgrade. 3. Clock Skew Warning: Open Windows Date & Time settings, enable Set time automatically, click Sync now (or run w32tm /resync in Administrator PowerShell), and relaunch Studio.

1. One-Time TensorRT Compile (1–2 Minutes): The very first time you select TensorRT or load a new model on an NVIDIA GPU, TensorRT compiles a hardware-specific engine cache. Let it finish without closing Studio — every subsequent launch loads in seconds. 2. Model Download Interrupted: Open the Marketplace / Models view inside Sentinel Studio and re-download the model cleanly if a network drop corrupted the file. 3. Antivirus Lock: Make sure C:\SentinelVision isn't being locked mid-load by third-party antivirus software.

Freshly published native desktop builds can trigger heuristic false-positives before SmartScreen reputation builds up. 1. Verify the source: Only run builds downloaded from sentinelcorevision.com/downloads.html. 2. Restore & Exclude: Open Windows Security → Virus & threat protection → Protection history to restore any quarantined file, add C:\SentinelVision as a folder exclusion, and re-extract if needed.

Need help with your setup? Check our Setup Guide for wiring diagrams. For direct 1-on-1 assistance, reach out via our Contact Page or email [email protected] with your GPU model, capture card make and model, a safe Studio status-bar screenshot, and only the last 4 characters of your SCV- key.