On-Demand AI Model Weights

Mission Control keeps its base installation package lightweight (~60 MB) by offering optional AI neural network weights on-demand. This modular design ensures users can download only the specific neural capabilities required for their gaming setup.

---

1. Available Model Weights

YOLOv8n Vision Model (6.2 MB)

• Architecture: YOLOv8 Nano (yolov8n.pt)
• Primary Function: Ultra-fast HUD target tracking & low-latency object detection.
• Inference Speed: ~2–5 ms per frame (GPU dependent).
• VRAM Footprint: Minimal (~150 MB VRAM overhead).
• Best For: Entry-level GPUs, budget setups, and competitive high-FPS gameplay requiring sub-5ms overlay tracking.

YOLOv8s High Precision (22.5 MB)

• Architecture: YOLOv8 Small (yolov8s.pt)
• Primary Function: High-accuracy neural tracking & complex scene understanding.
• Inference Speed: ~6–12 ms per frame.
• VRAM Footprint: ~400 MB VRAM overhead.
• Best For: High-end GPUs, 1440p / 4K monitors, and complex game scenes requiring precise detection of small UI elements and distant enemy outlines.

Whisper-Tiny Voice AI (39.0 MB)

• Architecture: OpenAI Whisper Tiny (whisper-tiny.pt)
• Primary Function: Offline neural Speech-to-Text (STT) & voice command engine.
• Processing Mode: 100% On-Device local processing.
• Best For: Hands-free voice commands, AI assistant voice interactions, and offline operational privacy with zero cloud API latency.

---

2. Technical Execution Pipeline

When a model weight is downloaded via the Vision or Voice dashboard:

1. The app fetches pre-trained neural weights directly from Mission Control's secure release mirror.
2. Weights are stored locally under %APPDATA%\MissionControl\models\ (Windows) or ~/.config/MissionControl/models/ (Linux).
3. If an NVIDIA GPU is active, TensorRT 10.x automatically compiles .pt weights into optimized FP16/INT8 .engine binaries tailored to your GPU's exact compute architecture.

---

3. Managing Model Weights

• Installation: Open Vision AI or Agent Settings in Mission Control and click Download next to the desired model card.
• Storage Location:
  • Windows: %APPDATA%\MissionControl\models\
  • Linux: ~/.config/MissionControl/models/
• Removal: Deleting model weight files from disk automatically reverts the UI state back to "Not Installed" without affecting base application functions.