Reverse engineered prompt

Build me a simple Python tool that can take a road traffic video and estimate how fast each vehicle is moving. I want it to detect cars and other vehicles frame by frame, track them across the video, draw boxes and speed labels on screen, and save an annotated output video. It should work with common video files like mp4, avi, and mov, and it would be great if it runs right away on the included sample video as well as my own clips.

Please keep setup easy with a clear main entry point and a requirements file. Add a basic calibration option so I can tune the speed estimates using something like real world distance or pixels to meters, since different camera angles will change accuracy. If there is already a plate detection model included, feel free to wire that in as an optional extra, but the main goal is reliable vehicle speed estimation from video.

Make it practical, readable, and easy to run. Look up current docs online if you need to.

Want more depth? Deep Reverse