Torqdrop

TorqDrop is a mechanic's eye and ear in your pocket. Most people can't tell what's wrong with their car, and when they buy a used one they mostly hope for the best. Mechanics can tell, but they're rarely there when you need them, so we built an Android app that gives everyone that first expert opinion. The main feature is an AI engine sound check. You film your engine for 20 seconds while the app guides you to hold steady, give one gentle rev, and let it settle back to idle. Meta's Muse Spark model watches the video, listens to the engine and reads live data from the car, all in one request. It tells you what it heard, such as a knock, tick, squeal or misfire, marks when each sound happened on an rpm timeline, notes any smoke or leaks it can see, and suggests what to check next. With a cheap Bluetooth OBD-II adapter, TorqDrop also shows live engine data, explains trouble codes in plain language, gives the car a health score, logs trips automatically and keeps a garage with service reminders. It recognises the car even when it doesn't report a VIN. For buyers and inspectors, there's a guided 200-point used-car inspection that shows where to look and what to photograph. AI reviews every photo for leaks, damage and wear, and reads the RC and insurance papers. The inspector confirms or dismisses each finding, then signs off a report they can share as a link or PDF. We built TorqDrop with Expo and React Native. A custom Bluetooth OBD-II stack runs on the phone, and Firebase handles sign-in, storage and Cloud Functions. All AI calls go through the Meta Model API on the server. Because a wrong answer about a car can cost money or be unsafe, safety warnings come from fixed rules that the AI can never override, every finding shows its evidence and confidence, and nothing reaches a report unless a person confirms it. Engine clips are deleted right after analysis.