Let me start with a simple truth — drowsy driving kills. Every year, thousands of people lose their lives because a driver fell asleep at the wheel. Traditional drowsiness detection systems try to solve this problem, but they have a fundamental flaw — they treat every driver the same way. Fixed thresholds for eye closure and yawning simply do not work for everyone. A driver with naturally small eyes gets false alerts constantly. Another driver who is genuinely drowsy but has wide eyes slips through. In this paper, we present a completely different ap- proach using Generative AI, specifically a Conditional Generative Adversarial Network (cGAN). Our system learns each driver’s unique alert face during the first few minutes of driving. Then it continuously compares their real-time face against a personalized baseline. If something looks off — even subtle drowsiness before the eyes fully close — the system triggers an alert. We trained and tested this system on the NTHU Drowsy Driver Dataset, and the results were impressive — 94.8% detection accuracy with only 2.1% false positives. The system runs in under 150 milliseconds per frame on standard hardware. This is not just another drowsiness detector. This is a system that actually adapts to the person behind the wheel.