Experiment #6: The Count

So this experiment hasn’t become all that I had hoped it would be just yet, but with failure being such an important part of experimentation (and eventual musical technology victory), I wanted to share it anyway.

What The Count ideally does, is it adds Count Von Count himself counting “1… 2… 3… 4…” perfectly and musically in time over any track you give it. After all, he really loves to count and I’m pretty sure he finds music a wonderful thing just aching to be counted.

In practice though, while The Count’s tempo is pretty dead on, the counts aren’t always falling exactly on the beats. As with my Reverse Click Track-er, I’m using detailed analysis of the track provided by EchoNest to know the exact points in time where the songs beats and bars start. Unfortunately, the information that I’m getting from EchoNest don’t seem to always line up as it should. Sometimes it’s dead on, sometimes it’s on the upbeat instead of the downbeat, aaaaaand¬†sometimes it’s not even close at all.

I’ve posted to EchoNest’s dev forum about the inconsistencies I’m seeing and I’m hoping they’ll be able to shed some light on how I can either make my app much more accurate or in a worst-case scenario, explain that this is currently the best that their musical brain can do. But based on other EchoNest experiments, like Tristan Jehan’s awesome The Swinger, it really seems like I should be able to get very accurate beat and bar results from EchoNest… Either way, I’ll definitely share what I find and will hopefully be able to have to a fully functional and much more accurate Count Von Count music machine.

Update: Tristan responded to my post and explained that certain mp3 decoders handle error correction differently, which affects the offset timing, and the next EchoNest update will remedy this issue a bit. I’m also going to see if there’s any consistency in the offset created by Flash’s mp3 decoding and if I can work around the issue at all…

You can try The Count for yourself below, or listen to the Count-i-fied “We R Who We R” by Ke$ha which was one of the tracks I found to line up pretty perfectly:

WeRWhoWeCount.mp3

  1. No trackbacks yet.