Audio Spike Detector

As the social chair of the northeastern women's ultimate frisbee team, I spend a lot of time cutting highlight reels from full game recordings. Most of that time is spent going through long footage trying to find the best plays. I wanted a faster way to identify key moments without relying on heavy editing software or AI tools. This project explores whether audio analysis can replace that manual search. The main question I explored was whether audio alone could be used to point me to the most important parts of a video. The final result is a browser-based tool built with JavaScript and p5.js. Users can upload a video, and the tool analyzes the audio using the Web Audio API to map volume over time. This data is visualized as a waveform, where peaks represent louder moments. Users can click into those peaks, adjust how much time before and after each spike they want to keep, and build a timeline of clips. The tool also includes a preview mode that plays selected segments back-to-back, along with an export feature that records and downloads the final video. One of my biggest takeaways is how relatively simple the audio processing side of the code actually is, and how effective it is at predicting potential highlight moments. While it is not perfect, it significantly reduces the time needed to find clips. I also learned a lot about user experience and interface design through multiple iterations focused on improving usability and workflow.