Why would it need that much training data? It takes me 3 or 4 videos to pick up the pattern. It’s super patternistic, and I almost have a hard time believing it would require that much training data. It’s almost like the Shazam app, I think you can literally compare the graph of the song and identify it quickly.