Awesome. How does the audio analysis work? Is it all happening client side?
Also, am I right in saying that you did your own fft rather than use the built in analyser? I tried to do one with the built in analyser a while back and had mixed results.