For a long time I was bothered by how modern TVs still struggle with stable audio levels. Commercials are louder, dialogues are quieter, and sudden spikes can wake up half the house. Even with built‑in “volume leveling”, the problem persists — especially during ads and streaming.

After trying different solutions over the years, nothing worked reliably. So I built a small offline system that listens to TV audio in real time, detects sudden loudness spikes, and sends IR volume adjustments locally.

No cloud, no APIs, no smart‑home ecosystem — just local audio analysis + a simple IR blaster.

The hardest part wasn’t the code, but figuring out when audio truly requires intervention. That led me to design a small signal‑behavior model that looks at micro‑events, context and dynamics before deciding whether to react.

The prototype is now stable enough for daily use. Not perfect yet, but the foundation works.