Also capture the audio sample from the same record alongside the video. Repeat with lots of records. Train a model to predict the audio from the video. Voila! Play records with just the camera “stylus”...