Timed to release right after Super Bowl 2024, we listened to 138 National Anthem performances and compared each to the standard melody to see which artists deviated the most.
First, we collected performances of the “Star Spangled Banner” across major sporting events (Super Bowl, NBA AllStar Game, MLB World Series, etc.) and political events (Inaguration, Democratic/Republican National Conventions, July 4th celebrations, etc.). We trimmed the clips, isolated the vocals from the rest of the audio, and extracted the pitch and notes from each rendition.
We then used the Dynamic Time Warping algorithm to measure the amount a performer deviates from the standard, strictly melodically. With these numbers, we were able to find the biggest diva (or deviator!) for each phase and for the full anthem.
We built an interactive webpage that walked the audience through each phrase and then allowed them to listen to the deviations (and the divas!) themselves in a giant heatmap. We also partnered with Switched on Pop to record a podcast episode and made a video/audio mashup anthem of the biggest diva from each of the 16 phrases.
Totally fell into the rabbit hole and found myself forgetting I was a judge and instead enjoying playing with listening and comparing the different versions. The concept was ingenious and the execution made it even better. The explanation and dataviz was immersive and the way it used AI to calculate and differentiate shows real innovation. Totally hooked.
