Back to Work

AI Game Coach

2026
Client / Project
AI Game Coach
Year
2026
Industry
Gaming AI / Computer Vision / Desktop Applications
Tech Stack
Python 3.10+, Ollama, MiniCPM-V, mss, Pillow, Tkinter, pyttsx3, PyInstaller

Key Features

Real-Time Screen Capture

Captures the primary display directly using mss with minimal latency and no game integration.

Local Vision AI Analysis

Processes screenshots using the MiniCPM-V vision model running entirely through Ollama.

Floating Coaching Overlay

Displays tactical advice in a draggable, borderless, always-on-top interface.

Voice Coaching

Reads AI-generated guidance aloud using Windows text-to-speech for hands-free assistance.

Game-Specific Profiles

Automatically adjusts coaching frequency and behavior for different game genres.

Session Logging

Records every scene analysis and coaching recommendation in structured JSON files.

Markdown Reports

Generates complete coaching timelines after each session for later review.

Offline Operation

Runs entirely on the local machine without requiring cloud services or internet connectivity.

Business Impact

Zero Privacy Concerns

All processing happens locally, ensuring that gameplay data never leaves the user’s PC.

No Performance Impact

The game process is never injected or modified, preserving game stability and performance.

Improved Player Learning

Continuous tactical feedback helps players identify mistakes and improve over time.

Cross-Game Scalability

The architecture supports multiple genres with customizable coaching behaviors.

Reviewable Progress

Session reports create a measurable record of player development.

Accessible Coaching

Voice feedback enables players to receive guidance without taking their eyes off gameplay.

Ready to build something great?

Tell us your idea. We'll tell you how to make it real.

Start a Conversation